go/src/cmd/compile/internal/ssa/_gen/generic.rules

2777 lines
135 KiB
Text
Raw Normal View History

// Copyright 2015 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Simplifications that apply to all backend architectures. As an example, this
// Go source code
//
// y := 0 * x
//
// can be translated into y := 0 without losing any information, which saves a
// pointless multiplication instruction. Other .rules files in this directory
// (for example AMD64.rules) contain rules specific to the architecture in the
// filename. The rules here apply to every architecture.
//
// The code for parsing this file lives in rulegen.go; this file generates
// ssa/rewritegeneric.go.
// values are specified using the following format:
// (op <type> [auxint] {aux} arg0 arg1 ...)
// the type, aux, and auxint fields are optional
// on the matching side
// - the type, aux, and auxint fields must match if they are specified.
// - the first occurrence of a variable defines that variable. Subsequent
// uses must match (be == to) the first use.
// - v is defined to be the value matched.
// - an additional conditional can be provided after the match pattern with "&&".
// on the generated side
// - the type of the top-level expression is the same as the one on the left-hand side.
// - the type of any subexpressions must be specified explicitly (or
// be specified in the op's type field).
// - auxint will be 0 if not specified.
// - aux will be nil if not specified.
// blocks are specified using the following format:
// (kind controlvalue succ0 succ1 ...)
// controlvalue must be "nil" or a value expression
// succ* fields must be variables
// For now, the generated successors must be a permutation of the matched successors.
// constant folding
(Trunc16to8 (Const16 [c])) => (Const8 [int8(c)])
(Trunc32to8 (Const32 [c])) => (Const8 [int8(c)])
(Trunc32to16 (Const32 [c])) => (Const16 [int16(c)])
(Trunc64to8 (Const64 [c])) => (Const8 [int8(c)])
(Trunc64to16 (Const64 [c])) => (Const16 [int16(c)])
(Trunc64to32 (Const64 [c])) => (Const32 [int32(c)])
(Cvt64Fto32F (Const64F [c])) => (Const32F [float32(c)])
(Cvt32Fto64F (Const32F [c])) => (Const64F [float64(c)])
(Cvt32to32F (Const32 [c])) => (Const32F [float32(c)])
(Cvt32to64F (Const32 [c])) => (Const64F [float64(c)])
(Cvt64to32F (Const64 [c])) => (Const32F [float32(c)])
(Cvt64to64F (Const64 [c])) => (Const64F [float64(c)])
(Cvt32Fto32 (Const32F [c])) => (Const32 [int32(c)])
(Cvt32Fto64 (Const32F [c])) => (Const64 [int64(c)])
(Cvt64Fto32 (Const64F [c])) => (Const32 [int32(c)])
(Cvt64Fto64 (Const64F [c])) => (Const64 [int64(c)])
(Round32F x:(Const32F)) => x
(Round64F x:(Const64F)) => x
(CvtBoolToUint8 (ConstBool [false])) => (Const8 [0])
(CvtBoolToUint8 (ConstBool [true])) => (Const8 [1])
(BitLen64 (Const64 [c])) && config.PtrSize == 8 => (Const64 [int64(bits.Len64(uint64(c)))])
(BitLen32 (Const32 [c])) && config.PtrSize == 8 => (Const64 [int64(bits.Len32(uint32(c)))])
(BitLen16 (Const16 [c])) && config.PtrSize == 8 => (Const64 [int64(bits.Len16(uint16(c)))])
(BitLen8 (Const8 [c])) && config.PtrSize == 8 => (Const64 [int64(bits.Len8(uint8(c)))])
(BitLen64 (Const64 [c])) && config.PtrSize == 4 => (Const32 [int32(bits.Len64(uint64(c)))])
(BitLen32 (Const32 [c])) && config.PtrSize == 4 => (Const32 [int32(bits.Len32(uint32(c)))])
(BitLen16 (Const16 [c])) && config.PtrSize == 4 => (Const32 [int32(bits.Len16(uint16(c)))])
(BitLen8 (Const8 [c])) && config.PtrSize == 4 => (Const32 [int32(bits.Len8(uint8(c)))])
(Trunc16to8 (ZeroExt8to16 x)) => x
(Trunc32to8 (ZeroExt8to32 x)) => x
(Trunc32to16 (ZeroExt8to32 x)) => (ZeroExt8to16 x)
(Trunc32to16 (ZeroExt16to32 x)) => x
(Trunc64to8 (ZeroExt8to64 x)) => x
(Trunc64to16 (ZeroExt8to64 x)) => (ZeroExt8to16 x)
(Trunc64to16 (ZeroExt16to64 x)) => x
(Trunc64to32 (ZeroExt8to64 x)) => (ZeroExt8to32 x)
(Trunc64to32 (ZeroExt16to64 x)) => (ZeroExt16to32 x)
(Trunc64to32 (ZeroExt32to64 x)) => x
(Trunc16to8 (SignExt8to16 x)) => x
(Trunc32to8 (SignExt8to32 x)) => x
(Trunc32to16 (SignExt8to32 x)) => (SignExt8to16 x)
(Trunc32to16 (SignExt16to32 x)) => x
(Trunc64to8 (SignExt8to64 x)) => x
(Trunc64to16 (SignExt8to64 x)) => (SignExt8to16 x)
(Trunc64to16 (SignExt16to64 x)) => x
(Trunc64to32 (SignExt8to64 x)) => (SignExt8to32 x)
(Trunc64to32 (SignExt16to64 x)) => (SignExt16to32 x)
(Trunc64to32 (SignExt32to64 x)) => x
(ZeroExt8to16 (Const8 [c])) => (Const16 [int16( uint8(c))])
(ZeroExt8to32 (Const8 [c])) => (Const32 [int32( uint8(c))])
(ZeroExt8to64 (Const8 [c])) => (Const64 [int64( uint8(c))])
(ZeroExt16to32 (Const16 [c])) => (Const32 [int32(uint16(c))])
(ZeroExt16to64 (Const16 [c])) => (Const64 [int64(uint16(c))])
(ZeroExt32to64 (Const32 [c])) => (Const64 [int64(uint32(c))])
(SignExt8to16 (Const8 [c])) => (Const16 [int16(c)])
(SignExt8to32 (Const8 [c])) => (Const32 [int32(c)])
(SignExt8to64 (Const8 [c])) => (Const64 [int64(c)])
(SignExt16to32 (Const16 [c])) => (Const32 [int32(c)])
(SignExt16to64 (Const16 [c])) => (Const64 [int64(c)])
(SignExt32to64 (Const32 [c])) => (Const64 [int64(c)])
(Neg8 (Const8 [c])) => (Const8 [-c])
(Neg16 (Const16 [c])) => (Const16 [-c])
(Neg32 (Const32 [c])) => (Const32 [-c])
(Neg64 (Const64 [c])) => (Const64 [-c])
(Neg32F (Const32F [c])) && c != 0 => (Const32F [-c])
(Neg64F (Const64F [c])) && c != 0 => (Const64F [-c])
(Add8 (Const8 [c]) (Const8 [d])) => (Const8 [c+d])
(Add16 (Const16 [c]) (Const16 [d])) => (Const16 [c+d])
(Add32 (Const32 [c]) (Const32 [d])) => (Const32 [c+d])
(Add64 (Const64 [c]) (Const64 [d])) => (Const64 [c+d])
(Add32F (Const32F [c]) (Const32F [d])) && c+d == c+d => (Const32F [c+d])
(Add64F (Const64F [c]) (Const64F [d])) && c+d == c+d => (Const64F [c+d])
(AddPtr <t> x (Const64 [c])) => (OffPtr <t> x [c])
(AddPtr <t> x (Const32 [c])) => (OffPtr <t> x [int64(c)])
(Sub8 (Const8 [c]) (Const8 [d])) => (Const8 [c-d])
(Sub16 (Const16 [c]) (Const16 [d])) => (Const16 [c-d])
(Sub32 (Const32 [c]) (Const32 [d])) => (Const32 [c-d])
(Sub64 (Const64 [c]) (Const64 [d])) => (Const64 [c-d])
(Sub32F (Const32F [c]) (Const32F [d])) && c-d == c-d => (Const32F [c-d])
(Sub64F (Const64F [c]) (Const64F [d])) && c-d == c-d => (Const64F [c-d])
(Mul8 (Const8 [c]) (Const8 [d])) => (Const8 [c*d])
(Mul16 (Const16 [c]) (Const16 [d])) => (Const16 [c*d])
(Mul32 (Const32 [c]) (Const32 [d])) => (Const32 [c*d])
(Mul64 (Const64 [c]) (Const64 [d])) => (Const64 [c*d])
(Mul32F (Const32F [c]) (Const32F [d])) && c*d == c*d => (Const32F [c*d])
(Mul64F (Const64F [c]) (Const64F [d])) && c*d == c*d => (Const64F [c*d])
(And8 (Const8 [c]) (Const8 [d])) => (Const8 [c&d])
(And16 (Const16 [c]) (Const16 [d])) => (Const16 [c&d])
(And32 (Const32 [c]) (Const32 [d])) => (Const32 [c&d])
(And64 (Const64 [c]) (Const64 [d])) => (Const64 [c&d])
(Or8 (Const8 [c]) (Const8 [d])) => (Const8 [c|d])
(Or16 (Const16 [c]) (Const16 [d])) => (Const16 [c|d])
(Or32 (Const32 [c]) (Const32 [d])) => (Const32 [c|d])
(Or64 (Const64 [c]) (Const64 [d])) => (Const64 [c|d])
(Xor8 (Const8 [c]) (Const8 [d])) => (Const8 [c^d])
(Xor16 (Const16 [c]) (Const16 [d])) => (Const16 [c^d])
(Xor32 (Const32 [c]) (Const32 [d])) => (Const32 [c^d])
(Xor64 (Const64 [c]) (Const64 [d])) => (Const64 [c^d])
(Ctz64 (Const64 [c])) && config.PtrSize == 4 => (Const32 [int32(ntz64(c))])
(Ctz32 (Const32 [c])) && config.PtrSize == 4 => (Const32 [int32(ntz32(c))])
(Ctz16 (Const16 [c])) && config.PtrSize == 4 => (Const32 [int32(ntz16(c))])
(Ctz8 (Const8 [c])) && config.PtrSize == 4 => (Const32 [int32(ntz8(c))])
(Ctz64 (Const64 [c])) && config.PtrSize == 8 => (Const64 [int64(ntz64(c))])
(Ctz32 (Const32 [c])) && config.PtrSize == 8 => (Const64 [int64(ntz32(c))])
(Ctz16 (Const16 [c])) && config.PtrSize == 8 => (Const64 [int64(ntz16(c))])
(Ctz8 (Const8 [c])) && config.PtrSize == 8 => (Const64 [int64(ntz8(c))])
(Div8 (Const8 [c]) (Const8 [d])) && d != 0 => (Const8 [c/d])
(Div16 (Const16 [c]) (Const16 [d])) && d != 0 => (Const16 [c/d])
(Div32 (Const32 [c]) (Const32 [d])) && d != 0 => (Const32 [c/d])
(Div64 (Const64 [c]) (Const64 [d])) && d != 0 => (Const64 [c/d])
(Div8u (Const8 [c]) (Const8 [d])) && d != 0 => (Const8 [int8(uint8(c)/uint8(d))])
(Div16u (Const16 [c]) (Const16 [d])) && d != 0 => (Const16 [int16(uint16(c)/uint16(d))])
(Div32u (Const32 [c]) (Const32 [d])) && d != 0 => (Const32 [int32(uint32(c)/uint32(d))])
(Div64u (Const64 [c]) (Const64 [d])) && d != 0 => (Const64 [int64(uint64(c)/uint64(d))])
(Div32F (Const32F [c]) (Const32F [d])) && c/d == c/d => (Const32F [c/d])
(Div64F (Const64F [c]) (Const64F [d])) && c/d == c/d => (Const64F [c/d])
(Select0 (Div128u (Const64 [0]) lo y)) => (Div64u lo y)
(Select1 (Div128u (Const64 [0]) lo y)) => (Mod64u lo y)
(Not (ConstBool [c])) => (ConstBool [!c])
(Floor (Const64F [c])) => (Const64F [math.Floor(c)])
(Ceil (Const64F [c])) => (Const64F [math.Ceil(c)])
(Trunc (Const64F [c])) => (Const64F [math.Trunc(c)])
(RoundToEven (Const64F [c])) => (Const64F [math.RoundToEven(c)])
// Convert x * 1 to x.
(Mul(8|16|32|64) (Const(8|16|32|64) [1]) x) => x
(Select0 (Mul(32|64)uover (Const(32|64) [1]) x)) => x
(Select1 (Mul(32|64)uover (Const(32|64) [1]) x)) => (ConstBool [false])
// Convert x * -1 to -x.
(Mul(8|16|32|64) (Const(8|16|32|64) [-1]) x) => (Neg(8|16|32|64) x)
// DeMorgan's Laws
(And(8|16|32|64) <t> (Com(8|16|32|64) x) (Com(8|16|32|64) y)) => (Com(8|16|32|64) (Or(8|16|32|64) <t> x y))
(Or(8|16|32|64) <t> (Com(8|16|32|64) x) (Com(8|16|32|64) y)) => (Com(8|16|32|64) (And(8|16|32|64) <t> x y))
// Convert multiplication by a power of two to a shift.
(Mul8 <t> n (Const8 [c])) && isPowerOfTwo(c) => (Lsh8x64 <t> n (Const64 <typ.UInt64> [log8(c)]))
(Mul16 <t> n (Const16 [c])) && isPowerOfTwo(c) => (Lsh16x64 <t> n (Const64 <typ.UInt64> [log16(c)]))
(Mul32 <t> n (Const32 [c])) && isPowerOfTwo(c) => (Lsh32x64 <t> n (Const64 <typ.UInt64> [log32(c)]))
(Mul64 <t> n (Const64 [c])) && isPowerOfTwo(c) => (Lsh64x64 <t> n (Const64 <typ.UInt64> [log64(c)]))
(Mul8 <t> n (Const8 [c])) && t.IsSigned() && isPowerOfTwo(-c) => (Neg8 (Lsh8x64 <t> n (Const64 <typ.UInt64> [log8(-c)])))
(Mul16 <t> n (Const16 [c])) && t.IsSigned() && isPowerOfTwo(-c) => (Neg16 (Lsh16x64 <t> n (Const64 <typ.UInt64> [log16(-c)])))
(Mul32 <t> n (Const32 [c])) && t.IsSigned() && isPowerOfTwo(-c) => (Neg32 (Lsh32x64 <t> n (Const64 <typ.UInt64> [log32(-c)])))
(Mul64 <t> n (Const64 [c])) && t.IsSigned() && isPowerOfTwo(-c) => (Neg64 (Lsh64x64 <t> n (Const64 <typ.UInt64> [log64(-c)])))
(Mod8 (Const8 [c]) (Const8 [d])) && d != 0 => (Const8 [c % d])
(Mod16 (Const16 [c]) (Const16 [d])) && d != 0 => (Const16 [c % d])
(Mod32 (Const32 [c]) (Const32 [d])) && d != 0 => (Const32 [c % d])
(Mod64 (Const64 [c]) (Const64 [d])) && d != 0 => (Const64 [c % d])
(Mod8u (Const8 [c]) (Const8 [d])) && d != 0 => (Const8 [int8(uint8(c) % uint8(d))])
(Mod16u (Const16 [c]) (Const16 [d])) && d != 0 => (Const16 [int16(uint16(c) % uint16(d))])
(Mod32u (Const32 [c]) (Const32 [d])) && d != 0 => (Const32 [int32(uint32(c) % uint32(d))])
(Mod64u (Const64 [c]) (Const64 [d])) && d != 0 => (Const64 [int64(uint64(c) % uint64(d))])
(Lsh64x64 (Const64 [c]) (Const64 [d])) => (Const64 [c << uint64(d)])
(Rsh64x64 (Const64 [c]) (Const64 [d])) => (Const64 [c >> uint64(d)])
(Rsh64Ux64 (Const64 [c]) (Const64 [d])) => (Const64 [int64(uint64(c) >> uint64(d))])
(Lsh32x64 (Const32 [c]) (Const64 [d])) => (Const32 [c << uint64(d)])
(Rsh32x64 (Const32 [c]) (Const64 [d])) => (Const32 [c >> uint64(d)])
(Rsh32Ux64 (Const32 [c]) (Const64 [d])) => (Const32 [int32(uint32(c) >> uint64(d))])
(Lsh16x64 (Const16 [c]) (Const64 [d])) => (Const16 [c << uint64(d)])
(Rsh16x64 (Const16 [c]) (Const64 [d])) => (Const16 [c >> uint64(d)])
(Rsh16Ux64 (Const16 [c]) (Const64 [d])) => (Const16 [int16(uint16(c) >> uint64(d))])
(Lsh8x64 (Const8 [c]) (Const64 [d])) => (Const8 [c << uint64(d)])
(Rsh8x64 (Const8 [c]) (Const64 [d])) => (Const8 [c >> uint64(d)])
(Rsh8Ux64 (Const8 [c]) (Const64 [d])) => (Const8 [int8(uint8(c) >> uint64(d))])
// Fold IsInBounds when the range of the index cannot exceed the limit.
(IsInBounds (ZeroExt8to32 _) (Const32 [c])) && (1 << 8) <= c => (ConstBool [true])
(IsInBounds (ZeroExt8to64 _) (Const64 [c])) && (1 << 8) <= c => (ConstBool [true])
(IsInBounds (ZeroExt16to32 _) (Const32 [c])) && (1 << 16) <= c => (ConstBool [true])
(IsInBounds (ZeroExt16to64 _) (Const64 [c])) && (1 << 16) <= c => (ConstBool [true])
(IsInBounds x x) => (ConstBool [false])
(IsInBounds (And8 (Const8 [c]) _) (Const8 [d])) && 0 <= c && c < d => (ConstBool [true])
(IsInBounds (ZeroExt8to16 (And8 (Const8 [c]) _)) (Const16 [d])) && 0 <= c && int16(c) < d => (ConstBool [true])
(IsInBounds (ZeroExt8to32 (And8 (Const8 [c]) _)) (Const32 [d])) && 0 <= c && int32(c) < d => (ConstBool [true])
(IsInBounds (ZeroExt8to64 (And8 (Const8 [c]) _)) (Const64 [d])) && 0 <= c && int64(c) < d => (ConstBool [true])
(IsInBounds (And16 (Const16 [c]) _) (Const16 [d])) && 0 <= c && c < d => (ConstBool [true])
(IsInBounds (ZeroExt16to32 (And16 (Const16 [c]) _)) (Const32 [d])) && 0 <= c && int32(c) < d => (ConstBool [true])
(IsInBounds (ZeroExt16to64 (And16 (Const16 [c]) _)) (Const64 [d])) && 0 <= c && int64(c) < d => (ConstBool [true])
(IsInBounds (And32 (Const32 [c]) _) (Const32 [d])) && 0 <= c && c < d => (ConstBool [true])
(IsInBounds (ZeroExt32to64 (And32 (Const32 [c]) _)) (Const64 [d])) && 0 <= c && int64(c) < d => (ConstBool [true])
(IsInBounds (And64 (Const64 [c]) _) (Const64 [d])) && 0 <= c && c < d => (ConstBool [true])
(IsInBounds (Const32 [c]) (Const32 [d])) => (ConstBool [0 <= c && c < d])
(IsInBounds (Const64 [c]) (Const64 [d])) => (ConstBool [0 <= c && c < d])
// (Mod64u x y) is always between 0 (inclusive) and y (exclusive).
(IsInBounds (Mod32u _ y) y) => (ConstBool [true])
(IsInBounds (Mod64u _ y) y) => (ConstBool [true])
// Right shifting an unsigned number limits its value.
(IsInBounds (ZeroExt8to64 (Rsh8Ux64 _ (Const64 [c]))) (Const64 [d])) && 0 < c && c < 8 && 1<<uint( 8-c)-1 < d => (ConstBool [true])
(IsInBounds (ZeroExt8to32 (Rsh8Ux64 _ (Const64 [c]))) (Const32 [d])) && 0 < c && c < 8 && 1<<uint( 8-c)-1 < d => (ConstBool [true])
(IsInBounds (ZeroExt8to16 (Rsh8Ux64 _ (Const64 [c]))) (Const16 [d])) && 0 < c && c < 8 && 1<<uint( 8-c)-1 < d => (ConstBool [true])
(IsInBounds (Rsh8Ux64 _ (Const64 [c])) (Const64 [d])) && 0 < c && c < 8 && 1<<uint( 8-c)-1 < d => (ConstBool [true])
(IsInBounds (ZeroExt16to64 (Rsh16Ux64 _ (Const64 [c]))) (Const64 [d])) && 0 < c && c < 16 && 1<<uint(16-c)-1 < d => (ConstBool [true])
(IsInBounds (ZeroExt16to32 (Rsh16Ux64 _ (Const64 [c]))) (Const64 [d])) && 0 < c && c < 16 && 1<<uint(16-c)-1 < d => (ConstBool [true])
(IsInBounds (Rsh16Ux64 _ (Const64 [c])) (Const64 [d])) && 0 < c && c < 16 && 1<<uint(16-c)-1 < d => (ConstBool [true])
(IsInBounds (ZeroExt32to64 (Rsh32Ux64 _ (Const64 [c]))) (Const64 [d])) && 0 < c && c < 32 && 1<<uint(32-c)-1 < d => (ConstBool [true])
(IsInBounds (Rsh32Ux64 _ (Const64 [c])) (Const64 [d])) && 0 < c && c < 32 && 1<<uint(32-c)-1 < d => (ConstBool [true])
(IsInBounds (Rsh64Ux64 _ (Const64 [c])) (Const64 [d])) && 0 < c && c < 64 && 1<<uint(64-c)-1 < d => (ConstBool [true])
(IsSliceInBounds x x) => (ConstBool [true])
(IsSliceInBounds (And32 (Const32 [c]) _) (Const32 [d])) && 0 <= c && c <= d => (ConstBool [true])
(IsSliceInBounds (And64 (Const64 [c]) _) (Const64 [d])) && 0 <= c && c <= d => (ConstBool [true])
(IsSliceInBounds (Const32 [0]) _) => (ConstBool [true])
(IsSliceInBounds (Const64 [0]) _) => (ConstBool [true])
(IsSliceInBounds (Const32 [c]) (Const32 [d])) => (ConstBool [0 <= c && c <= d])
(IsSliceInBounds (Const64 [c]) (Const64 [d])) => (ConstBool [0 <= c && c <= d])
(IsSliceInBounds (SliceLen x) (SliceCap x)) => (ConstBool [true])
(Eq(64|32|16|8) x x) => (ConstBool [true])
(EqB (ConstBool [c]) (ConstBool [d])) => (ConstBool [c == d])
(EqB (ConstBool [false]) x) => (Not x)
(EqB (ConstBool [true]) x) => x
(Neq(64|32|16|8) x x) => (ConstBool [false])
(NeqB (ConstBool [c]) (ConstBool [d])) => (ConstBool [c != d])
(NeqB (ConstBool [false]) x) => x
(NeqB (ConstBool [true]) x) => (Not x)
(NeqB (Not x) (Not y)) => (NeqB x y)
(Eq64 (Const64 <t> [c]) (Add64 (Const64 <t> [d]) x)) => (Eq64 (Const64 <t> [c-d]) x)
(Eq32 (Const32 <t> [c]) (Add32 (Const32 <t> [d]) x)) => (Eq32 (Const32 <t> [c-d]) x)
(Eq16 (Const16 <t> [c]) (Add16 (Const16 <t> [d]) x)) => (Eq16 (Const16 <t> [c-d]) x)
(Eq8 (Const8 <t> [c]) (Add8 (Const8 <t> [d]) x)) => (Eq8 (Const8 <t> [c-d]) x)
(Neq64 (Const64 <t> [c]) (Add64 (Const64 <t> [d]) x)) => (Neq64 (Const64 <t> [c-d]) x)
(Neq32 (Const32 <t> [c]) (Add32 (Const32 <t> [d]) x)) => (Neq32 (Const32 <t> [c-d]) x)
(Neq16 (Const16 <t> [c]) (Add16 (Const16 <t> [d]) x)) => (Neq16 (Const16 <t> [c-d]) x)
(Neq8 (Const8 <t> [c]) (Add8 (Const8 <t> [d]) x)) => (Neq8 (Const8 <t> [c-d]) x)
cmd/compile: optimize integer-in-range checks This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-20 11:55:56 -07:00
// signed integer range: ( c <= x && x (<|<=) d ) -> ( unsigned(x-c) (<|<=) unsigned(d-c) )
(AndB (Leq64 (Const64 [c]) x) ((Less|Leq)64 x (Const64 [d]))) && d >= c => ((Less|Leq)64U (Sub64 <x.Type> x (Const64 <x.Type> [c])) (Const64 <x.Type> [d-c]))
(AndB (Leq32 (Const32 [c]) x) ((Less|Leq)32 x (Const32 [d]))) && d >= c => ((Less|Leq)32U (Sub32 <x.Type> x (Const32 <x.Type> [c])) (Const32 <x.Type> [d-c]))
(AndB (Leq16 (Const16 [c]) x) ((Less|Leq)16 x (Const16 [d]))) && d >= c => ((Less|Leq)16U (Sub16 <x.Type> x (Const16 <x.Type> [c])) (Const16 <x.Type> [d-c]))
(AndB (Leq8 (Const8 [c]) x) ((Less|Leq)8 x (Const8 [d]))) && d >= c => ((Less|Leq)8U (Sub8 <x.Type> x (Const8 <x.Type> [c])) (Const8 <x.Type> [d-c]))
cmd/compile: optimize integer-in-range checks This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-20 11:55:56 -07:00
// signed integer range: ( c < x && x (<|<=) d ) -> ( unsigned(x-(c+1)) (<|<=) unsigned(d-(c+1)) )
(AndB (Less64 (Const64 [c]) x) ((Less|Leq)64 x (Const64 [d]))) && d >= c+1 && c+1 > c => ((Less|Leq)64U (Sub64 <x.Type> x (Const64 <x.Type> [c+1])) (Const64 <x.Type> [d-c-1]))
(AndB (Less32 (Const32 [c]) x) ((Less|Leq)32 x (Const32 [d]))) && d >= c+1 && c+1 > c => ((Less|Leq)32U (Sub32 <x.Type> x (Const32 <x.Type> [c+1])) (Const32 <x.Type> [d-c-1]))
(AndB (Less16 (Const16 [c]) x) ((Less|Leq)16 x (Const16 [d]))) && d >= c+1 && c+1 > c => ((Less|Leq)16U (Sub16 <x.Type> x (Const16 <x.Type> [c+1])) (Const16 <x.Type> [d-c-1]))
(AndB (Less8 (Const8 [c]) x) ((Less|Leq)8 x (Const8 [d]))) && d >= c+1 && c+1 > c => ((Less|Leq)8U (Sub8 <x.Type> x (Const8 <x.Type> [c+1])) (Const8 <x.Type> [d-c-1]))
cmd/compile: optimize integer-in-range checks This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-20 11:55:56 -07:00
// unsigned integer range: ( c <= x && x (<|<=) d ) -> ( x-c (<|<=) d-c )
(AndB (Leq64U (Const64 [c]) x) ((Less|Leq)64U x (Const64 [d]))) && uint64(d) >= uint64(c) => ((Less|Leq)64U (Sub64 <x.Type> x (Const64 <x.Type> [c])) (Const64 <x.Type> [d-c]))
(AndB (Leq32U (Const32 [c]) x) ((Less|Leq)32U x (Const32 [d]))) && uint32(d) >= uint32(c) => ((Less|Leq)32U (Sub32 <x.Type> x (Const32 <x.Type> [c])) (Const32 <x.Type> [d-c]))
(AndB (Leq16U (Const16 [c]) x) ((Less|Leq)16U x (Const16 [d]))) && uint16(d) >= uint16(c) => ((Less|Leq)16U (Sub16 <x.Type> x (Const16 <x.Type> [c])) (Const16 <x.Type> [d-c]))
(AndB (Leq8U (Const8 [c]) x) ((Less|Leq)8U x (Const8 [d]))) && uint8(d) >= uint8(c) => ((Less|Leq)8U (Sub8 <x.Type> x (Const8 <x.Type> [c])) (Const8 <x.Type> [d-c]))
cmd/compile: optimize integer-in-range checks This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-20 11:55:56 -07:00
// unsigned integer range: ( c < x && x (<|<=) d ) -> ( x-(c+1) (<|<=) d-(c+1) )
(AndB (Less64U (Const64 [c]) x) ((Less|Leq)64U x (Const64 [d]))) && uint64(d) >= uint64(c+1) && uint64(c+1) > uint64(c) => ((Less|Leq)64U (Sub64 <x.Type> x (Const64 <x.Type> [c+1])) (Const64 <x.Type> [d-c-1]))
(AndB (Less32U (Const32 [c]) x) ((Less|Leq)32U x (Const32 [d]))) && uint32(d) >= uint32(c+1) && uint32(c+1) > uint32(c) => ((Less|Leq)32U (Sub32 <x.Type> x (Const32 <x.Type> [c+1])) (Const32 <x.Type> [d-c-1]))
(AndB (Less16U (Const16 [c]) x) ((Less|Leq)16U x (Const16 [d]))) && uint16(d) >= uint16(c+1) && uint16(c+1) > uint16(c) => ((Less|Leq)16U (Sub16 <x.Type> x (Const16 <x.Type> [c+1])) (Const16 <x.Type> [d-c-1]))
(AndB (Less8U (Const8 [c]) x) ((Less|Leq)8U x (Const8 [d]))) && uint8(d) >= uint8(c+1) && uint8(c+1) > uint8(c) => ((Less|Leq)8U (Sub8 <x.Type> x (Const8 <x.Type> [c+1])) (Const8 <x.Type> [d-c-1]))
cmd/compile: optimize integer-in-range checks This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-20 11:55:56 -07:00
// signed integer range: ( c (<|<=) x || x < d ) -> ( unsigned(c-d) (<|<=) unsigned(x-d) )
(OrB ((Less|Leq)64 (Const64 [c]) x) (Less64 x (Const64 [d]))) && c >= d => ((Less|Leq)64U (Const64 <x.Type> [c-d]) (Sub64 <x.Type> x (Const64 <x.Type> [d])))
(OrB ((Less|Leq)32 (Const32 [c]) x) (Less32 x (Const32 [d]))) && c >= d => ((Less|Leq)32U (Const32 <x.Type> [c-d]) (Sub32 <x.Type> x (Const32 <x.Type> [d])))
(OrB ((Less|Leq)16 (Const16 [c]) x) (Less16 x (Const16 [d]))) && c >= d => ((Less|Leq)16U (Const16 <x.Type> [c-d]) (Sub16 <x.Type> x (Const16 <x.Type> [d])))
(OrB ((Less|Leq)8 (Const8 [c]) x) (Less8 x (Const8 [d]))) && c >= d => ((Less|Leq)8U (Const8 <x.Type> [c-d]) (Sub8 <x.Type> x (Const8 <x.Type> [d])))
cmd/compile: optimize integer-in-range checks This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-20 11:55:56 -07:00
// signed integer range: ( c (<|<=) x || x <= d ) -> ( unsigned(c-(d+1)) (<|<=) unsigned(x-(d+1)) )
(OrB ((Less|Leq)64 (Const64 [c]) x) (Leq64 x (Const64 [d]))) && c >= d+1 && d+1 > d => ((Less|Leq)64U (Const64 <x.Type> [c-d-1]) (Sub64 <x.Type> x (Const64 <x.Type> [d+1])))
(OrB ((Less|Leq)32 (Const32 [c]) x) (Leq32 x (Const32 [d]))) && c >= d+1 && d+1 > d => ((Less|Leq)32U (Const32 <x.Type> [c-d-1]) (Sub32 <x.Type> x (Const32 <x.Type> [d+1])))
(OrB ((Less|Leq)16 (Const16 [c]) x) (Leq16 x (Const16 [d]))) && c >= d+1 && d+1 > d => ((Less|Leq)16U (Const16 <x.Type> [c-d-1]) (Sub16 <x.Type> x (Const16 <x.Type> [d+1])))
(OrB ((Less|Leq)8 (Const8 [c]) x) (Leq8 x (Const8 [d]))) && c >= d+1 && d+1 > d => ((Less|Leq)8U (Const8 <x.Type> [c-d-1]) (Sub8 <x.Type> x (Const8 <x.Type> [d+1])))
cmd/compile: optimize integer-in-range checks This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-20 11:55:56 -07:00
// unsigned integer range: ( c (<|<=) x || x < d ) -> ( c-d (<|<=) x-d )
(OrB ((Less|Leq)64U (Const64 [c]) x) (Less64U x (Const64 [d]))) && uint64(c) >= uint64(d) => ((Less|Leq)64U (Const64 <x.Type> [c-d]) (Sub64 <x.Type> x (Const64 <x.Type> [d])))
(OrB ((Less|Leq)32U (Const32 [c]) x) (Less32U x (Const32 [d]))) && uint32(c) >= uint32(d) => ((Less|Leq)32U (Const32 <x.Type> [c-d]) (Sub32 <x.Type> x (Const32 <x.Type> [d])))
(OrB ((Less|Leq)16U (Const16 [c]) x) (Less16U x (Const16 [d]))) && uint16(c) >= uint16(d) => ((Less|Leq)16U (Const16 <x.Type> [c-d]) (Sub16 <x.Type> x (Const16 <x.Type> [d])))
(OrB ((Less|Leq)8U (Const8 [c]) x) (Less8U x (Const8 [d]))) && uint8(c) >= uint8(d) => ((Less|Leq)8U (Const8 <x.Type> [c-d]) (Sub8 <x.Type> x (Const8 <x.Type> [d])))
cmd/compile: optimize integer-in-range checks This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-20 11:55:56 -07:00
// unsigned integer range: ( c (<|<=) x || x <= d ) -> ( c-(d+1) (<|<=) x-(d+1) )
(OrB ((Less|Leq)64U (Const64 [c]) x) (Leq64U x (Const64 [d]))) && uint64(c) >= uint64(d+1) && uint64(d+1) > uint64(d) => ((Less|Leq)64U (Const64 <x.Type> [c-d-1]) (Sub64 <x.Type> x (Const64 <x.Type> [d+1])))
(OrB ((Less|Leq)32U (Const32 [c]) x) (Leq32U x (Const32 [d]))) && uint32(c) >= uint32(d+1) && uint32(d+1) > uint32(d) => ((Less|Leq)32U (Const32 <x.Type> [c-d-1]) (Sub32 <x.Type> x (Const32 <x.Type> [d+1])))
(OrB ((Less|Leq)16U (Const16 [c]) x) (Leq16U x (Const16 [d]))) && uint16(c) >= uint16(d+1) && uint16(d+1) > uint16(d) => ((Less|Leq)16U (Const16 <x.Type> [c-d-1]) (Sub16 <x.Type> x (Const16 <x.Type> [d+1])))
(OrB ((Less|Leq)8U (Const8 [c]) x) (Leq8U x (Const8 [d]))) && uint8(c) >= uint8(d+1) && uint8(d+1) > uint8(d) => ((Less|Leq)8U (Const8 <x.Type> [c-d-1]) (Sub8 <x.Type> x (Const8 <x.Type> [d+1])))
cmd/compile: optimize integer-in-range checks This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-20 11:55:56 -07:00
cmd/compile: automatically handle commuting ops in rewrite rules Note that this is a redo of an undo of the original buggy CL 38666. We have lots of rewrite rules that vary only in the fact that we have 2 versions for the 2 different orderings of various commuting ops. For example: (ADDL x (MOVLconst [c])) -> (ADDLconst [c] x) (ADDL (MOVLconst [c]) x) -> (ADDLconst [c] x) It can get unwieldly quickly, especially when there is more than one commuting op in a rule. Our existing "fix" for this problem is to have rules that canonicalize the operations first. For example: (Eq64 x (Const64 <t> [c])) && x.Op != OpConst64 -> (Eq64 (Const64 <t> [c]) x) Subsequent rules can then assume if there is a constant arg to Eq64, it will be the first one. This fix kinda works, but it is fragile and only works when we remember to include the required extra rules. The fundamental problem is that the rule matcher doesn't know anything about commuting ops. This CL fixes that fact. We already have information about which ops commute. (The register allocator takes advantage of commutivity.) The rule generator now automatically generates multiple rules for a single source rule when there are commutative ops in the rule. We can now drop all of our almost-duplicate source-level rules and the canonicalization rules. I have some CLs in progress that will be a lot less verbose when the rule generator handles commutivity for me. I had to reorganize the load-combining rules a bit. The 8-way OR rules generated 128 different reorderings, which was causing the generator to put too much code in the rewrite*.go files (the big ones were going from 25K lines to 132K lines). Instead I reorganized the rules to combine pairs of loads at a time. The generated rule files are now actually a bit (5%) smaller. Make.bash times are ~unchanged. Compiler benchmarks are not observably different. Probably because we don't spend much compiler time in rule matching anyway. I've also done a pass over all of our ops adding commutative markings for ops which hadn't had them previously. Fixes #18292 Change-Id: Ic1c0e43fbf579539f459971625f69690c9ab8805 Reviewed-on: https://go-review.googlesource.com/38801 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2017-03-30 03:30:22 +00:00
// Canonicalize x-const to x+(-const)
(Sub64 x (Const64 <t> [c])) && x.Op != OpConst64 => (Add64 (Const64 <t> [-c]) x)
(Sub32 x (Const32 <t> [c])) && x.Op != OpConst32 => (Add32 (Const32 <t> [-c]) x)
(Sub16 x (Const16 <t> [c])) && x.Op != OpConst16 => (Add16 (Const16 <t> [-c]) x)
(Sub8 x (Const8 <t> [c])) && x.Op != OpConst8 => (Add8 (Const8 <t> [-c]) x)
// fold negation into comparison operators
(Not (Eq(64|32|16|8|B|Ptr|64F|32F) x y)) => (Neq(64|32|16|8|B|Ptr|64F|32F) x y)
(Not (Neq(64|32|16|8|B|Ptr|64F|32F) x y)) => (Eq(64|32|16|8|B|Ptr|64F|32F) x y)
(Not (Less(64|32|16|8) x y)) => (Leq(64|32|16|8) y x)
(Not (Less(64|32|16|8)U x y)) => (Leq(64|32|16|8)U y x)
(Not (Leq(64|32|16|8) x y)) => (Less(64|32|16|8) y x)
(Not (Leq(64|32|16|8)U x y)) => (Less(64|32|16|8)U y x)
// Distribute multiplication c * (d+x) -> c*d + c*x. Useful for:
// a[i].b = ...; a[i+1].b = ...
(Mul64 (Const64 <t> [c]) (Add64 <t> (Const64 <t> [d]) x)) =>
(Add64 (Const64 <t> [c*d]) (Mul64 <t> (Const64 <t> [c]) x))
(Mul32 (Const32 <t> [c]) (Add32 <t> (Const32 <t> [d]) x)) =>
(Add32 (Const32 <t> [c*d]) (Mul32 <t> (Const32 <t> [c]) x))
(Mul16 (Const16 <t> [c]) (Add16 <t> (Const16 <t> [d]) x)) =>
(Add16 (Const16 <t> [c*d]) (Mul16 <t> (Const16 <t> [c]) x))
(Mul8 (Const8 <t> [c]) (Add8 <t> (Const8 <t> [d]) x)) =>
(Add8 (Const8 <t> [c*d]) (Mul8 <t> (Const8 <t> [c]) x))
// Rewrite x*y ± x*z to x*(y±z)
(Add(64|32|16|8) <t> (Mul(64|32|16|8) x y) (Mul(64|32|16|8) x z))
=> (Mul(64|32|16|8) x (Add(64|32|16|8) <t> y z))
(Sub(64|32|16|8) <t> (Mul(64|32|16|8) x y) (Mul(64|32|16|8) x z))
=> (Mul(64|32|16|8) x (Sub(64|32|16|8) <t> y z))
// rewrite shifts of 8/16/32 bit consts into 64 bit consts to reduce
// the number of the other rewrite rules for const shifts
(Lsh64x32 <t> x (Const32 [c])) => (Lsh64x64 x (Const64 <t> [int64(uint32(c))]))
(Lsh64x16 <t> x (Const16 [c])) => (Lsh64x64 x (Const64 <t> [int64(uint16(c))]))
(Lsh64x8 <t> x (Const8 [c])) => (Lsh64x64 x (Const64 <t> [int64(uint8(c))]))
(Rsh64x32 <t> x (Const32 [c])) => (Rsh64x64 x (Const64 <t> [int64(uint32(c))]))
(Rsh64x16 <t> x (Const16 [c])) => (Rsh64x64 x (Const64 <t> [int64(uint16(c))]))
(Rsh64x8 <t> x (Const8 [c])) => (Rsh64x64 x (Const64 <t> [int64(uint8(c))]))
(Rsh64Ux32 <t> x (Const32 [c])) => (Rsh64Ux64 x (Const64 <t> [int64(uint32(c))]))
(Rsh64Ux16 <t> x (Const16 [c])) => (Rsh64Ux64 x (Const64 <t> [int64(uint16(c))]))
(Rsh64Ux8 <t> x (Const8 [c])) => (Rsh64Ux64 x (Const64 <t> [int64(uint8(c))]))
(Lsh32x32 <t> x (Const32 [c])) => (Lsh32x64 x (Const64 <t> [int64(uint32(c))]))
(Lsh32x16 <t> x (Const16 [c])) => (Lsh32x64 x (Const64 <t> [int64(uint16(c))]))
(Lsh32x8 <t> x (Const8 [c])) => (Lsh32x64 x (Const64 <t> [int64(uint8(c))]))
(Rsh32x32 <t> x (Const32 [c])) => (Rsh32x64 x (Const64 <t> [int64(uint32(c))]))
(Rsh32x16 <t> x (Const16 [c])) => (Rsh32x64 x (Const64 <t> [int64(uint16(c))]))
(Rsh32x8 <t> x (Const8 [c])) => (Rsh32x64 x (Const64 <t> [int64(uint8(c))]))
(Rsh32Ux32 <t> x (Const32 [c])) => (Rsh32Ux64 x (Const64 <t> [int64(uint32(c))]))
(Rsh32Ux16 <t> x (Const16 [c])) => (Rsh32Ux64 x (Const64 <t> [int64(uint16(c))]))
(Rsh32Ux8 <t> x (Const8 [c])) => (Rsh32Ux64 x (Const64 <t> [int64(uint8(c))]))
(Lsh16x32 <t> x (Const32 [c])) => (Lsh16x64 x (Const64 <t> [int64(uint32(c))]))
(Lsh16x16 <t> x (Const16 [c])) => (Lsh16x64 x (Const64 <t> [int64(uint16(c))]))
(Lsh16x8 <t> x (Const8 [c])) => (Lsh16x64 x (Const64 <t> [int64(uint8(c))]))
(Rsh16x32 <t> x (Const32 [c])) => (Rsh16x64 x (Const64 <t> [int64(uint32(c))]))
(Rsh16x16 <t> x (Const16 [c])) => (Rsh16x64 x (Const64 <t> [int64(uint16(c))]))
(Rsh16x8 <t> x (Const8 [c])) => (Rsh16x64 x (Const64 <t> [int64(uint8(c))]))
(Rsh16Ux32 <t> x (Const32 [c])) => (Rsh16Ux64 x (Const64 <t> [int64(uint32(c))]))
(Rsh16Ux16 <t> x (Const16 [c])) => (Rsh16Ux64 x (Const64 <t> [int64(uint16(c))]))
(Rsh16Ux8 <t> x (Const8 [c])) => (Rsh16Ux64 x (Const64 <t> [int64(uint8(c))]))
(Lsh8x32 <t> x (Const32 [c])) => (Lsh8x64 x (Const64 <t> [int64(uint32(c))]))
(Lsh8x16 <t> x (Const16 [c])) => (Lsh8x64 x (Const64 <t> [int64(uint16(c))]))
(Lsh8x8 <t> x (Const8 [c])) => (Lsh8x64 x (Const64 <t> [int64(uint8(c))]))
(Rsh8x32 <t> x (Const32 [c])) => (Rsh8x64 x (Const64 <t> [int64(uint32(c))]))
(Rsh8x16 <t> x (Const16 [c])) => (Rsh8x64 x (Const64 <t> [int64(uint16(c))]))
(Rsh8x8 <t> x (Const8 [c])) => (Rsh8x64 x (Const64 <t> [int64(uint8(c))]))
(Rsh8Ux32 <t> x (Const32 [c])) => (Rsh8Ux64 x (Const64 <t> [int64(uint32(c))]))
(Rsh8Ux16 <t> x (Const16 [c])) => (Rsh8Ux64 x (Const64 <t> [int64(uint16(c))]))
(Rsh8Ux8 <t> x (Const8 [c])) => (Rsh8Ux64 x (Const64 <t> [int64(uint8(c))]))
// shifts by zero
(Lsh(64|32|16|8)x64 x (Const64 [0])) => x
(Rsh(64|32|16|8)x64 x (Const64 [0])) => x
(Rsh(64|32|16|8)Ux64 x (Const64 [0])) => x
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
// rotates by multiples of register width
(RotateLeft64 x (Const64 [c])) && c%64 == 0 => x
(RotateLeft32 x (Const32 [c])) && c%32 == 0 => x
(RotateLeft16 x (Const16 [c])) && c%16 == 0 => x
(RotateLeft8 x (Const8 [c])) && c%8 == 0 => x
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
// zero shifted
(Lsh64x(64|32|16|8) (Const64 [0]) _) => (Const64 [0])
(Rsh64x(64|32|16|8) (Const64 [0]) _) => (Const64 [0])
(Rsh64Ux(64|32|16|8) (Const64 [0]) _) => (Const64 [0])
(Lsh32x(64|32|16|8) (Const32 [0]) _) => (Const32 [0])
(Rsh32x(64|32|16|8) (Const32 [0]) _) => (Const32 [0])
(Rsh32Ux(64|32|16|8) (Const32 [0]) _) => (Const32 [0])
(Lsh16x(64|32|16|8) (Const16 [0]) _) => (Const16 [0])
(Rsh16x(64|32|16|8) (Const16 [0]) _) => (Const16 [0])
(Rsh16Ux(64|32|16|8) (Const16 [0]) _) => (Const16 [0])
(Lsh8x(64|32|16|8) (Const8 [0]) _) => (Const8 [0])
(Rsh8x(64|32|16|8) (Const8 [0]) _) => (Const8 [0])
(Rsh8Ux(64|32|16|8) (Const8 [0]) _) => (Const8 [0])
// large left shifts of all values, and right shifts of unsigned values
((Lsh64|Rsh64U)x64 _ (Const64 [c])) && uint64(c) >= 64 => (Const64 [0])
((Lsh32|Rsh32U)x64 _ (Const64 [c])) && uint64(c) >= 32 => (Const32 [0])
((Lsh16|Rsh16U)x64 _ (Const64 [c])) && uint64(c) >= 16 => (Const16 [0])
((Lsh8|Rsh8U)x64 _ (Const64 [c])) && uint64(c) >= 8 => (Const8 [0])
// combine const shifts
(Lsh64x64 <t> (Lsh64x64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Lsh64x64 x (Const64 <t> [c+d]))
(Lsh32x64 <t> (Lsh32x64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Lsh32x64 x (Const64 <t> [c+d]))
(Lsh16x64 <t> (Lsh16x64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Lsh16x64 x (Const64 <t> [c+d]))
(Lsh8x64 <t> (Lsh8x64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Lsh8x64 x (Const64 <t> [c+d]))
(Rsh64x64 <t> (Rsh64x64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Rsh64x64 x (Const64 <t> [c+d]))
(Rsh32x64 <t> (Rsh32x64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Rsh32x64 x (Const64 <t> [c+d]))
(Rsh16x64 <t> (Rsh16x64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Rsh16x64 x (Const64 <t> [c+d]))
(Rsh8x64 <t> (Rsh8x64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Rsh8x64 x (Const64 <t> [c+d]))
(Rsh64Ux64 <t> (Rsh64Ux64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Rsh64Ux64 x (Const64 <t> [c+d]))
(Rsh32Ux64 <t> (Rsh32Ux64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Rsh32Ux64 x (Const64 <t> [c+d]))
(Rsh16Ux64 <t> (Rsh16Ux64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Rsh16Ux64 x (Const64 <t> [c+d]))
(Rsh8Ux64 <t> (Rsh8Ux64 x (Const64 [c])) (Const64 [d])) && !uaddOvf(c,d) => (Rsh8Ux64 x (Const64 <t> [c+d]))
// Remove signed right shift before an unsigned right shift that extracts the sign bit.
(Rsh8Ux64 (Rsh8x64 x _) (Const64 <t> [7] )) => (Rsh8Ux64 x (Const64 <t> [7] ))
(Rsh16Ux64 (Rsh16x64 x _) (Const64 <t> [15])) => (Rsh16Ux64 x (Const64 <t> [15]))
(Rsh32Ux64 (Rsh32x64 x _) (Const64 <t> [31])) => (Rsh32Ux64 x (Const64 <t> [31]))
(Rsh64Ux64 (Rsh64x64 x _) (Const64 <t> [63])) => (Rsh64Ux64 x (Const64 <t> [63]))
// Convert x>>c<<c to x&^(1<<c-1)
(Lsh64x64 i:(Rsh(64|64U)x64 x (Const64 [c])) (Const64 [c])) && c >= 0 && c < 64 && i.Uses == 1 => (And64 x (Const64 <v.Type> [int64(-1) << c]))
(Lsh32x64 i:(Rsh(32|32U)x64 x (Const64 [c])) (Const64 [c])) && c >= 0 && c < 32 && i.Uses == 1 => (And32 x (Const32 <v.Type> [int32(-1) << c]))
(Lsh16x64 i:(Rsh(16|16U)x64 x (Const64 [c])) (Const64 [c])) && c >= 0 && c < 16 && i.Uses == 1 => (And16 x (Const16 <v.Type> [int16(-1) << c]))
(Lsh8x64 i:(Rsh(8|8U)x64 x (Const64 [c])) (Const64 [c])) && c >= 0 && c < 8 && i.Uses == 1 => (And8 x (Const8 <v.Type> [int8(-1) << c]))
// similarly for x<<c>>c
(Rsh64Ux64 i:(Lsh64x64 x (Const64 [c])) (Const64 [c])) && c >= 0 && c < 64 && i.Uses == 1 => (And64 x (Const64 <v.Type> [int64(^uint64(0)>>c)]))
(Rsh32Ux64 i:(Lsh32x64 x (Const64 [c])) (Const64 [c])) && c >= 0 && c < 32 && i.Uses == 1 => (And32 x (Const32 <v.Type> [int32(^uint32(0)>>c)]))
(Rsh16Ux64 i:(Lsh16x64 x (Const64 [c])) (Const64 [c])) && c >= 0 && c < 16 && i.Uses == 1 => (And16 x (Const16 <v.Type> [int16(^uint16(0)>>c)]))
(Rsh8Ux64 i:(Lsh8x64 x (Const64 [c])) (Const64 [c])) && c >= 0 && c < 8 && i.Uses == 1 => (And8 x (Const8 <v.Type> [int8 (^uint8 (0)>>c)]))
// ((x >> c1) << c2) >> c3
(Rsh(64|32|16|8)Ux64 (Lsh(64|32|16|8)x64 (Rsh(64|32|16|8)Ux64 x (Const64 [c1])) (Const64 [c2])) (Const64 [c3]))
&& uint64(c1) >= uint64(c2) && uint64(c3) >= uint64(c2) && !uaddOvf(c1-c2, c3)
=> (Rsh(64|32|16|8)Ux64 x (Const64 <typ.UInt64> [c1-c2+c3]))
// ((x << c1) >> c2) << c3
(Lsh(64|32|16|8)x64 (Rsh(64|32|16|8)Ux64 (Lsh(64|32|16|8)x64 x (Const64 [c1])) (Const64 [c2])) (Const64 [c3]))
&& uint64(c1) >= uint64(c2) && uint64(c3) >= uint64(c2) && !uaddOvf(c1-c2, c3)
=> (Lsh(64|32|16|8)x64 x (Const64 <typ.UInt64> [c1-c2+c3]))
// (x >> c) & uppermask = 0
(And64 (Const64 [m]) (Rsh64Ux64 _ (Const64 [c]))) && c >= int64(64-ntz64(m)) => (Const64 [0])
(And32 (Const32 [m]) (Rsh32Ux64 _ (Const64 [c]))) && c >= int64(32-ntz32(m)) => (Const32 [0])
(And16 (Const16 [m]) (Rsh16Ux64 _ (Const64 [c]))) && c >= int64(16-ntz16(m)) => (Const16 [0])
(And8 (Const8 [m]) (Rsh8Ux64 _ (Const64 [c]))) && c >= int64(8-ntz8(m)) => (Const8 [0])
// (x << c) & lowermask = 0
(And64 (Const64 [m]) (Lsh64x64 _ (Const64 [c]))) && c >= int64(64-nlz64(m)) => (Const64 [0])
(And32 (Const32 [m]) (Lsh32x64 _ (Const64 [c]))) && c >= int64(32-nlz32(m)) => (Const32 [0])
(And16 (Const16 [m]) (Lsh16x64 _ (Const64 [c]))) && c >= int64(16-nlz16(m)) => (Const16 [0])
(And8 (Const8 [m]) (Lsh8x64 _ (Const64 [c]))) && c >= int64(8-nlz8(m)) => (Const8 [0])
// replace shifts with zero extensions
(Rsh16Ux64 (Lsh16x64 x (Const64 [8])) (Const64 [8])) => (ZeroExt8to16 (Trunc16to8 <typ.UInt8> x))
(Rsh32Ux64 (Lsh32x64 x (Const64 [24])) (Const64 [24])) => (ZeroExt8to32 (Trunc32to8 <typ.UInt8> x))
(Rsh64Ux64 (Lsh64x64 x (Const64 [56])) (Const64 [56])) => (ZeroExt8to64 (Trunc64to8 <typ.UInt8> x))
(Rsh32Ux64 (Lsh32x64 x (Const64 [16])) (Const64 [16])) => (ZeroExt16to32 (Trunc32to16 <typ.UInt16> x))
(Rsh64Ux64 (Lsh64x64 x (Const64 [48])) (Const64 [48])) => (ZeroExt16to64 (Trunc64to16 <typ.UInt16> x))
(Rsh64Ux64 (Lsh64x64 x (Const64 [32])) (Const64 [32])) => (ZeroExt32to64 (Trunc64to32 <typ.UInt32> x))
// replace shifts with sign extensions
(Rsh16x64 (Lsh16x64 x (Const64 [8])) (Const64 [8])) => (SignExt8to16 (Trunc16to8 <typ.Int8> x))
(Rsh32x64 (Lsh32x64 x (Const64 [24])) (Const64 [24])) => (SignExt8to32 (Trunc32to8 <typ.Int8> x))
(Rsh64x64 (Lsh64x64 x (Const64 [56])) (Const64 [56])) => (SignExt8to64 (Trunc64to8 <typ.Int8> x))
(Rsh32x64 (Lsh32x64 x (Const64 [16])) (Const64 [16])) => (SignExt16to32 (Trunc32to16 <typ.Int16> x))
(Rsh64x64 (Lsh64x64 x (Const64 [48])) (Const64 [48])) => (SignExt16to64 (Trunc64to16 <typ.Int16> x))
(Rsh64x64 (Lsh64x64 x (Const64 [32])) (Const64 [32])) => (SignExt32to64 (Trunc64to32 <typ.Int32> x))
// ((x >> c) & d) << e
(Lsh64x64 (And64 (Rsh(64|64U)x64 <t> x (Const64 <t2> [c])) (Const64 [d])) (Const64 [e])) && c >= e => (And64 (Rsh(64|64U)x64 <t> x (Const64 <t2> [c-e])) (Const64 <t> [d<<e]))
(Lsh32x64 (And32 (Rsh(32|32U)x64 <t> x (Const64 <t2> [c])) (Const32 [d])) (Const64 [e])) && c >= e => (And32 (Rsh(32|32U)x64 <t> x (Const64 <t2> [c-e])) (Const32 <t> [d<<e]))
(Lsh16x64 (And16 (Rsh(16|16U)x64 <t> x (Const64 <t2> [c])) (Const16 [d])) (Const64 [e])) && c >= e => (And16 (Rsh(16|16U)x64 <t> x (Const64 <t2> [c-e])) (Const16 <t> [d<<e]))
(Lsh8x64 (And8 (Rsh(8|8U)x64 <t> x (Const64 <t2> [c])) (Const8 [d])) (Const64 [e])) && c >= e => (And8 (Rsh(8|8U)x64 <t> x (Const64 <t2> [c-e])) (Const8 <t> [d<<e]))
(Lsh64x64 (And64 (Rsh(64|64U)x64 <t> x (Const64 <t2> [c])) (Const64 [d])) (Const64 [e])) && c < e => (And64 (Lsh64x64 <t> x (Const64 <t2> [e-c])) (Const64 <t> [d<<e]))
(Lsh32x64 (And32 (Rsh(32|32U)x64 <t> x (Const64 <t2> [c])) (Const32 [d])) (Const64 [e])) && c < e => (And32 (Lsh32x64 <t> x (Const64 <t2> [e-c])) (Const32 <t> [d<<e]))
(Lsh16x64 (And16 (Rsh(16|16U)x64 <t> x (Const64 <t2> [c])) (Const16 [d])) (Const64 [e])) && c < e => (And16 (Lsh16x64 <t> x (Const64 <t2> [e-c])) (Const16 <t> [d<<e]))
(Lsh8x64 (And8 (Rsh(8|8U)x64 <t> x (Const64 <t2> [c])) (Const8 [d])) (Const64 [e])) && c < e => (And8 (Lsh8x64 <t> x (Const64 <t2> [e-c])) (Const8 <t> [d<<e]))
// constant comparisons
(Eq(64|32|16|8) (Const(64|32|16|8) [c]) (Const(64|32|16|8) [d])) => (ConstBool [c == d])
(Neq(64|32|16|8) (Const(64|32|16|8) [c]) (Const(64|32|16|8) [d])) => (ConstBool [c != d])
(Less(64|32|16|8) (Const(64|32|16|8) [c]) (Const(64|32|16|8) [d])) => (ConstBool [c < d])
(Leq(64|32|16|8) (Const(64|32|16|8) [c]) (Const(64|32|16|8) [d])) => (ConstBool [c <= d])
(Less64U (Const64 [c]) (Const64 [d])) => (ConstBool [uint64(c) < uint64(d)])
(Less32U (Const32 [c]) (Const32 [d])) => (ConstBool [uint32(c) < uint32(d)])
(Less16U (Const16 [c]) (Const16 [d])) => (ConstBool [uint16(c) < uint16(d)])
(Less8U (Const8 [c]) (Const8 [d])) => (ConstBool [ uint8(c) < uint8(d)])
(Leq64U (Const64 [c]) (Const64 [d])) => (ConstBool [uint64(c) <= uint64(d)])
(Leq32U (Const32 [c]) (Const32 [d])) => (ConstBool [uint32(c) <= uint32(d)])
(Leq16U (Const16 [c]) (Const16 [d])) => (ConstBool [uint16(c) <= uint16(d)])
(Leq8U (Const8 [c]) (Const8 [d])) => (ConstBool [ uint8(c) <= uint8(d)])
(Leq8 (Const8 [0]) (And8 _ (Const8 [c]))) && c >= 0 => (ConstBool [true])
(Leq16 (Const16 [0]) (And16 _ (Const16 [c]))) && c >= 0 => (ConstBool [true])
(Leq32 (Const32 [0]) (And32 _ (Const32 [c]))) && c >= 0 => (ConstBool [true])
(Leq64 (Const64 [0]) (And64 _ (Const64 [c]))) && c >= 0 => (ConstBool [true])
(Leq8 (Const8 [0]) (Rsh8Ux64 _ (Const64 [c]))) && c > 0 => (ConstBool [true])
(Leq16 (Const16 [0]) (Rsh16Ux64 _ (Const64 [c]))) && c > 0 => (ConstBool [true])
(Leq32 (Const32 [0]) (Rsh32Ux64 _ (Const64 [c]))) && c > 0 => (ConstBool [true])
(Leq64 (Const64 [0]) (Rsh64Ux64 _ (Const64 [c]))) && c > 0 => (ConstBool [true])
// prefer equalities with zero
(Less(64|32|16|8) (Const(64|32|16|8) <t> [0]) x) && isNonNegative(x) => (Neq(64|32|16|8) (Const(64|32|16|8) <t> [0]) x)
(Less(64|32|16|8) x (Const(64|32|16|8) <t> [1])) && isNonNegative(x) => (Eq(64|32|16|8) (Const(64|32|16|8) <t> [0]) x)
(Less(64|32|16|8)U x (Const(64|32|16|8) <t> [1])) => (Eq(64|32|16|8) (Const(64|32|16|8) <t> [0]) x)
(Leq(64|32|16|8)U (Const(64|32|16|8) <t> [1]) x) => (Neq(64|32|16|8) (Const(64|32|16|8) <t> [0]) x)
// prefer comparisons with zero
(Less(64|32|16|8) x (Const(64|32|16|8) <t> [1])) => (Leq(64|32|16|8) x (Const(64|32|16|8) <t> [0]))
(Leq(64|32|16|8) x (Const(64|32|16|8) <t> [-1])) => (Less(64|32|16|8) x (Const(64|32|16|8) <t> [0]))
(Leq(64|32|16|8) (Const(64|32|16|8) <t> [1]) x) => (Less(64|32|16|8) (Const(64|32|16|8) <t> [0]) x)
(Less(64|32|16|8) (Const(64|32|16|8) <t> [-1]) x) => (Leq(64|32|16|8) (Const(64|32|16|8) <t> [0]) x)
// constant floating point comparisons
(Eq32F (Const32F [c]) (Const32F [d])) => (ConstBool [c == d])
(Eq64F (Const64F [c]) (Const64F [d])) => (ConstBool [c == d])
(Neq32F (Const32F [c]) (Const32F [d])) => (ConstBool [c != d])
(Neq64F (Const64F [c]) (Const64F [d])) => (ConstBool [c != d])
(Less32F (Const32F [c]) (Const32F [d])) => (ConstBool [c < d])
(Less64F (Const64F [c]) (Const64F [d])) => (ConstBool [c < d])
(Leq32F (Const32F [c]) (Const32F [d])) => (ConstBool [c <= d])
(Leq64F (Const64F [c]) (Const64F [d])) => (ConstBool [c <= d])
// simplifications
(Or(64|32|16|8) x x) => x
(Or(64|32|16|8) (Const(64|32|16|8) [0]) x) => x
(Or(64|32|16|8) (Const(64|32|16|8) [-1]) _) => (Const(64|32|16|8) [-1])
(Or(64|32|16|8) (Com(64|32|16|8) x) x) => (Const(64|32|16|8) [-1])
(And(64|32|16|8) x x) => x
(And(64|32|16|8) (Const(64|32|16|8) [-1]) x) => x
(And(64|32|16|8) (Const(64|32|16|8) [0]) _) => (Const(64|32|16|8) [0])
(And(64|32|16|8) (Com(64|32|16|8) x) x) => (Const(64|32|16|8) [0])
(Xor(64|32|16|8) x x) => (Const(64|32|16|8) [0])
(Xor(64|32|16|8) (Const(64|32|16|8) [0]) x) => x
(Xor(64|32|16|8) (Com(64|32|16|8) x) x) => (Const(64|32|16|8) [-1])
(Add(64|32|16|8) (Const(64|32|16|8) [0]) x) => x
(Sub(64|32|16|8) x x) => (Const(64|32|16|8) [0])
(Mul(64|32|16|8) (Const(64|32|16|8) [0]) _) => (Const(64|32|16|8) [0])
(Select0 (Mul(64|32)uover (Const(64|32) [0]) x)) => (Const(64|32) [0])
(Select1 (Mul(64|32)uover (Const(64|32) [0]) x)) => (ConstBool [false])
(Com(64|32|16|8) (Com(64|32|16|8) x)) => x
(Com(64|32|16|8) (Const(64|32|16|8) [c])) => (Const(64|32|16|8) [^c])
(Neg(64|32|16|8) (Sub(64|32|16|8) x y)) => (Sub(64|32|16|8) y x)
(Add(64|32|16|8) x (Neg(64|32|16|8) y)) => (Sub(64|32|16|8) x y)
(Xor(64|32|16|8) (Const(64|32|16|8) [-1]) x) => (Com(64|32|16|8) x)
(Sub(64|32|16|8) (Neg(64|32|16|8) x) (Com(64|32|16|8) x)) => (Const(64|32|16|8) [1])
(Sub(64|32|16|8) (Com(64|32|16|8) x) (Neg(64|32|16|8) x)) => (Const(64|32|16|8) [-1])
(Add(64|32|16|8) (Com(64|32|16|8) x) x) => (Const(64|32|16|8) [-1])
// Simplification when involving common integer
// (t + x) - (t + y) == x - y
// (t + x) - (y + t) == x - y
// (x + t) - (y + t) == x - y
// (x + t) - (t + y) == x - y
// (x - t) + (t + y) == x + y
// (x - t) + (y + t) == x + y
(Sub(64|32|16|8) (Add(64|32|16|8) t x) (Add(64|32|16|8) t y)) => (Sub(64|32|16|8) x y)
(Add(64|32|16|8) (Sub(64|32|16|8) x t) (Add(64|32|16|8) t y)) => (Add(64|32|16|8) x y)
cmd/compile: add more Neg/Com optimizations This is a grab-bag of minor optimizations. While we're here, document the c != -(1<<31) constraints better (#31888). file before after Δ % go 14669924 14665828 -4096 -0.028% asm 4867088 4858896 -8192 -0.168% compile 23988320 23984224 -4096 -0.017% cover 5210856 5206760 -4096 -0.079% link 6084376 6080280 -4096 -0.067% total 132181084 132156508 -24576 -0.019% file before after Δ % archive/tar.a 516708 516702 -6 -0.001% bufio.a 182200 181974 -226 -0.124% bytes.a 217624 216890 -734 -0.337% cmd/compile/internal/gc.a 8865412 8865228 -184 -0.002% cmd/compile/internal/ssa.a 29921002 29933976 +12974 +0.043% cmd/go/internal/modfetch/codehost.a 530602 530430 -172 -0.032% cmd/go/internal/modfetch.a 679664 679578 -86 -0.013% cmd/go/internal/modfile.a 411102 410928 -174 -0.042% cmd/go/internal/test.a 315218 315126 -92 -0.029% cmd/go/internal/tlog.a 183242 183256 +14 +0.008% cmd/go/internal/txtar.a 23148 23060 -88 -0.380% cmd/internal/bio.a 132064 132060 -4 -0.003% cmd/internal/buildid.a 107174 107172 -2 -0.002% cmd/internal/edit.a 33208 33354 +146 +0.440% cmd/internal/obj/arm.a 416488 416432 -56 -0.013% cmd/internal/obj/arm64.a 2772626 2772622 -4 -0.000% cmd/internal/obj/x86.a 923186 923114 -72 -0.008% cmd/internal/obj.a 679834 679836 +2 +0.000% cmd/internal/objfile.a 358374 358372 -2 -0.001% cmd/internal/test2json.a 67482 67434 -48 -0.071% cmd/link/internal/ld.a 2836280 2836110 -170 -0.006% cmd/link/internal/loadpe.a 148234 147736 -498 -0.336% cmd/link/internal/objfile.a 144534 144434 -100 -0.069% cmd/link/internal/ppc64.a 170876 170382 -494 -0.289% cmd/vendor/github.com/google/pprof/internal/elfexec.a 49896 49892 -4 -0.008% cmd/vendor/github.com/google/pprof/internal/graph.a 437478 437404 -74 -0.017% cmd/vendor/github.com/google/pprof/profile.a 902040 902044 +4 +0.000% cmd/vendor/github.com/ianlancetaylor/demangle.a 1217856 1217854 -2 -0.000% cmd/vendor/golang.org/x/arch/x86/x86asm.a 561332 560684 -648 -0.115% cmd/vendor/golang.org/x/crypto/ssh/terminal.a 153788 153784 -4 -0.003% cmd/vendor/golang.org/x/sys/unix.a 1043894 1043814 -80 -0.008% cmd/vendor/golang.org/x/tools/go/analysis/passes/printf.a 288458 288414 -44 -0.015% compress/flate.a 369024 368132 -892 -0.242% crypto/aes.a 109058 108968 -90 -0.083% crypto/cipher.a 150410 150544 +134 +0.089% crypto/elliptic.a 323572 323758 +186 +0.057% crypto/md5.a 50868 50788 -80 -0.157% crypto/rsa.a 195292 195214 -78 -0.040% crypto/sha1.a 70936 70858 -78 -0.110% crypto/sha256.a 75316 75236 -80 -0.106% crypto/sha512.a 84846 84768 -78 -0.092% crypto/subtle.a 6520 6514 -6 -0.092% crypto/tls.a 1654916 1654852 -64 -0.004% crypto/x509.a 888674 888638 -36 -0.004% database/sql.a 730280 730198 -82 -0.011% debug/gosym.a 184936 184862 -74 -0.040% debug/macho.a 272138 272136 -2 -0.001% debug/plan9obj.a 78444 78368 -76 -0.097% encoding/base64.a 82126 81882 -244 -0.297% encoding/binary.a 187196 187150 -46 -0.025% encoding/gob.a 897868 897870 +2 +0.000% encoding/json.a 659934 659832 -102 -0.015% encoding/pem.a 59138 58870 -268 -0.453% encoding/xml.a 694054 693300 -754 -0.109% fmt.a 484518 484196 -322 -0.066% go/format.a 33962 33994 +32 +0.094% go/printer.a 437132 437134 +2 +0.000% go/scanner.a 141774 141772 -2 -0.001% go/token.a 125130 125126 -4 -0.003% go/types.a 2192086 2191994 -92 -0.004% html/template.a 599038 598770 -268 -0.045% html.a 184842 184710 -132 -0.071% image/draw.a 129592 129238 -354 -0.273% image/gif.a 171824 171716 -108 -0.063% image/internal/imageutil.a 20282 19272 -1010 -4.980% image/jpeg.a 275608 275114 -494 -0.179% image/png.a 343416 343620 +204 +0.059% image.a 362244 362210 -34 -0.009% index/suffixarray.a 113040 112954 -86 -0.076% internal/trace.a 518972 518838 -134 -0.026% math/big.a 1012670 1012354 -316 -0.031% math.a 219338 219334 -4 -0.002% mime/multipart.a 178854 178502 -352 -0.197% mime/quotedprintable.a 49226 48936 -290 -0.589% net/http/cgi.a 172328 172324 -4 -0.002% net/http.a 4000180 3999732 -448 -0.011% net.a 1858330 1858252 -78 -0.004% path/filepath.a 107496 107498 +2 +0.002% reflect.a 1439776 1439994 +218 +0.015% regexp/syntax.a 459430 459432 +2 +0.000% regexp.a 416394 416400 +6 +0.001% runtime/debug.a 42106 42100 -6 -0.014% runtime/pprof/internal/profile.a 608718 608720 +2 +0.000% runtime/pprof.a 355474 355476 +2 +0.001% runtime.a 3555748 3555796 +48 +0.001% strconv.a 294432 294410 -22 -0.007% strings.a 292148 292090 -58 -0.020% syscall.a 859682 859470 -212 -0.025% text/tabwriter.a 65614 65148 -466 -0.710% vendor/golang.org/x/crypto/chacha20poly1305.a 126736 126728 -8 -0.006% vendor/golang.org/x/crypto/cryptobyte.a 269112 269114 +2 +0.001% vendor/golang.org/x/crypto/internal/chacha20.a 61842 61262 -580 -0.938% vendor/golang.org/x/crypto/poly1305.a 47410 47404 -6 -0.013% vendor/golang.org/x/net/dns/dnsmessage.a 628700 628012 -688 -0.109% vendor/golang.org/x/net/idna.a 237678 237826 +148 +0.062% vendor/golang.org/x/net/route.a 187852 187458 -394 -0.210% vendor/golang.org/x/sys/unix.a 1022426 1022348 -78 -0.008% vendor/golang.org/x/text/transform.a 117954 118104 +150 +0.127% vendor/golang.org/x/text/unicode/bidi.a 291398 291404 +6 +0.002% vendor/golang.org/x/text/unicode/norm.a 534640 534540 -100 -0.019% total 128945190 128945128 -62 -0.000% Change-Id: I346dc31356d5ef7774b824cf202169610bd26432 Reviewed-on: https://go-review.googlesource.com/c/go/+/175778 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-04 16:56:08 -07:00
// ^(x-1) == ^x+1 == -x
(Add(64|32|16|8) (Const(64|32|16|8) [1]) (Com(64|32|16|8) x)) => (Neg(64|32|16|8) x)
(Com(64|32|16|8) (Add(64|32|16|8) (Const(64|32|16|8) [-1]) x)) => (Neg(64|32|16|8) x)
cmd/compile: add more Neg/Com optimizations This is a grab-bag of minor optimizations. While we're here, document the c != -(1<<31) constraints better (#31888). file before after Δ % go 14669924 14665828 -4096 -0.028% asm 4867088 4858896 -8192 -0.168% compile 23988320 23984224 -4096 -0.017% cover 5210856 5206760 -4096 -0.079% link 6084376 6080280 -4096 -0.067% total 132181084 132156508 -24576 -0.019% file before after Δ % archive/tar.a 516708 516702 -6 -0.001% bufio.a 182200 181974 -226 -0.124% bytes.a 217624 216890 -734 -0.337% cmd/compile/internal/gc.a 8865412 8865228 -184 -0.002% cmd/compile/internal/ssa.a 29921002 29933976 +12974 +0.043% cmd/go/internal/modfetch/codehost.a 530602 530430 -172 -0.032% cmd/go/internal/modfetch.a 679664 679578 -86 -0.013% cmd/go/internal/modfile.a 411102 410928 -174 -0.042% cmd/go/internal/test.a 315218 315126 -92 -0.029% cmd/go/internal/tlog.a 183242 183256 +14 +0.008% cmd/go/internal/txtar.a 23148 23060 -88 -0.380% cmd/internal/bio.a 132064 132060 -4 -0.003% cmd/internal/buildid.a 107174 107172 -2 -0.002% cmd/internal/edit.a 33208 33354 +146 +0.440% cmd/internal/obj/arm.a 416488 416432 -56 -0.013% cmd/internal/obj/arm64.a 2772626 2772622 -4 -0.000% cmd/internal/obj/x86.a 923186 923114 -72 -0.008% cmd/internal/obj.a 679834 679836 +2 +0.000% cmd/internal/objfile.a 358374 358372 -2 -0.001% cmd/internal/test2json.a 67482 67434 -48 -0.071% cmd/link/internal/ld.a 2836280 2836110 -170 -0.006% cmd/link/internal/loadpe.a 148234 147736 -498 -0.336% cmd/link/internal/objfile.a 144534 144434 -100 -0.069% cmd/link/internal/ppc64.a 170876 170382 -494 -0.289% cmd/vendor/github.com/google/pprof/internal/elfexec.a 49896 49892 -4 -0.008% cmd/vendor/github.com/google/pprof/internal/graph.a 437478 437404 -74 -0.017% cmd/vendor/github.com/google/pprof/profile.a 902040 902044 +4 +0.000% cmd/vendor/github.com/ianlancetaylor/demangle.a 1217856 1217854 -2 -0.000% cmd/vendor/golang.org/x/arch/x86/x86asm.a 561332 560684 -648 -0.115% cmd/vendor/golang.org/x/crypto/ssh/terminal.a 153788 153784 -4 -0.003% cmd/vendor/golang.org/x/sys/unix.a 1043894 1043814 -80 -0.008% cmd/vendor/golang.org/x/tools/go/analysis/passes/printf.a 288458 288414 -44 -0.015% compress/flate.a 369024 368132 -892 -0.242% crypto/aes.a 109058 108968 -90 -0.083% crypto/cipher.a 150410 150544 +134 +0.089% crypto/elliptic.a 323572 323758 +186 +0.057% crypto/md5.a 50868 50788 -80 -0.157% crypto/rsa.a 195292 195214 -78 -0.040% crypto/sha1.a 70936 70858 -78 -0.110% crypto/sha256.a 75316 75236 -80 -0.106% crypto/sha512.a 84846 84768 -78 -0.092% crypto/subtle.a 6520 6514 -6 -0.092% crypto/tls.a 1654916 1654852 -64 -0.004% crypto/x509.a 888674 888638 -36 -0.004% database/sql.a 730280 730198 -82 -0.011% debug/gosym.a 184936 184862 -74 -0.040% debug/macho.a 272138 272136 -2 -0.001% debug/plan9obj.a 78444 78368 -76 -0.097% encoding/base64.a 82126 81882 -244 -0.297% encoding/binary.a 187196 187150 -46 -0.025% encoding/gob.a 897868 897870 +2 +0.000% encoding/json.a 659934 659832 -102 -0.015% encoding/pem.a 59138 58870 -268 -0.453% encoding/xml.a 694054 693300 -754 -0.109% fmt.a 484518 484196 -322 -0.066% go/format.a 33962 33994 +32 +0.094% go/printer.a 437132 437134 +2 +0.000% go/scanner.a 141774 141772 -2 -0.001% go/token.a 125130 125126 -4 -0.003% go/types.a 2192086 2191994 -92 -0.004% html/template.a 599038 598770 -268 -0.045% html.a 184842 184710 -132 -0.071% image/draw.a 129592 129238 -354 -0.273% image/gif.a 171824 171716 -108 -0.063% image/internal/imageutil.a 20282 19272 -1010 -4.980% image/jpeg.a 275608 275114 -494 -0.179% image/png.a 343416 343620 +204 +0.059% image.a 362244 362210 -34 -0.009% index/suffixarray.a 113040 112954 -86 -0.076% internal/trace.a 518972 518838 -134 -0.026% math/big.a 1012670 1012354 -316 -0.031% math.a 219338 219334 -4 -0.002% mime/multipart.a 178854 178502 -352 -0.197% mime/quotedprintable.a 49226 48936 -290 -0.589% net/http/cgi.a 172328 172324 -4 -0.002% net/http.a 4000180 3999732 -448 -0.011% net.a 1858330 1858252 -78 -0.004% path/filepath.a 107496 107498 +2 +0.002% reflect.a 1439776 1439994 +218 +0.015% regexp/syntax.a 459430 459432 +2 +0.000% regexp.a 416394 416400 +6 +0.001% runtime/debug.a 42106 42100 -6 -0.014% runtime/pprof/internal/profile.a 608718 608720 +2 +0.000% runtime/pprof.a 355474 355476 +2 +0.001% runtime.a 3555748 3555796 +48 +0.001% strconv.a 294432 294410 -22 -0.007% strings.a 292148 292090 -58 -0.020% syscall.a 859682 859470 -212 -0.025% text/tabwriter.a 65614 65148 -466 -0.710% vendor/golang.org/x/crypto/chacha20poly1305.a 126736 126728 -8 -0.006% vendor/golang.org/x/crypto/cryptobyte.a 269112 269114 +2 +0.001% vendor/golang.org/x/crypto/internal/chacha20.a 61842 61262 -580 -0.938% vendor/golang.org/x/crypto/poly1305.a 47410 47404 -6 -0.013% vendor/golang.org/x/net/dns/dnsmessage.a 628700 628012 -688 -0.109% vendor/golang.org/x/net/idna.a 237678 237826 +148 +0.062% vendor/golang.org/x/net/route.a 187852 187458 -394 -0.210% vendor/golang.org/x/sys/unix.a 1022426 1022348 -78 -0.008% vendor/golang.org/x/text/transform.a 117954 118104 +150 +0.127% vendor/golang.org/x/text/unicode/bidi.a 291398 291404 +6 +0.002% vendor/golang.org/x/text/unicode/norm.a 534640 534540 -100 -0.019% total 128945190 128945128 -62 -0.000% Change-Id: I346dc31356d5ef7774b824cf202169610bd26432 Reviewed-on: https://go-review.googlesource.com/c/go/+/175778 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-04 16:56:08 -07:00
// -(-x) == x
(Neg(64|32|16|8) (Neg(64|32|16|8) x)) => x
cmd/compile: add more Neg/Com optimizations This is a grab-bag of minor optimizations. While we're here, document the c != -(1<<31) constraints better (#31888). file before after Δ % go 14669924 14665828 -4096 -0.028% asm 4867088 4858896 -8192 -0.168% compile 23988320 23984224 -4096 -0.017% cover 5210856 5206760 -4096 -0.079% link 6084376 6080280 -4096 -0.067% total 132181084 132156508 -24576 -0.019% file before after Δ % archive/tar.a 516708 516702 -6 -0.001% bufio.a 182200 181974 -226 -0.124% bytes.a 217624 216890 -734 -0.337% cmd/compile/internal/gc.a 8865412 8865228 -184 -0.002% cmd/compile/internal/ssa.a 29921002 29933976 +12974 +0.043% cmd/go/internal/modfetch/codehost.a 530602 530430 -172 -0.032% cmd/go/internal/modfetch.a 679664 679578 -86 -0.013% cmd/go/internal/modfile.a 411102 410928 -174 -0.042% cmd/go/internal/test.a 315218 315126 -92 -0.029% cmd/go/internal/tlog.a 183242 183256 +14 +0.008% cmd/go/internal/txtar.a 23148 23060 -88 -0.380% cmd/internal/bio.a 132064 132060 -4 -0.003% cmd/internal/buildid.a 107174 107172 -2 -0.002% cmd/internal/edit.a 33208 33354 +146 +0.440% cmd/internal/obj/arm.a 416488 416432 -56 -0.013% cmd/internal/obj/arm64.a 2772626 2772622 -4 -0.000% cmd/internal/obj/x86.a 923186 923114 -72 -0.008% cmd/internal/obj.a 679834 679836 +2 +0.000% cmd/internal/objfile.a 358374 358372 -2 -0.001% cmd/internal/test2json.a 67482 67434 -48 -0.071% cmd/link/internal/ld.a 2836280 2836110 -170 -0.006% cmd/link/internal/loadpe.a 148234 147736 -498 -0.336% cmd/link/internal/objfile.a 144534 144434 -100 -0.069% cmd/link/internal/ppc64.a 170876 170382 -494 -0.289% cmd/vendor/github.com/google/pprof/internal/elfexec.a 49896 49892 -4 -0.008% cmd/vendor/github.com/google/pprof/internal/graph.a 437478 437404 -74 -0.017% cmd/vendor/github.com/google/pprof/profile.a 902040 902044 +4 +0.000% cmd/vendor/github.com/ianlancetaylor/demangle.a 1217856 1217854 -2 -0.000% cmd/vendor/golang.org/x/arch/x86/x86asm.a 561332 560684 -648 -0.115% cmd/vendor/golang.org/x/crypto/ssh/terminal.a 153788 153784 -4 -0.003% cmd/vendor/golang.org/x/sys/unix.a 1043894 1043814 -80 -0.008% cmd/vendor/golang.org/x/tools/go/analysis/passes/printf.a 288458 288414 -44 -0.015% compress/flate.a 369024 368132 -892 -0.242% crypto/aes.a 109058 108968 -90 -0.083% crypto/cipher.a 150410 150544 +134 +0.089% crypto/elliptic.a 323572 323758 +186 +0.057% crypto/md5.a 50868 50788 -80 -0.157% crypto/rsa.a 195292 195214 -78 -0.040% crypto/sha1.a 70936 70858 -78 -0.110% crypto/sha256.a 75316 75236 -80 -0.106% crypto/sha512.a 84846 84768 -78 -0.092% crypto/subtle.a 6520 6514 -6 -0.092% crypto/tls.a 1654916 1654852 -64 -0.004% crypto/x509.a 888674 888638 -36 -0.004% database/sql.a 730280 730198 -82 -0.011% debug/gosym.a 184936 184862 -74 -0.040% debug/macho.a 272138 272136 -2 -0.001% debug/plan9obj.a 78444 78368 -76 -0.097% encoding/base64.a 82126 81882 -244 -0.297% encoding/binary.a 187196 187150 -46 -0.025% encoding/gob.a 897868 897870 +2 +0.000% encoding/json.a 659934 659832 -102 -0.015% encoding/pem.a 59138 58870 -268 -0.453% encoding/xml.a 694054 693300 -754 -0.109% fmt.a 484518 484196 -322 -0.066% go/format.a 33962 33994 +32 +0.094% go/printer.a 437132 437134 +2 +0.000% go/scanner.a 141774 141772 -2 -0.001% go/token.a 125130 125126 -4 -0.003% go/types.a 2192086 2191994 -92 -0.004% html/template.a 599038 598770 -268 -0.045% html.a 184842 184710 -132 -0.071% image/draw.a 129592 129238 -354 -0.273% image/gif.a 171824 171716 -108 -0.063% image/internal/imageutil.a 20282 19272 -1010 -4.980% image/jpeg.a 275608 275114 -494 -0.179% image/png.a 343416 343620 +204 +0.059% image.a 362244 362210 -34 -0.009% index/suffixarray.a 113040 112954 -86 -0.076% internal/trace.a 518972 518838 -134 -0.026% math/big.a 1012670 1012354 -316 -0.031% math.a 219338 219334 -4 -0.002% mime/multipart.a 178854 178502 -352 -0.197% mime/quotedprintable.a 49226 48936 -290 -0.589% net/http/cgi.a 172328 172324 -4 -0.002% net/http.a 4000180 3999732 -448 -0.011% net.a 1858330 1858252 -78 -0.004% path/filepath.a 107496 107498 +2 +0.002% reflect.a 1439776 1439994 +218 +0.015% regexp/syntax.a 459430 459432 +2 +0.000% regexp.a 416394 416400 +6 +0.001% runtime/debug.a 42106 42100 -6 -0.014% runtime/pprof/internal/profile.a 608718 608720 +2 +0.000% runtime/pprof.a 355474 355476 +2 +0.001% runtime.a 3555748 3555796 +48 +0.001% strconv.a 294432 294410 -22 -0.007% strings.a 292148 292090 -58 -0.020% syscall.a 859682 859470 -212 -0.025% text/tabwriter.a 65614 65148 -466 -0.710% vendor/golang.org/x/crypto/chacha20poly1305.a 126736 126728 -8 -0.006% vendor/golang.org/x/crypto/cryptobyte.a 269112 269114 +2 +0.001% vendor/golang.org/x/crypto/internal/chacha20.a 61842 61262 -580 -0.938% vendor/golang.org/x/crypto/poly1305.a 47410 47404 -6 -0.013% vendor/golang.org/x/net/dns/dnsmessage.a 628700 628012 -688 -0.109% vendor/golang.org/x/net/idna.a 237678 237826 +148 +0.062% vendor/golang.org/x/net/route.a 187852 187458 -394 -0.210% vendor/golang.org/x/sys/unix.a 1022426 1022348 -78 -0.008% vendor/golang.org/x/text/transform.a 117954 118104 +150 +0.127% vendor/golang.org/x/text/unicode/bidi.a 291398 291404 +6 +0.002% vendor/golang.org/x/text/unicode/norm.a 534640 534540 -100 -0.019% total 128945190 128945128 -62 -0.000% Change-Id: I346dc31356d5ef7774b824cf202169610bd26432 Reviewed-on: https://go-review.googlesource.com/c/go/+/175778 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-04 16:56:08 -07:00
// -^x == x+1
(Neg(64|32|16|8) <t> (Com(64|32|16|8) x)) => (Add(64|32|16|8) (Const(64|32|16|8) <t> [1]) x)
(And(64|32|16|8) x (And(64|32|16|8) x y)) => (And(64|32|16|8) x y)
(Or(64|32|16|8) x (Or(64|32|16|8) x y)) => (Or(64|32|16|8) x y)
(Xor(64|32|16|8) x (Xor(64|32|16|8) x y)) => y
// Fold comparisons with numeric bounds
(Less(64|32|16|8)U _ (Const(64|32|16|8) [0])) => (ConstBool [false])
(Leq(64|32|16|8)U (Const(64|32|16|8) [0]) _) => (ConstBool [true])
(Less(64|32|16|8)U (Const(64|32|16|8) [-1]) _) => (ConstBool [false])
(Leq(64|32|16|8)U _ (Const(64|32|16|8) [-1])) => (ConstBool [true])
(Less64 _ (Const64 [math.MinInt64])) => (ConstBool [false])
(Less32 _ (Const32 [math.MinInt32])) => (ConstBool [false])
(Less16 _ (Const16 [math.MinInt16])) => (ConstBool [false])
(Less8 _ (Const8 [math.MinInt8 ])) => (ConstBool [false])
(Leq64 (Const64 [math.MinInt64]) _) => (ConstBool [true])
(Leq32 (Const32 [math.MinInt32]) _) => (ConstBool [true])
(Leq16 (Const16 [math.MinInt16]) _) => (ConstBool [true])
(Leq8 (Const8 [math.MinInt8 ]) _) => (ConstBool [true])
(Less64 (Const64 [math.MaxInt64]) _) => (ConstBool [false])
(Less32 (Const32 [math.MaxInt32]) _) => (ConstBool [false])
(Less16 (Const16 [math.MaxInt16]) _) => (ConstBool [false])
(Less8 (Const8 [math.MaxInt8 ]) _) => (ConstBool [false])
(Leq64 _ (Const64 [math.MaxInt64])) => (ConstBool [true])
(Leq32 _ (Const32 [math.MaxInt32])) => (ConstBool [true])
(Leq16 _ (Const16 [math.MaxInt16])) => (ConstBool [true])
(Leq8 _ (Const8 [math.MaxInt8 ])) => (ConstBool [true])
// Canonicalize <= on numeric bounds and < near numeric bounds to ==
(Leq(64|32|16|8)U x c:(Const(64|32|16|8) [0])) => (Eq(64|32|16|8) x c)
(Leq(64|32|16|8)U c:(Const(64|32|16|8) [-1]) x) => (Eq(64|32|16|8) x c)
(Less(64|32|16|8)U x (Const(64|32|16|8) <t> [1])) => (Eq(64|32|16|8) x (Const(64|32|16|8) <t> [0]))
(Less(64|32|16|8)U (Const(64|32|16|8) <t> [-2]) x) => (Eq(64|32|16|8) x (Const(64|32|16|8) <t> [-1]))
(Leq64 x c:(Const64 [math.MinInt64])) => (Eq64 x c)
(Leq32 x c:(Const32 [math.MinInt32])) => (Eq32 x c)
(Leq16 x c:(Const16 [math.MinInt16])) => (Eq16 x c)
(Leq8 x c:(Const8 [math.MinInt8 ])) => (Eq8 x c)
(Leq64 c:(Const64 [math.MaxInt64]) x) => (Eq64 x c)
(Leq32 c:(Const32 [math.MaxInt32]) x) => (Eq32 x c)
(Leq16 c:(Const16 [math.MaxInt16]) x) => (Eq16 x c)
(Leq8 c:(Const8 [math.MaxInt8 ]) x) => (Eq8 x c)
(Less64 x (Const64 <t> [math.MinInt64+1])) => (Eq64 x (Const64 <t> [math.MinInt64]))
(Less32 x (Const32 <t> [math.MinInt32+1])) => (Eq32 x (Const32 <t> [math.MinInt32]))
(Less16 x (Const16 <t> [math.MinInt16+1])) => (Eq16 x (Const16 <t> [math.MinInt16]))
(Less8 x (Const8 <t> [math.MinInt8 +1])) => (Eq8 x (Const8 <t> [math.MinInt8 ]))
(Less64 (Const64 <t> [math.MaxInt64-1]) x) => (Eq64 x (Const64 <t> [math.MaxInt64]))
(Less32 (Const32 <t> [math.MaxInt32-1]) x) => (Eq32 x (Const32 <t> [math.MaxInt32]))
(Less16 (Const16 <t> [math.MaxInt16-1]) x) => (Eq16 x (Const16 <t> [math.MaxInt16]))
(Less8 (Const8 <t> [math.MaxInt8 -1]) x) => (Eq8 x (Const8 <t> [math.MaxInt8 ]))
// Ands clear bits. Ors set bits.
// If a subsequent Or will set all the bits
// that an And cleared, we can skip the And.
// This happens in bitmasking code like:
// x &^= 3 << shift // clear two old bits
// x |= v << shift // set two new bits
// when shift is a small constant and v ends up a constant 3.
(Or8 (And8 x (Const8 [c2])) (Const8 <t> [c1])) && ^(c1 | c2) == 0 => (Or8 (Const8 <t> [c1]) x)
(Or16 (And16 x (Const16 [c2])) (Const16 <t> [c1])) && ^(c1 | c2) == 0 => (Or16 (Const16 <t> [c1]) x)
(Or32 (And32 x (Const32 [c2])) (Const32 <t> [c1])) && ^(c1 | c2) == 0 => (Or32 (Const32 <t> [c1]) x)
(Or64 (And64 x (Const64 [c2])) (Const64 <t> [c1])) && ^(c1 | c2) == 0 => (Or64 (Const64 <t> [c1]) x)
(Trunc64to8 (And64 (Const64 [y]) x)) && y&0xFF == 0xFF => (Trunc64to8 x)
(Trunc64to16 (And64 (Const64 [y]) x)) && y&0xFFFF == 0xFFFF => (Trunc64to16 x)
(Trunc64to32 (And64 (Const64 [y]) x)) && y&0xFFFFFFFF == 0xFFFFFFFF => (Trunc64to32 x)
(Trunc32to8 (And32 (Const32 [y]) x)) && y&0xFF == 0xFF => (Trunc32to8 x)
(Trunc32to16 (And32 (Const32 [y]) x)) && y&0xFFFF == 0xFFFF => (Trunc32to16 x)
(Trunc16to8 (And16 (Const16 [y]) x)) && y&0xFF == 0xFF => (Trunc16to8 x)
(ZeroExt8to64 (Trunc64to8 x:(Rsh64Ux64 _ (Const64 [s])))) && s >= 56 => x
(ZeroExt16to64 (Trunc64to16 x:(Rsh64Ux64 _ (Const64 [s])))) && s >= 48 => x
(ZeroExt32to64 (Trunc64to32 x:(Rsh64Ux64 _ (Const64 [s])))) && s >= 32 => x
(ZeroExt8to32 (Trunc32to8 x:(Rsh32Ux64 _ (Const64 [s])))) && s >= 24 => x
(ZeroExt16to32 (Trunc32to16 x:(Rsh32Ux64 _ (Const64 [s])))) && s >= 16 => x
(ZeroExt8to16 (Trunc16to8 x:(Rsh16Ux64 _ (Const64 [s])))) && s >= 8 => x
(SignExt8to64 (Trunc64to8 x:(Rsh64x64 _ (Const64 [s])))) && s >= 56 => x
(SignExt16to64 (Trunc64to16 x:(Rsh64x64 _ (Const64 [s])))) && s >= 48 => x
(SignExt32to64 (Trunc64to32 x:(Rsh64x64 _ (Const64 [s])))) && s >= 32 => x
(SignExt8to32 (Trunc32to8 x:(Rsh32x64 _ (Const64 [s])))) && s >= 24 => x
(SignExt16to32 (Trunc32to16 x:(Rsh32x64 _ (Const64 [s])))) && s >= 16 => x
(SignExt8to16 (Trunc16to8 x:(Rsh16x64 _ (Const64 [s])))) && s >= 8 => x
(Slicemask (Const32 [x])) && x > 0 => (Const32 [-1])
(Slicemask (Const32 [0])) => (Const32 [0])
(Slicemask (Const64 [x])) && x > 0 => (Const64 [-1])
(Slicemask (Const64 [0])) => (Const64 [0])
// simplifications often used for lengths. e.g. len(s[i:i+5])==5
(Sub(64|32|16|8) (Add(64|32|16|8) x y) x) => y
(Sub(64|32|16|8) (Add(64|32|16|8) x y) y) => x
cmd/compile: generic SSA rules for simplifying 2 and 3 operand integer arithmetic expressions This applies the following generic integer addition/subtraction transformations: x - (x + y) = -y (x - y) - x = -y y + (x - y) = x y + (z + (x - y) = x + z There's over 40 unique functions matching in Go. Hits 2 funcs in the runtime itself: runtime.stackfree() runtime.runqdrain() Go binary size reduced by 0.05% on Linux x86_64. StackCopy bench (perflocked Cascade Lake x86): name old time/op new time/op delta StackCopyPtr-8 87.3ms ± 1% 86.9ms ± 0% -0.45% (p=0.000 n=20+20) StackCopy-8 77.6ms ± 1% 77.0ms ± 0% -0.76% (p=0.000 n=20+20) StackCopyNoCache-8 2.28ms ± 2% 2.26ms ± 2% -0.93% (p=0.008 n=19+20) test/bench/go1 benchmarks (perflocked Cascade Lake x86): name old time/op new time/op delta BinaryTree17-8 1.88s ± 1% 1.88s ± 0% ~ (p=0.373 n=15+12) Fannkuch11-8 2.31s ± 0% 2.35s ± 0% +1.52% (p=0.000 n=15+14) FmtFprintfEmpty-8 26.6ns ± 0% 26.6ns ± 0% ~ (p=0.081 n=14+13) FmtFprintfString-8 48.6ns ± 0% 50.0ns ± 0% +2.86% (p=0.000 n=15+14) FmtFprintfInt-8 56.9ns ± 0% 54.8ns ± 0% -3.70% (p=0.000 n=15+15) FmtFprintfIntInt-8 90.4ns ± 0% 88.8ns ± 0% -1.78% (p=0.000 n=15+15) FmtFprintfPrefixedInt-8 104ns ± 0% 104ns ± 0% ~ (p=0.905 n=14+13) FmtFprintfFloat-8 148ns ± 0% 144ns ± 0% -2.19% (p=0.000 n=14+15) FmtManyArgs-8 389ns ± 0% 390ns ± 0% +0.35% (p=0.000 n=12+15) GobDecode-8 3.90ms ± 1% 3.88ms ± 0% -0.49% (p=0.000 n=15+14) GobEncode-8 2.73ms ± 0% 2.73ms ± 0% ~ (p=0.425 n=15+14) Gzip-8 169ms ± 0% 168ms ± 0% -0.52% (p=0.000 n=13+13) Gunzip-8 24.7ms ± 0% 24.8ms ± 0% +0.61% (p=0.000 n=15+15) HTTPClientServer-8 60.5µs ± 6% 60.4µs ± 7% ~ (p=0.595 n=15+15) JSONEncode-8 6.97ms ± 1% 6.93ms ± 0% -0.69% (p=0.000 n=14+14) JSONDecode-8 31.2ms ± 1% 30.8ms ± 1% -1.27% (p=0.000 n=14+14) Mandelbrot200-8 3.87ms ± 0% 3.87ms ± 0% ~ (p=0.652 n=15+14) GoParse-8 2.65ms ± 2% 2.64ms ± 1% ~ (p=0.202 n=15+15) RegexpMatchEasy0_32-8 45.1ns ± 0% 45.9ns ± 0% +1.68% (p=0.000 n=14+15) RegexpMatchEasy0_1K-8 140ns ± 0% 139ns ± 0% -0.44% (p=0.000 n=15+14) RegexpMatchEasy1_32-8 40.9ns ± 3% 40.5ns ± 0% -0.88% (p=0.000 n=15+13) RegexpMatchEasy1_1K-8 215ns ± 1% 220ns ± 1% +2.27% (p=0.000 n=15+15) RegexpMatchMedium_32-8 783ns ± 7% 738ns ± 0% ~ (p=0.361 n=15+15) RegexpMatchMedium_1K-8 24.1µs ± 6% 23.4µs ± 6% -2.94% (p=0.004 n=15+15) RegexpMatchHard_32-8 1.10µs ± 1% 1.09µs ± 1% -0.40% (p=0.006 n=15+14) RegexpMatchHard_1K-8 33.0µs ± 0% 33.0µs ± 0% ~ (p=0.535 n=12+14) Revcomp-8 354ms ± 0% 353ms ± 0% -0.23% (p=0.002 n=15+13) Template-8 42.0ms ± 1% 41.8ms ± 2% -0.37% (p=0.023 n=14+15) TimeParse-8 181ns ± 0% 180ns ± 1% -0.18% (p=0.014 n=12+13) TimeFormat-8 240ns ± 0% 242ns ± 1% +0.69% (p=0.000 n=12+15) [Geo mean] 35.2µs 35.1µs -0.43% name old speed new speed delta GobDecode-8 197MB/s ± 1% 198MB/s ± 0% +0.49% (p=0.000 n=15+14) GobEncode-8 281MB/s ± 0% 281MB/s ± 0% ~ (p=0.419 n=15+14) Gzip-8 115MB/s ± 0% 115MB/s ± 0% +0.52% (p=0.000 n=13+13) Gunzip-8 786MB/s ± 0% 781MB/s ± 0% -0.60% (p=0.000 n=15+15) JSONEncode-8 278MB/s ± 1% 280MB/s ± 0% +0.69% (p=0.000 n=14+14) JSONDecode-8 62.3MB/s ± 1% 63.1MB/s ± 1% +1.29% (p=0.000 n=14+14) GoParse-8 21.9MB/s ± 2% 22.0MB/s ± 1% ~ (p=0.205 n=15+15) RegexpMatchEasy0_32-8 709MB/s ± 0% 697MB/s ± 0% -1.65% (p=0.000 n=14+15) RegexpMatchEasy0_1K-8 7.34GB/s ± 0% 7.37GB/s ± 0% +0.43% (p=0.000 n=15+15) RegexpMatchEasy1_32-8 783MB/s ± 2% 790MB/s ± 0% +0.88% (p=0.000 n=15+13) RegexpMatchEasy1_1K-8 4.77GB/s ± 1% 4.66GB/s ± 1% -2.23% (p=0.000 n=15+15) RegexpMatchMedium_32-8 41.0MB/s ± 7% 43.3MB/s ± 0% ~ (p=0.360 n=15+15) RegexpMatchMedium_1K-8 42.5MB/s ± 6% 43.8MB/s ± 6% +3.07% (p=0.004 n=15+15) RegexpMatchHard_32-8 29.2MB/s ± 1% 29.3MB/s ± 1% +0.41% (p=0.006 n=15+14) RegexpMatchHard_1K-8 31.1MB/s ± 0% 31.1MB/s ± 0% ~ (p=0.495 n=12+14) Revcomp-8 718MB/s ± 0% 720MB/s ± 0% +0.23% (p=0.002 n=15+13) Template-8 46.3MB/s ± 1% 46.4MB/s ± 2% +0.38% (p=0.021 n=14+15) [Geo mean] 205MB/s 206MB/s +0.57% Change-Id: Ibd1afdf8b6c0b08087dcc3acd8f943637eb95ac0 Reviewed-on: https://go-review.googlesource.com/c/go/+/344930 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Go Bot <gobot@golang.org> Trust: Josh Bleecher Snyder <josharian@gmail.com>
2021-08-25 12:36:17 +02:00
(Sub(64|32|16|8) (Sub(64|32|16|8) x y) x) => (Neg(64|32|16|8) y)
(Sub(64|32|16|8) x (Add(64|32|16|8) x y)) => (Neg(64|32|16|8) y)
(Add(64|32|16|8) x (Sub(64|32|16|8) y x)) => y
(Add(64|32|16|8) x (Add(64|32|16|8) y (Sub(64|32|16|8) z x))) => (Add(64|32|16|8) y z)
// basic phi simplifications
(Phi (Const8 [c]) (Const8 [c])) => (Const8 [c])
(Phi (Const16 [c]) (Const16 [c])) => (Const16 [c])
(Phi (Const32 [c]) (Const32 [c])) => (Const32 [c])
(Phi (Const64 [c]) (Const64 [c])) => (Const64 [c])
// slice and interface comparisons
// The frontend ensures that we can only compare against nil,
// so we need only compare the first word (interface type or slice ptr).
(EqInter x y) => (EqPtr (ITab x) (ITab y))
(NeqInter x y) => (NeqPtr (ITab x) (ITab y))
(EqSlice x y) => (EqPtr (SlicePtr x) (SlicePtr y))
(NeqSlice x y) => (NeqPtr (SlicePtr x) (SlicePtr y))
// Load of store of same address, with compatibly typed value and same size
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> p1 (Store {t2} p2 x _))
&& isSamePtr(p1, p2)
&& t1.Compare(x.Type) == types.CMPeq
&& t1.Size() == t2.Size()
=> x
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> p1 (Store {t2} p2 _ (Store {t3} p3 x _)))
&& isSamePtr(p1, p3)
&& t1.Compare(x.Type) == types.CMPeq
&& t1.Size() == t2.Size()
&& disjoint(p3, t3.Size(), p2, t2.Size())
=> x
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> p1 (Store {t2} p2 _ (Store {t3} p3 _ (Store {t4} p4 x _))))
&& isSamePtr(p1, p4)
&& t1.Compare(x.Type) == types.CMPeq
&& t1.Size() == t2.Size()
&& disjoint(p4, t4.Size(), p2, t2.Size())
&& disjoint(p4, t4.Size(), p3, t3.Size())
=> x
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> p1 (Store {t2} p2 _ (Store {t3} p3 _ (Store {t4} p4 _ (Store {t5} p5 x _)))))
&& isSamePtr(p1, p5)
&& t1.Compare(x.Type) == types.CMPeq
&& t1.Size() == t2.Size()
&& disjoint(p5, t5.Size(), p2, t2.Size())
&& disjoint(p5, t5.Size(), p3, t3.Size())
&& disjoint(p5, t5.Size(), p4, t4.Size())
=> x
// Pass constants through math.Float{32,64}bits and math.Float{32,64}frombits
(Load <t1> p1 (Store {t2} p2 (Const64 [x]) _)) && isSamePtr(p1,p2) && t2.Size() == 8 && is64BitFloat(t1) && !math.IsNaN(math.Float64frombits(uint64(x))) => (Const64F [math.Float64frombits(uint64(x))])
(Load <t1> p1 (Store {t2} p2 (Const32 [x]) _)) && isSamePtr(p1,p2) && t2.Size() == 4 && is32BitFloat(t1) && !math.IsNaN(float64(math.Float32frombits(uint32(x)))) => (Const32F [math.Float32frombits(uint32(x))])
(Load <t1> p1 (Store {t2} p2 (Const64F [x]) _)) && isSamePtr(p1,p2) && t2.Size() == 8 && is64BitInt(t1) => (Const64 [int64(math.Float64bits(x))])
(Load <t1> p1 (Store {t2} p2 (Const32F [x]) _)) && isSamePtr(p1,p2) && t2.Size() == 4 && is32BitInt(t1) => (Const32 [int32(math.Float32bits(x))])
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Float Loads up to Zeros so they can be constant folded.
(Load <t1> op:(OffPtr [o1] p1)
(Store {t2} p2 _
mem:(Zero [n] p3 _)))
&& o1 >= 0 && o1+t1.Size() <= n && isSamePtr(p1, p3)
&& CanSSA(t1)
&& disjoint(op, t1.Size(), p2, t2.Size())
=> @mem.Block (Load <t1> (OffPtr <op.Type> [o1] p3) mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> op:(OffPtr [o1] p1)
(Store {t2} p2 _
(Store {t3} p3 _
mem:(Zero [n] p4 _))))
&& o1 >= 0 && o1+t1.Size() <= n && isSamePtr(p1, p4)
&& CanSSA(t1)
&& disjoint(op, t1.Size(), p2, t2.Size())
&& disjoint(op, t1.Size(), p3, t3.Size())
=> @mem.Block (Load <t1> (OffPtr <op.Type> [o1] p4) mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> op:(OffPtr [o1] p1)
(Store {t2} p2 _
(Store {t3} p3 _
(Store {t4} p4 _
mem:(Zero [n] p5 _)))))
&& o1 >= 0 && o1+t1.Size() <= n && isSamePtr(p1, p5)
&& CanSSA(t1)
&& disjoint(op, t1.Size(), p2, t2.Size())
&& disjoint(op, t1.Size(), p3, t3.Size())
&& disjoint(op, t1.Size(), p4, t4.Size())
=> @mem.Block (Load <t1> (OffPtr <op.Type> [o1] p5) mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> op:(OffPtr [o1] p1)
(Store {t2} p2 _
(Store {t3} p3 _
(Store {t4} p4 _
(Store {t5} p5 _
mem:(Zero [n] p6 _))))))
&& o1 >= 0 && o1+t1.Size() <= n && isSamePtr(p1, p6)
&& CanSSA(t1)
&& disjoint(op, t1.Size(), p2, t2.Size())
&& disjoint(op, t1.Size(), p3, t3.Size())
&& disjoint(op, t1.Size(), p4, t4.Size())
&& disjoint(op, t1.Size(), p5, t5.Size())
=> @mem.Block (Load <t1> (OffPtr <op.Type> [o1] p6) mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Zero to Load forwarding.
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& t1.IsBoolean()
&& isSamePtr(p1, p2)
&& n >= o + 1
=> (ConstBool [false])
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is8BitInt(t1)
&& isSamePtr(p1, p2)
&& n >= o + 1
=> (Const8 [0])
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is16BitInt(t1)
&& isSamePtr(p1, p2)
&& n >= o + 2
=> (Const16 [0])
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is32BitInt(t1)
&& isSamePtr(p1, p2)
&& n >= o + 4
=> (Const32 [0])
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is64BitInt(t1)
&& isSamePtr(p1, p2)
&& n >= o + 8
=> (Const64 [0])
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is32BitFloat(t1)
&& isSamePtr(p1, p2)
&& n >= o + 4
=> (Const32F [0])
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is64BitFloat(t1)
&& isSamePtr(p1, p2)
&& n >= o + 8
=> (Const64F [0])
// Eliminate stores of values that have just been loaded from the same location.
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// We also handle the common case where there are some intermediate stores.
(Store {t1} p1 (Load <t2> p2 mem) mem)
&& isSamePtr(p1, p2)
&& t2.Size() == t1.Size()
=> mem
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} p1 (Load <t2> p2 oldmem) mem:(Store {t3} p3 _ oldmem))
&& isSamePtr(p1, p2)
&& t2.Size() == t1.Size()
&& disjoint(p1, t1.Size(), p3, t3.Size())
=> mem
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} p1 (Load <t2> p2 oldmem) mem:(Store {t3} p3 _ (Store {t4} p4 _ oldmem)))
&& isSamePtr(p1, p2)
&& t2.Size() == t1.Size()
&& disjoint(p1, t1.Size(), p3, t3.Size())
&& disjoint(p1, t1.Size(), p4, t4.Size())
=> mem
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} p1 (Load <t2> p2 oldmem) mem:(Store {t3} p3 _ (Store {t4} p4 _ (Store {t5} p5 _ oldmem))))
&& isSamePtr(p1, p2)
&& t2.Size() == t1.Size()
&& disjoint(p1, t1.Size(), p3, t3.Size())
&& disjoint(p1, t1.Size(), p4, t4.Size())
&& disjoint(p1, t1.Size(), p5, t5.Size())
=> mem
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Don't Store zeros to cleared variables.
(Store {t} (OffPtr [o] p1) x mem:(Zero [n] p2 _))
&& isConstZero(x)
&& o >= 0 && t.Size() + o <= n && isSamePtr(p1, p2)
=> mem
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} op:(OffPtr [o1] p1) x mem:(Store {t2} p2 _ (Zero [n] p3 _)))
&& isConstZero(x)
&& o1 >= 0 && t1.Size() + o1 <= n && isSamePtr(p1, p3)
&& disjoint(op, t1.Size(), p2, t2.Size())
=> mem
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} op:(OffPtr [o1] p1) x mem:(Store {t2} p2 _ (Store {t3} p3 _ (Zero [n] p4 _))))
&& isConstZero(x)
&& o1 >= 0 && t1.Size() + o1 <= n && isSamePtr(p1, p4)
&& disjoint(op, t1.Size(), p2, t2.Size())
&& disjoint(op, t1.Size(), p3, t3.Size())
=> mem
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} op:(OffPtr [o1] p1) x mem:(Store {t2} p2 _ (Store {t3} p3 _ (Store {t4} p4 _ (Zero [n] p5 _)))))
&& isConstZero(x)
&& o1 >= 0 && t1.Size() + o1 <= n && isSamePtr(p1, p5)
&& disjoint(op, t1.Size(), p2, t2.Size())
&& disjoint(op, t1.Size(), p3, t3.Size())
&& disjoint(op, t1.Size(), p4, t4.Size())
=> mem
// Collapse OffPtr
cmd/compile: disallow rewrite rules from declaring reserved names If I change a rule in ARM64.rules to use the variable name "b" in a conflicting way, rulegen would previously not complain, and the compiler would later give a confusing error: $ go run *.go && go build cmd/compile/internal/ssa # cmd/compile/internal/ssa ../rewriteARM64.go:24236:10: b.NewValue0 undefined (type int64 has no field or method NewValue0) Make rulegen complain early about those cases. Sometimes they might happen to be harmless, but in general they can easily cause confusion or unintended effect due to shadowing. After the change, with the same conflicting rule: $ go run *.go && go build cmd/compile/internal/ssa 2021/03/22 11:31:49 rule ARM64.rules:495 uses the reserved name b exit status 1 Note that 24 existing rules were using reserved names. It seems like the shadowing was harmless, as it wasn't causing typechecking issues nor did it seem to cause unintended behavior when the rule rewrite code ran. The bool values "b" were renamed "t", since that seems to have a precedent in other rules and in the fmt package. Sequential values like "a b c" were renamed to "x y z", since "b" is reserved. Finally, "typ" was renamed to "_typ", since there doesn't seem to be an obviously better answer. Passes all three of: $ GOARCH=amd64 go build -toolexec 'toolstash -cmp' -a std $ GOARCH=arm64 go build -toolexec 'toolstash -cmp' -a std $ GOARCH=mips64 go build -toolexec 'toolstash -cmp' -a std Fixes #45154. Change-Id: I1cce194dc7b477886a9c218c17973e996bcedccf Reviewed-on: https://go-review.googlesource.com/c/go/+/303549 Trust: Daniel Martí <mvdan@mvdan.cc> Run-TryBot: Daniel Martí <mvdan@mvdan.cc> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2021-03-22 11:35:02 +01:00
(OffPtr (OffPtr p [y]) [x]) => (OffPtr p [x+y])
(OffPtr p [0]) && v.Type.Compare(p.Type) == types.CMPeq => p
// indexing operations
// Note: bounds check has already been done
(PtrIndex <t> ptr idx) && config.PtrSize == 4 && is32Bit(t.Elem().Size()) => (AddPtr ptr (Mul32 <typ.Int> idx (Const32 <typ.Int> [int32(t.Elem().Size())])))
(PtrIndex <t> ptr idx) && config.PtrSize == 8 => (AddPtr ptr (Mul64 <typ.Int> idx (Const64 <typ.Int> [t.Elem().Size()])))
// struct operations
(StructSelect [i] x:(StructMake ___)) => x.Args[i]
(Load <t> _ _) && t.IsStruct() && CanSSA(t) => rewriteStructLoad(v)
(Store _ (StructMake ___) _) => rewriteStructStore(v)
(StructSelect [i] x:(Load <t> ptr mem)) && !CanSSA(t) =>
@x.Block (Load <v.Type> (OffPtr <v.Type.PtrTo()> [t.FieldOff(int(i))] ptr) mem)
// Putting struct{*byte} and similar into direct interfaces.
(IMake _typ (StructMake val)) => (IMake _typ val)
(StructSelect [0] (IData x)) => (IData x)
cmd/compile: better job of naming compound types Compound AUTO types weren't named previously. That was because live variable analysis (plive.go) doesn't handle spilling to compound types. It can't handle them because there is no valid place to put VARDEFs when regalloc is spilling compound types. compound types = multiword builtin types: complex, string, slice, and interface. Instead, we split named AUTOs into individual one-word variables. For example, a string s gets split into a byte ptr s.ptr and an integer s.len. Those two variables can be spilled to / restored from independently. As a result, live variable analysis can handle them because they are one-word objects. This CL will change how AUTOs are described in DWARF information. Consider the code: func f(s string, i int) int { x := s[i:i+5] g() return lookup(x) } The old compiler would spill x to two consecutive slots on the stack, both named x (at offsets 0 and 8). The new compiler spills the pointer of x to a slot named x.ptr. It doesn't spill x.len at all, as it is a constant (5) and can be rematerialized for the call to lookup. So compound objects may not be spilled in their entirety, and even if they are they won't necessarily be contiguous. Such is the price of optimization. Re-enable live variable analysis tests. One test remains disabled, it fails because of #14904. Change-Id: I8ef2b5ab91e43a0d2136bfc231c05d100ec0b801 Reviewed-on: https://go-review.googlesource.com/21233 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-03-28 11:25:17 -07:00
// un-SSAable values use mem->mem copies
(Store {t} dst (Load src mem) mem) && !CanSSA(t) =>
(Move {t} [t.Size()] dst src mem)
(Store {t} dst (Load src mem) (VarDef {x} mem)) && !CanSSA(t) =>
(Move {t} [t.Size()] dst src (VarDef {x} mem))
// array ops
(ArraySelect (ArrayMake1 x)) => x
(Load <t> _ _) && t.IsArray() && t.NumElem() == 0 =>
(ArrayMake0)
(Load <t> ptr mem) && t.IsArray() && t.NumElem() == 1 && CanSSA(t) =>
(ArrayMake1 (Load <t.Elem()> ptr mem))
(Store _ (ArrayMake0) mem) => mem
(Store dst (ArrayMake1 e) mem) => (Store {e.Type} dst e mem)
// Putting [1]*byte and similar into direct interfaces.
cmd/compile: disallow rewrite rules from declaring reserved names If I change a rule in ARM64.rules to use the variable name "b" in a conflicting way, rulegen would previously not complain, and the compiler would later give a confusing error: $ go run *.go && go build cmd/compile/internal/ssa # cmd/compile/internal/ssa ../rewriteARM64.go:24236:10: b.NewValue0 undefined (type int64 has no field or method NewValue0) Make rulegen complain early about those cases. Sometimes they might happen to be harmless, but in general they can easily cause confusion or unintended effect due to shadowing. After the change, with the same conflicting rule: $ go run *.go && go build cmd/compile/internal/ssa 2021/03/22 11:31:49 rule ARM64.rules:495 uses the reserved name b exit status 1 Note that 24 existing rules were using reserved names. It seems like the shadowing was harmless, as it wasn't causing typechecking issues nor did it seem to cause unintended behavior when the rule rewrite code ran. The bool values "b" were renamed "t", since that seems to have a precedent in other rules and in the fmt package. Sequential values like "a b c" were renamed to "x y z", since "b" is reserved. Finally, "typ" was renamed to "_typ", since there doesn't seem to be an obviously better answer. Passes all three of: $ GOARCH=amd64 go build -toolexec 'toolstash -cmp' -a std $ GOARCH=arm64 go build -toolexec 'toolstash -cmp' -a std $ GOARCH=mips64 go build -toolexec 'toolstash -cmp' -a std Fixes #45154. Change-Id: I1cce194dc7b477886a9c218c17973e996bcedccf Reviewed-on: https://go-review.googlesource.com/c/go/+/303549 Trust: Daniel Martí <mvdan@mvdan.cc> Run-TryBot: Daniel Martí <mvdan@mvdan.cc> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2021-03-22 11:35:02 +01:00
(IMake _typ (ArrayMake1 val)) => (IMake _typ val)
(ArraySelect [0] (IData x)) => (IData x)
// string ops
// Decomposing StringMake and lowering of StringPtr and StringLen
// happens in a later pass, dec, so that these operations are available
cmd/compile: better job of naming compound types Compound AUTO types weren't named previously. That was because live variable analysis (plive.go) doesn't handle spilling to compound types. It can't handle them because there is no valid place to put VARDEFs when regalloc is spilling compound types. compound types = multiword builtin types: complex, string, slice, and interface. Instead, we split named AUTOs into individual one-word variables. For example, a string s gets split into a byte ptr s.ptr and an integer s.len. Those two variables can be spilled to / restored from independently. As a result, live variable analysis can handle them because they are one-word objects. This CL will change how AUTOs are described in DWARF information. Consider the code: func f(s string, i int) int { x := s[i:i+5] g() return lookup(x) } The old compiler would spill x to two consecutive slots on the stack, both named x (at offsets 0 and 8). The new compiler spills the pointer of x to a slot named x.ptr. It doesn't spill x.len at all, as it is a constant (5) and can be rematerialized for the call to lookup. So compound objects may not be spilled in their entirety, and even if they are they won't necessarily be contiguous. Such is the price of optimization. Re-enable live variable analysis tests. One test remains disabled, it fails because of #14904. Change-Id: I8ef2b5ab91e43a0d2136bfc231c05d100ec0b801 Reviewed-on: https://go-review.googlesource.com/21233 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-03-28 11:25:17 -07:00
// to other passes for optimizations.
(StringPtr (StringMake (Addr <t> {s} base) _)) => (Addr <t> {s} base)
(StringLen (StringMake _ (Const64 <t> [c]))) => (Const64 <t> [c])
cmd/compile: start implementing strongly typed aux and auxint fields Right now the Aux and AuxInt fields of ssa.Values are typed as interface{} and int64, respectively. Each rule that uses these values must cast them to the type they actually are (*obj.LSym, or int32, or ValAndOff, etc.), use them, and then cast them back to interface{} or int64. We know for each opcode what the types of the Aux and AuxInt fields should be. So let's modify the rule generator to declare the types to be what we know they should be, autoconverting to and from the generic types for us. That way we can make the rules more type safe. It's difficult to make a single CL for this, so I've coopted the "=>" token to indicate a rule that is strongly typed. "->" rules are processed as before. That will let us migrate a few rules at a time in separate CLs. Hopefully we can reach a state where all rules are strongly typed and we can drop the distinction. This CL changes just a few rules to get a feel for what this transition would look like. I've decided not to put explicit types in the rules. I think it makes the rules somewhat clearer, but definitely more verbose. In particular, the passthrough rules that don't modify the fields in question are verbose for no real reason. Change-Id: I63a1b789ac5702e7caf7934cd49f784235d1d73d Reviewed-on: https://go-review.googlesource.com/c/go/+/190197 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2020-03-19 16:25:08 -07:00
(ConstString {str}) && config.PtrSize == 4 && str == "" =>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(StringMake (ConstNil) (Const32 <typ.Int> [0]))
cmd/compile: start implementing strongly typed aux and auxint fields Right now the Aux and AuxInt fields of ssa.Values are typed as interface{} and int64, respectively. Each rule that uses these values must cast them to the type they actually are (*obj.LSym, or int32, or ValAndOff, etc.), use them, and then cast them back to interface{} or int64. We know for each opcode what the types of the Aux and AuxInt fields should be. So let's modify the rule generator to declare the types to be what we know they should be, autoconverting to and from the generic types for us. That way we can make the rules more type safe. It's difficult to make a single CL for this, so I've coopted the "=>" token to indicate a rule that is strongly typed. "->" rules are processed as before. That will let us migrate a few rules at a time in separate CLs. Hopefully we can reach a state where all rules are strongly typed and we can drop the distinction. This CL changes just a few rules to get a feel for what this transition would look like. I've decided not to put explicit types in the rules. I think it makes the rules somewhat clearer, but definitely more verbose. In particular, the passthrough rules that don't modify the fields in question are verbose for no real reason. Change-Id: I63a1b789ac5702e7caf7934cd49f784235d1d73d Reviewed-on: https://go-review.googlesource.com/c/go/+/190197 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2020-03-19 16:25:08 -07:00
(ConstString {str}) && config.PtrSize == 8 && str == "" =>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(StringMake (ConstNil) (Const64 <typ.Int> [0]))
cmd/compile: start implementing strongly typed aux and auxint fields Right now the Aux and AuxInt fields of ssa.Values are typed as interface{} and int64, respectively. Each rule that uses these values must cast them to the type they actually are (*obj.LSym, or int32, or ValAndOff, etc.), use them, and then cast them back to interface{} or int64. We know for each opcode what the types of the Aux and AuxInt fields should be. So let's modify the rule generator to declare the types to be what we know they should be, autoconverting to and from the generic types for us. That way we can make the rules more type safe. It's difficult to make a single CL for this, so I've coopted the "=>" token to indicate a rule that is strongly typed. "->" rules are processed as before. That will let us migrate a few rules at a time in separate CLs. Hopefully we can reach a state where all rules are strongly typed and we can drop the distinction. This CL changes just a few rules to get a feel for what this transition would look like. I've decided not to put explicit types in the rules. I think it makes the rules somewhat clearer, but definitely more verbose. In particular, the passthrough rules that don't modify the fields in question are verbose for no real reason. Change-Id: I63a1b789ac5702e7caf7934cd49f784235d1d73d Reviewed-on: https://go-review.googlesource.com/c/go/+/190197 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2020-03-19 16:25:08 -07:00
(ConstString {str}) && config.PtrSize == 4 && str != "" =>
(StringMake
cmd/compile: start implementing strongly typed aux and auxint fields Right now the Aux and AuxInt fields of ssa.Values are typed as interface{} and int64, respectively. Each rule that uses these values must cast them to the type they actually are (*obj.LSym, or int32, or ValAndOff, etc.), use them, and then cast them back to interface{} or int64. We know for each opcode what the types of the Aux and AuxInt fields should be. So let's modify the rule generator to declare the types to be what we know they should be, autoconverting to and from the generic types for us. That way we can make the rules more type safe. It's difficult to make a single CL for this, so I've coopted the "=>" token to indicate a rule that is strongly typed. "->" rules are processed as before. That will let us migrate a few rules at a time in separate CLs. Hopefully we can reach a state where all rules are strongly typed and we can drop the distinction. This CL changes just a few rules to get a feel for what this transition would look like. I've decided not to put explicit types in the rules. I think it makes the rules somewhat clearer, but definitely more verbose. In particular, the passthrough rules that don't modify the fields in question are verbose for no real reason. Change-Id: I63a1b789ac5702e7caf7934cd49f784235d1d73d Reviewed-on: https://go-review.googlesource.com/c/go/+/190197 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2020-03-19 16:25:08 -07:00
(Addr <typ.BytePtr> {fe.StringData(str)}
(SB))
cmd/compile: start implementing strongly typed aux and auxint fields Right now the Aux and AuxInt fields of ssa.Values are typed as interface{} and int64, respectively. Each rule that uses these values must cast them to the type they actually are (*obj.LSym, or int32, or ValAndOff, etc.), use them, and then cast them back to interface{} or int64. We know for each opcode what the types of the Aux and AuxInt fields should be. So let's modify the rule generator to declare the types to be what we know they should be, autoconverting to and from the generic types for us. That way we can make the rules more type safe. It's difficult to make a single CL for this, so I've coopted the "=>" token to indicate a rule that is strongly typed. "->" rules are processed as before. That will let us migrate a few rules at a time in separate CLs. Hopefully we can reach a state where all rules are strongly typed and we can drop the distinction. This CL changes just a few rules to get a feel for what this transition would look like. I've decided not to put explicit types in the rules. I think it makes the rules somewhat clearer, but definitely more verbose. In particular, the passthrough rules that don't modify the fields in question are verbose for no real reason. Change-Id: I63a1b789ac5702e7caf7934cd49f784235d1d73d Reviewed-on: https://go-review.googlesource.com/c/go/+/190197 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2020-03-19 16:25:08 -07:00
(Const32 <typ.Int> [int32(len(str))]))
(ConstString {str}) && config.PtrSize == 8 && str != "" =>
(StringMake
cmd/compile: start implementing strongly typed aux and auxint fields Right now the Aux and AuxInt fields of ssa.Values are typed as interface{} and int64, respectively. Each rule that uses these values must cast them to the type they actually are (*obj.LSym, or int32, or ValAndOff, etc.), use them, and then cast them back to interface{} or int64. We know for each opcode what the types of the Aux and AuxInt fields should be. So let's modify the rule generator to declare the types to be what we know they should be, autoconverting to and from the generic types for us. That way we can make the rules more type safe. It's difficult to make a single CL for this, so I've coopted the "=>" token to indicate a rule that is strongly typed. "->" rules are processed as before. That will let us migrate a few rules at a time in separate CLs. Hopefully we can reach a state where all rules are strongly typed and we can drop the distinction. This CL changes just a few rules to get a feel for what this transition would look like. I've decided not to put explicit types in the rules. I think it makes the rules somewhat clearer, but definitely more verbose. In particular, the passthrough rules that don't modify the fields in question are verbose for no real reason. Change-Id: I63a1b789ac5702e7caf7934cd49f784235d1d73d Reviewed-on: https://go-review.googlesource.com/c/go/+/190197 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2020-03-19 16:25:08 -07:00
(Addr <typ.BytePtr> {fe.StringData(str)}
(SB))
cmd/compile: start implementing strongly typed aux and auxint fields Right now the Aux and AuxInt fields of ssa.Values are typed as interface{} and int64, respectively. Each rule that uses these values must cast them to the type they actually are (*obj.LSym, or int32, or ValAndOff, etc.), use them, and then cast them back to interface{} or int64. We know for each opcode what the types of the Aux and AuxInt fields should be. So let's modify the rule generator to declare the types to be what we know they should be, autoconverting to and from the generic types for us. That way we can make the rules more type safe. It's difficult to make a single CL for this, so I've coopted the "=>" token to indicate a rule that is strongly typed. "->" rules are processed as before. That will let us migrate a few rules at a time in separate CLs. Hopefully we can reach a state where all rules are strongly typed and we can drop the distinction. This CL changes just a few rules to get a feel for what this transition would look like. I've decided not to put explicit types in the rules. I think it makes the rules somewhat clearer, but definitely more verbose. In particular, the passthrough rules that don't modify the fields in question are verbose for no real reason. Change-Id: I63a1b789ac5702e7caf7934cd49f784235d1d73d Reviewed-on: https://go-review.googlesource.com/c/go/+/190197 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2020-03-19 16:25:08 -07:00
(Const64 <typ.Int> [int64(len(str))]))
// slice ops
cmd/compile: better job of naming compound types Compound AUTO types weren't named previously. That was because live variable analysis (plive.go) doesn't handle spilling to compound types. It can't handle them because there is no valid place to put VARDEFs when regalloc is spilling compound types. compound types = multiword builtin types: complex, string, slice, and interface. Instead, we split named AUTOs into individual one-word variables. For example, a string s gets split into a byte ptr s.ptr and an integer s.len. Those two variables can be spilled to / restored from independently. As a result, live variable analysis can handle them because they are one-word objects. This CL will change how AUTOs are described in DWARF information. Consider the code: func f(s string, i int) int { x := s[i:i+5] g() return lookup(x) } The old compiler would spill x to two consecutive slots on the stack, both named x (at offsets 0 and 8). The new compiler spills the pointer of x to a slot named x.ptr. It doesn't spill x.len at all, as it is a constant (5) and can be rematerialized for the call to lookup. So compound objects may not be spilled in their entirety, and even if they are they won't necessarily be contiguous. Such is the price of optimization. Re-enable live variable analysis tests. One test remains disabled, it fails because of #14904. Change-Id: I8ef2b5ab91e43a0d2136bfc231c05d100ec0b801 Reviewed-on: https://go-review.googlesource.com/21233 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-03-28 11:25:17 -07:00
// Only a few slice rules are provided here. See dec.rules for
// a more comprehensive set.
(SliceLen (SliceMake _ (Const64 <t> [c]) _)) => (Const64 <t> [c])
(SliceCap (SliceMake _ _ (Const64 <t> [c]))) => (Const64 <t> [c])
(SliceLen (SliceMake _ (Const32 <t> [c]) _)) => (Const32 <t> [c])
(SliceCap (SliceMake _ _ (Const32 <t> [c]))) => (Const32 <t> [c])
(SlicePtr (SliceMake (SlicePtr x) _ _)) => (SlicePtr x)
(SliceLen (SliceMake _ (SliceLen x) _)) => (SliceLen x)
(SliceCap (SliceMake _ _ (SliceCap x))) => (SliceCap x)
(SliceCap (SliceMake _ _ (SliceLen x))) => (SliceLen x)
(ConstSlice) && config.PtrSize == 4 =>
(SliceMake
(ConstNil <v.Type.Elem().PtrTo()>)
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Const32 <typ.Int> [0])
(Const32 <typ.Int> [0]))
(ConstSlice) && config.PtrSize == 8 =>
(SliceMake
(ConstNil <v.Type.Elem().PtrTo()>)
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Const64 <typ.Int> [0])
(Const64 <typ.Int> [0]))
// interface ops
(ConstInterface) =>
(IMake
(ConstNil <typ.Uintptr>)
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(ConstNil <typ.BytePtr>))
(NilCheck ptr:(GetG mem) mem) => ptr
(If (Not cond) yes no) => (If cond no yes)
(If (ConstBool [c]) yes no) && c => (First yes no)
(If (ConstBool [c]) yes no) && !c => (First no yes)
(Phi <t> nx:(Not x) ny:(Not y)) && nx.Uses == 1 && ny.Uses == 1 => (Not (Phi <t> x y))
// Get rid of Convert ops for pointer arithmetic on unsafe.Pointer.
(Convert (Add(64|32) (Convert ptr mem) off) mem) => (AddPtr ptr off)
(Convert (Convert ptr mem) mem) => ptr
// Note: it is important that the target rewrite is ptr+(off1+off2), not (ptr+off1)+off2.
// We must ensure that no intermediate computations are invalid pointers.
(Convert a:(Add(64|32) (Add(64|32) (Convert ptr mem) off1) off2) mem) => (AddPtr ptr (Add(64|32) <a.Type> off1 off2))
// strength reduction of divide by a constant.
// See ../magic.go for a detailed description of these algorithms.
// Unsigned divide by power of 2. Strength reduce to a shift.
(Div8u n (Const8 [c])) && isPowerOfTwo(c) => (Rsh8Ux64 n (Const64 <typ.UInt64> [log8(c)]))
(Div16u n (Const16 [c])) && isPowerOfTwo(c) => (Rsh16Ux64 n (Const64 <typ.UInt64> [log16(c)]))
(Div32u n (Const32 [c])) && isPowerOfTwo(c) => (Rsh32Ux64 n (Const64 <typ.UInt64> [log32(c)]))
(Div64u n (Const64 [c])) && isPowerOfTwo(c) => (Rsh64Ux64 n (Const64 <typ.UInt64> [log64(c)]))
(Div64u n (Const64 [-1<<63])) => (Rsh64Ux64 n (Const64 <typ.UInt64> [63]))
// Signed non-negative divide by power of 2.
(Div8 n (Const8 [c])) && isNonNegative(n) && isPowerOfTwo(c) => (Rsh8Ux64 n (Const64 <typ.UInt64> [log8(c)]))
(Div16 n (Const16 [c])) && isNonNegative(n) && isPowerOfTwo(c) => (Rsh16Ux64 n (Const64 <typ.UInt64> [log16(c)]))
(Div32 n (Const32 [c])) && isNonNegative(n) && isPowerOfTwo(c) => (Rsh32Ux64 n (Const64 <typ.UInt64> [log32(c)]))
(Div64 n (Const64 [c])) && isNonNegative(n) && isPowerOfTwo(c) => (Rsh64Ux64 n (Const64 <typ.UInt64> [log64(c)]))
(Div64 n (Const64 [-1<<63])) && isNonNegative(n) => (Const64 [0])
// Unsigned divide, not a power of 2. Strength reduce to a multiply.
// For 8-bit divides, we just do a direct 9-bit by 8-bit multiply.
(Div8u x (Const8 [c])) && umagicOK8(c) =>
(Trunc32to8
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh32Ux64 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(1<<8+umagic8(c).m)])
(ZeroExt8to32 x))
(Const64 <typ.UInt64> [8+umagic8(c).s])))
// For 16-bit divides on 64-bit machines, we do a direct 17-bit by 16-bit multiply.
(Div16u x (Const16 [c])) && umagicOK16(c) && config.RegSize == 8 =>
(Trunc64to16
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh64Ux64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(1<<16+umagic16(c).m)])
(ZeroExt16to64 x))
(Const64 <typ.UInt64> [16+umagic16(c).s])))
// For 16-bit divides on 32-bit machines
(Div16u x (Const16 [c])) && umagicOK16(c) && config.RegSize == 4 && umagic16(c).m&1 == 0 =>
(Trunc32to16
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh32Ux64 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(1<<15+umagic16(c).m/2)])
(ZeroExt16to32 x))
(Const64 <typ.UInt64> [16+umagic16(c).s-1])))
(Div16u x (Const16 [c])) && umagicOK16(c) && config.RegSize == 4 && c&1 == 0 =>
(Trunc32to16
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh32Ux64 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(1<<15+(umagic16(c).m+1)/2)])
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh32Ux64 <typ.UInt32> (ZeroExt16to32 x) (Const64 <typ.UInt64> [1])))
(Const64 <typ.UInt64> [16+umagic16(c).s-2])))
(Div16u x (Const16 [c])) && umagicOK16(c) && config.RegSize == 4 && config.useAvg =>
(Trunc32to16
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh32Ux64 <typ.UInt32>
(Avg32u
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Lsh32x64 <typ.UInt32> (ZeroExt16to32 x) (Const64 <typ.UInt64> [16]))
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(umagic16(c).m)])
(ZeroExt16to32 x)))
(Const64 <typ.UInt64> [16+umagic16(c).s-1])))
// For 32-bit divides on 32-bit machines
(Div32u x (Const32 [c])) && umagicOK32(c) && config.RegSize == 4 && umagic32(c).m&1 == 0 && config.useHmul =>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh32Ux64 <typ.UInt32>
(Hmul32u <typ.UInt32>
(Const32 <typ.UInt32> [int32(1<<31+umagic32(c).m/2)])
x)
(Const64 <typ.UInt64> [umagic32(c).s-1]))
(Div32u x (Const32 [c])) && umagicOK32(c) && config.RegSize == 4 && c&1 == 0 && config.useHmul =>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh32Ux64 <typ.UInt32>
(Hmul32u <typ.UInt32>
(Const32 <typ.UInt32> [int32(1<<31+(umagic32(c).m+1)/2)])
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh32Ux64 <typ.UInt32> x (Const64 <typ.UInt64> [1])))
(Const64 <typ.UInt64> [umagic32(c).s-2]))
(Div32u x (Const32 [c])) && umagicOK32(c) && config.RegSize == 4 && config.useAvg && config.useHmul =>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh32Ux64 <typ.UInt32>
(Avg32u
x
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Hmul32u <typ.UInt32>
(Const32 <typ.UInt32> [int32(umagic32(c).m)])
x))
(Const64 <typ.UInt64> [umagic32(c).s-1]))
// For 32-bit divides on 64-bit machines
// We'll use a regular (non-hi) multiply for this case.
(Div32u x (Const32 [c])) && umagicOK32(c) && config.RegSize == 8 && umagic32(c).m&1 == 0 =>
(Trunc64to32
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh64Ux64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(1<<31+umagic32(c).m/2)])
(ZeroExt32to64 x))
(Const64 <typ.UInt64> [32+umagic32(c).s-1])))
(Div32u x (Const32 [c])) && umagicOK32(c) && config.RegSize == 8 && c&1 == 0 =>
(Trunc64to32
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh64Ux64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(1<<31+(umagic32(c).m+1)/2)])
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh64Ux64 <typ.UInt64> (ZeroExt32to64 x) (Const64 <typ.UInt64> [1])))
(Const64 <typ.UInt64> [32+umagic32(c).s-2])))
(Div32u x (Const32 [c])) && umagicOK32(c) && config.RegSize == 8 && config.useAvg =>
(Trunc64to32
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh64Ux64 <typ.UInt64>
(Avg64u
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Lsh64x64 <typ.UInt64> (ZeroExt32to64 x) (Const64 <typ.UInt64> [32]))
(Mul64 <typ.UInt64>
(Const64 <typ.UInt32> [int64(umagic32(c).m)])
(ZeroExt32to64 x)))
(Const64 <typ.UInt64> [32+umagic32(c).s-1])))
// For unsigned 64-bit divides on 32-bit machines,
// if the constant fits in 16 bits (so that the last term
// fits in 32 bits), convert to three 32-bit divides by a constant.
//
// If 1<<32 = Q * c + R
// and x = hi << 32 + lo
//
// Then x = (hi/c*c + hi%c) << 32 + lo
// = hi/c*c<<32 + hi%c<<32 + lo
// = hi/c*c<<32 + (hi%c)*(Q*c+R) + lo/c*c + lo%c
// = hi/c*c<<32 + (hi%c)*Q*c + lo/c*c + (hi%c*R+lo%c)
// and x / c = (hi/c)<<32 + (hi%c)*Q + lo/c + (hi%c*R+lo%c)/c
(Div64u x (Const64 [c])) && c > 0 && c <= 0xFFFF && umagicOK32(int32(c)) && config.RegSize == 4 && config.useHmul =>
(Add64
(Add64 <typ.UInt64>
(Add64 <typ.UInt64>
(Lsh64x64 <typ.UInt64>
(ZeroExt32to64
(Div32u <typ.UInt32>
(Trunc64to32 <typ.UInt32> (Rsh64Ux64 <typ.UInt64> x (Const64 <typ.UInt64> [32])))
(Const32 <typ.UInt32> [int32(c)])))
(Const64 <typ.UInt64> [32]))
(ZeroExt32to64 (Div32u <typ.UInt32> (Trunc64to32 <typ.UInt32> x) (Const32 <typ.UInt32> [int32(c)]))))
(Mul64 <typ.UInt64>
(ZeroExt32to64 <typ.UInt64>
(Mod32u <typ.UInt32>
(Trunc64to32 <typ.UInt32> (Rsh64Ux64 <typ.UInt64> x (Const64 <typ.UInt64> [32])))
(Const32 <typ.UInt32> [int32(c)])))
(Const64 <typ.UInt64> [int64((1<<32)/c)])))
(ZeroExt32to64
(Div32u <typ.UInt32>
(Add32 <typ.UInt32>
(Mod32u <typ.UInt32> (Trunc64to32 <typ.UInt32> x) (Const32 <typ.UInt32> [int32(c)]))
(Mul32 <typ.UInt32>
(Mod32u <typ.UInt32>
(Trunc64to32 <typ.UInt32> (Rsh64Ux64 <typ.UInt64> x (Const64 <typ.UInt64> [32])))
(Const32 <typ.UInt32> [int32(c)]))
(Const32 <typ.UInt32> [int32((1<<32)%c)])))
(Const32 <typ.UInt32> [int32(c)]))))
// For 64-bit divides on 64-bit machines
// (64-bit divides on 32-bit machines are lowered to a runtime call by the walk pass.)
(Div64u x (Const64 [c])) && umagicOK64(c) && config.RegSize == 8 && umagic64(c).m&1 == 0 && config.useHmul =>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh64Ux64 <typ.UInt64>
(Hmul64u <typ.UInt64>
(Const64 <typ.UInt64> [int64(1<<63+umagic64(c).m/2)])
x)
(Const64 <typ.UInt64> [umagic64(c).s-1]))
(Div64u x (Const64 [c])) && umagicOK64(c) && config.RegSize == 8 && c&1 == 0 && config.useHmul =>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh64Ux64 <typ.UInt64>
(Hmul64u <typ.UInt64>
(Const64 <typ.UInt64> [int64(1<<63+(umagic64(c).m+1)/2)])
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh64Ux64 <typ.UInt64> x (Const64 <typ.UInt64> [1])))
(Const64 <typ.UInt64> [umagic64(c).s-2]))
(Div64u x (Const64 [c])) && umagicOK64(c) && config.RegSize == 8 && config.useAvg && config.useHmul =>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Rsh64Ux64 <typ.UInt64>
(Avg64u
x
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Hmul64u <typ.UInt64>
(Const64 <typ.UInt64> [int64(umagic64(c).m)])
x))
(Const64 <typ.UInt64> [umagic64(c).s-1]))
// Signed divide by a negative constant. Rewrite to divide by a positive constant.
(Div8 <t> n (Const8 [c])) && c < 0 && c != -1<<7 => (Neg8 (Div8 <t> n (Const8 <t> [-c])))
(Div16 <t> n (Const16 [c])) && c < 0 && c != -1<<15 => (Neg16 (Div16 <t> n (Const16 <t> [-c])))
(Div32 <t> n (Const32 [c])) && c < 0 && c != -1<<31 => (Neg32 (Div32 <t> n (Const32 <t> [-c])))
(Div64 <t> n (Const64 [c])) && c < 0 && c != -1<<63 => (Neg64 (Div64 <t> n (Const64 <t> [-c])))
// Dividing by the most-negative number. Result is always 0 except
// if the input is also the most-negative number.
// We can detect that using the sign bit of x & -x.
(Div8 <t> x (Const8 [-1<<7 ])) => (Rsh8Ux64 (And8 <t> x (Neg8 <t> x)) (Const64 <typ.UInt64> [7 ]))
(Div16 <t> x (Const16 [-1<<15])) => (Rsh16Ux64 (And16 <t> x (Neg16 <t> x)) (Const64 <typ.UInt64> [15]))
(Div32 <t> x (Const32 [-1<<31])) => (Rsh32Ux64 (And32 <t> x (Neg32 <t> x)) (Const64 <typ.UInt64> [31]))
(Div64 <t> x (Const64 [-1<<63])) => (Rsh64Ux64 (And64 <t> x (Neg64 <t> x)) (Const64 <typ.UInt64> [63]))
// Signed divide by power of 2.
// n / c = n >> log(c) if n >= 0
// = (n+c-1) >> log(c) if n < 0
// We conditionally add c-1 by adding n>>63>>(64-log(c)) (first shift signed, second shift unsigned).
(Div8 <t> n (Const8 [c])) && isPowerOfTwo(c) =>
(Rsh8x64
(Add8 <t> n (Rsh8Ux64 <t> (Rsh8x64 <t> n (Const64 <typ.UInt64> [ 7])) (Const64 <typ.UInt64> [int64( 8-log8(c))])))
(Const64 <typ.UInt64> [int64(log8(c))]))
(Div16 <t> n (Const16 [c])) && isPowerOfTwo(c) =>
(Rsh16x64
(Add16 <t> n (Rsh16Ux64 <t> (Rsh16x64 <t> n (Const64 <typ.UInt64> [15])) (Const64 <typ.UInt64> [int64(16-log16(c))])))
(Const64 <typ.UInt64> [int64(log16(c))]))
(Div32 <t> n (Const32 [c])) && isPowerOfTwo(c) =>
(Rsh32x64
(Add32 <t> n (Rsh32Ux64 <t> (Rsh32x64 <t> n (Const64 <typ.UInt64> [31])) (Const64 <typ.UInt64> [int64(32-log32(c))])))
(Const64 <typ.UInt64> [int64(log32(c))]))
(Div64 <t> n (Const64 [c])) && isPowerOfTwo(c) =>
(Rsh64x64
(Add64 <t> n (Rsh64Ux64 <t> (Rsh64x64 <t> n (Const64 <typ.UInt64> [63])) (Const64 <typ.UInt64> [int64(64-log64(c))])))
(Const64 <typ.UInt64> [int64(log64(c))]))
// Signed divide, not a power of 2. Strength reduce to a multiply.
(Div8 <t> x (Const8 [c])) && smagicOK8(c) =>
(Sub8 <t>
(Rsh32x64 <t>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(smagic8(c).m)])
(SignExt8to32 x))
(Const64 <typ.UInt64> [8+smagic8(c).s]))
(Rsh32x64 <t>
(SignExt8to32 x)
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Const64 <typ.UInt64> [31])))
(Div16 <t> x (Const16 [c])) && smagicOK16(c) =>
(Sub16 <t>
(Rsh32x64 <t>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(smagic16(c).m)])
(SignExt16to32 x))
(Const64 <typ.UInt64> [16+smagic16(c).s]))
(Rsh32x64 <t>
(SignExt16to32 x)
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Const64 <typ.UInt64> [31])))
(Div32 <t> x (Const32 [c])) && smagicOK32(c) && config.RegSize == 8 =>
(Sub32 <t>
(Rsh64x64 <t>
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(smagic32(c).m)])
(SignExt32to64 x))
(Const64 <typ.UInt64> [32+smagic32(c).s]))
(Rsh64x64 <t>
(SignExt32to64 x)
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Const64 <typ.UInt64> [63])))
(Div32 <t> x (Const32 [c])) && smagicOK32(c) && config.RegSize == 4 && smagic32(c).m&1 == 0 && config.useHmul =>
(Sub32 <t>
(Rsh32x64 <t>
(Hmul32 <t>
(Const32 <typ.UInt32> [int32(smagic32(c).m/2)])
x)
(Const64 <typ.UInt64> [smagic32(c).s-1]))
(Rsh32x64 <t>
x
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Const64 <typ.UInt64> [31])))
(Div32 <t> x (Const32 [c])) && smagicOK32(c) && config.RegSize == 4 && smagic32(c).m&1 != 0 && config.useHmul =>
(Sub32 <t>
(Rsh32x64 <t>
(Add32 <t>
(Hmul32 <t>
(Const32 <typ.UInt32> [int32(smagic32(c).m)])
x)
x)
(Const64 <typ.UInt64> [smagic32(c).s]))
(Rsh32x64 <t>
x
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Const64 <typ.UInt64> [31])))
(Div64 <t> x (Const64 [c])) && smagicOK64(c) && smagic64(c).m&1 == 0 && config.useHmul =>
(Sub64 <t>
(Rsh64x64 <t>
(Hmul64 <t>
(Const64 <typ.UInt64> [int64(smagic64(c).m/2)])
x)
(Const64 <typ.UInt64> [smagic64(c).s-1]))
(Rsh64x64 <t>
x
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Const64 <typ.UInt64> [63])))
(Div64 <t> x (Const64 [c])) && smagicOK64(c) && smagic64(c).m&1 != 0 && config.useHmul =>
(Sub64 <t>
(Rsh64x64 <t>
(Add64 <t>
(Hmul64 <t>
(Const64 <typ.UInt64> [int64(smagic64(c).m)])
x)
x)
(Const64 <typ.UInt64> [smagic64(c).s]))
(Rsh64x64 <t>
x
cmd/compile: change ssa.Type into *types.Type When package ssa was created, Type was in package gc. To avoid circular dependencies, we used an interface (ssa.Type) to represent type information in SSA. In the Go 1.9 cycle, gri extricated the Type type from package gc. As a result, we can now use it in package ssa. Now, instead of package types depending on package ssa, it is the other way. This is a more sensible dependency tree, and helps compiler performance a bit. Though this is a big CL, most of the changes are mechanical and uninteresting. Interesting bits: * Add new singleton globals to package types for the special SSA types Memory, Void, Invalid, Flags, and Int128. * Add two new Types, TSSA for the special types, and TTUPLE, for SSA tuple types. ssa.MakeTuple is now types.NewTuple. * Move type comparison result constants CMPlt, CMPeq, and CMPgt to package types. * We had picked the name "types" in our rules for the handy list of types provided by ssa.Config. That conflicted with the types package name, so change it to "typ". * Update the type comparison routine to handle tuples and special types inline. * Teach gc/fmt.go how to print special types. * We can now eliminate ElemTypes in favor of just Elem, and probably also some other duplicated Type methods designed to return ssa.Type instead of *types.Type. * The ssa tests were using their own dummy types, and they were not particularly careful about types in general. Of necessity, this CL switches them to use *types.Type; it does not make them more type-accurate. Unfortunately, using types.Type means initializing a bit of the types universe. This is prime for refactoring and improvement. This shrinks ssa.Value; it now fits in a smaller size class on 64 bit systems. This doesn't have a giant impact, though, since most Values are preallocated in a chunk. name old alloc/op new alloc/op delta Template 37.9MB ± 0% 37.7MB ± 0% -0.57% (p=0.000 n=10+8) Unicode 28.9MB ± 0% 28.7MB ± 0% -0.52% (p=0.000 n=10+10) GoTypes 110MB ± 0% 109MB ± 0% -0.88% (p=0.000 n=10+10) Flate 24.7MB ± 0% 24.6MB ± 0% -0.66% (p=0.000 n=10+10) GoParser 31.1MB ± 0% 30.9MB ± 0% -0.61% (p=0.000 n=10+9) Reflect 73.9MB ± 0% 73.4MB ± 0% -0.62% (p=0.000 n=10+8) Tar 25.8MB ± 0% 25.6MB ± 0% -0.77% (p=0.000 n=9+10) XML 41.2MB ± 0% 40.9MB ± 0% -0.80% (p=0.000 n=10+10) [Geo mean] 40.5MB 40.3MB -0.68% name old allocs/op new allocs/op delta Template 385k ± 0% 386k ± 0% ~ (p=0.356 n=10+9) Unicode 343k ± 1% 344k ± 0% ~ (p=0.481 n=10+10) GoTypes 1.16M ± 0% 1.16M ± 0% -0.16% (p=0.004 n=10+10) Flate 238k ± 1% 238k ± 1% ~ (p=0.853 n=10+10) GoParser 320k ± 0% 320k ± 0% ~ (p=0.720 n=10+9) Reflect 957k ± 0% 957k ± 0% ~ (p=0.460 n=10+8) Tar 252k ± 0% 252k ± 0% ~ (p=0.133 n=9+10) XML 400k ± 0% 400k ± 0% ~ (p=0.796 n=10+10) [Geo mean] 428k 428k -0.01% Removing all the interface calls helps non-trivially with CPU, though. name old time/op new time/op delta Template 178ms ± 4% 173ms ± 3% -2.90% (p=0.000 n=94+96) Unicode 85.0ms ± 4% 83.9ms ± 4% -1.23% (p=0.000 n=96+96) GoTypes 543ms ± 3% 528ms ± 3% -2.73% (p=0.000 n=98+96) Flate 116ms ± 3% 113ms ± 4% -2.34% (p=0.000 n=96+99) GoParser 144ms ± 3% 140ms ± 4% -2.80% (p=0.000 n=99+97) Reflect 344ms ± 3% 334ms ± 4% -3.02% (p=0.000 n=100+99) Tar 106ms ± 5% 103ms ± 4% -3.30% (p=0.000 n=98+94) XML 198ms ± 5% 192ms ± 4% -2.88% (p=0.000 n=92+95) [Geo mean] 178ms 173ms -2.65% name old user-time/op new user-time/op delta Template 229ms ± 5% 224ms ± 5% -2.36% (p=0.000 n=95+99) Unicode 107ms ± 6% 106ms ± 5% -1.13% (p=0.001 n=93+95) GoTypes 696ms ± 4% 679ms ± 4% -2.45% (p=0.000 n=97+99) Flate 137ms ± 4% 134ms ± 5% -2.66% (p=0.000 n=99+96) GoParser 176ms ± 5% 172ms ± 8% -2.27% (p=0.000 n=98+100) Reflect 430ms ± 6% 411ms ± 5% -4.46% (p=0.000 n=100+92) Tar 128ms ±13% 123ms ±13% -4.21% (p=0.000 n=100+100) XML 239ms ± 6% 233ms ± 6% -2.50% (p=0.000 n=95+97) [Geo mean] 220ms 213ms -2.76% Change-Id: I15c7d6268347f8358e75066dfdbd77db24e8d0c1 Reviewed-on: https://go-review.googlesource.com/42145 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-28 14:12:28 -07:00
(Const64 <typ.UInt64> [63])))
// Unsigned mod by power of 2 constant.
(Mod8u <t> n (Const8 [c])) && isPowerOfTwo(c) => (And8 n (Const8 <t> [c-1]))
(Mod16u <t> n (Const16 [c])) && isPowerOfTwo(c) => (And16 n (Const16 <t> [c-1]))
(Mod32u <t> n (Const32 [c])) && isPowerOfTwo(c) => (And32 n (Const32 <t> [c-1]))
(Mod64u <t> n (Const64 [c])) && isPowerOfTwo(c) => (And64 n (Const64 <t> [c-1]))
(Mod64u <t> n (Const64 [-1<<63])) => (And64 n (Const64 <t> [1<<63-1]))
// Signed non-negative mod by power of 2 constant.
(Mod8 <t> n (Const8 [c])) && isNonNegative(n) && isPowerOfTwo(c) => (And8 n (Const8 <t> [c-1]))
(Mod16 <t> n (Const16 [c])) && isNonNegative(n) && isPowerOfTwo(c) => (And16 n (Const16 <t> [c-1]))
(Mod32 <t> n (Const32 [c])) && isNonNegative(n) && isPowerOfTwo(c) => (And32 n (Const32 <t> [c-1]))
(Mod64 <t> n (Const64 [c])) && isNonNegative(n) && isPowerOfTwo(c) => (And64 n (Const64 <t> [c-1]))
(Mod64 n (Const64 [-1<<63])) && isNonNegative(n) => n
// Signed mod by negative constant.
(Mod8 <t> n (Const8 [c])) && c < 0 && c != -1<<7 => (Mod8 <t> n (Const8 <t> [-c]))
(Mod16 <t> n (Const16 [c])) && c < 0 && c != -1<<15 => (Mod16 <t> n (Const16 <t> [-c]))
(Mod32 <t> n (Const32 [c])) && c < 0 && c != -1<<31 => (Mod32 <t> n (Const32 <t> [-c]))
(Mod64 <t> n (Const64 [c])) && c < 0 && c != -1<<63 => (Mod64 <t> n (Const64 <t> [-c]))
// All other mods by constants, do A%B = A-(A/B*B).
// This implements % with two * and a bunch of ancillary ops.
// One of the * is free if the user's code also computes A/B.
(Mod8 <t> x (Const8 [c])) && x.Op != OpConst8 && (c > 0 || c == -1<<7)
=> (Sub8 x (Mul8 <t> (Div8 <t> x (Const8 <t> [c])) (Const8 <t> [c])))
(Mod16 <t> x (Const16 [c])) && x.Op != OpConst16 && (c > 0 || c == -1<<15)
=> (Sub16 x (Mul16 <t> (Div16 <t> x (Const16 <t> [c])) (Const16 <t> [c])))
(Mod32 <t> x (Const32 [c])) && x.Op != OpConst32 && (c > 0 || c == -1<<31)
=> (Sub32 x (Mul32 <t> (Div32 <t> x (Const32 <t> [c])) (Const32 <t> [c])))
(Mod64 <t> x (Const64 [c])) && x.Op != OpConst64 && (c > 0 || c == -1<<63)
=> (Sub64 x (Mul64 <t> (Div64 <t> x (Const64 <t> [c])) (Const64 <t> [c])))
(Mod8u <t> x (Const8 [c])) && x.Op != OpConst8 && c > 0 && umagicOK8( c)
=> (Sub8 x (Mul8 <t> (Div8u <t> x (Const8 <t> [c])) (Const8 <t> [c])))
(Mod16u <t> x (Const16 [c])) && x.Op != OpConst16 && c > 0 && umagicOK16(c)
=> (Sub16 x (Mul16 <t> (Div16u <t> x (Const16 <t> [c])) (Const16 <t> [c])))
(Mod32u <t> x (Const32 [c])) && x.Op != OpConst32 && c > 0 && umagicOK32(c)
=> (Sub32 x (Mul32 <t> (Div32u <t> x (Const32 <t> [c])) (Const32 <t> [c])))
(Mod64u <t> x (Const64 [c])) && x.Op != OpConst64 && c > 0 && umagicOK64(c)
=> (Sub64 x (Mul64 <t> (Div64u <t> x (Const64 <t> [c])) (Const64 <t> [c])))
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
// For architectures without rotates on less than 32-bits, promote these checks to 32-bit.
(Eq8 (Mod8u x (Const8 [c])) (Const8 [0])) && x.Op != OpConst8 && udivisibleOK8(c) && !hasSmallRotate(config) =>
(Eq32 (Mod32u <typ.UInt32> (ZeroExt8to32 <typ.UInt32> x) (Const32 <typ.UInt32> [int32(uint8(c))])) (Const32 <typ.UInt32> [0]))
(Eq16 (Mod16u x (Const16 [c])) (Const16 [0])) && x.Op != OpConst16 && udivisibleOK16(c) && !hasSmallRotate(config) =>
(Eq32 (Mod32u <typ.UInt32> (ZeroExt16to32 <typ.UInt32> x) (Const32 <typ.UInt32> [int32(uint16(c))])) (Const32 <typ.UInt32> [0]))
(Eq8 (Mod8 x (Const8 [c])) (Const8 [0])) && x.Op != OpConst8 && sdivisibleOK8(c) && !hasSmallRotate(config) =>
(Eq32 (Mod32 <typ.Int32> (SignExt8to32 <typ.Int32> x) (Const32 <typ.Int32> [int32(c)])) (Const32 <typ.Int32> [0]))
(Eq16 (Mod16 x (Const16 [c])) (Const16 [0])) && x.Op != OpConst16 && sdivisibleOK16(c) && !hasSmallRotate(config) =>
(Eq32 (Mod32 <typ.Int32> (SignExt16to32 <typ.Int32> x) (Const32 <typ.Int32> [int32(c)])) (Const32 <typ.Int32> [0]))
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
// Divisibility checks x%c == 0 convert to multiply and rotate.
// Note, x%c == 0 is rewritten as x == c*(x/c) during the opt pass
// where (x/c) is performed using multiplication with magic constants.
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
// To rewrite x%c == 0 requires pattern matching the rewritten expression
// and checking that the division by the same constant wasn't already calculated.
// This check is made by counting uses of the magic constant multiplication.
// Note that if there were an intermediate opt pass, this rule could be applied
// directly on the Div op and magic division rewrites could be delayed to late opt.
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
// Unsigned divisibility checks convert to multiply and rotate.
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(Eq8 x (Mul8 (Const8 [c])
(Trunc32to8
(Rsh32Ux64
mul:(Mul32
(Const32 [m])
(ZeroExt8to32 x))
(Const64 [s])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(1<<8+umagic8(c).m) && s == 8+umagic8(c).s
&& x.Op != OpConst8 && udivisibleOK8(c)
=> (Leq8U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft8 <typ.UInt8>
(Mul8 <typ.UInt8>
(Const8 <typ.UInt8> [int8(udivisible8(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const8 <typ.UInt8> [int8(8-udivisible8(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const8 <typ.UInt8> [int8(udivisible8(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq16 x (Mul16 (Const16 [c])
(Trunc64to16
(Rsh64Ux64
mul:(Mul64
(Const64 [m])
(ZeroExt16to64 x))
(Const64 [s])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(1<<16+umagic16(c).m) && s == 16+umagic16(c).s
&& x.Op != OpConst16 && udivisibleOK16(c)
=> (Leq16U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int16(udivisible16(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const16 <typ.UInt16> [int16(16-udivisible16(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const16 <typ.UInt16> [int16(udivisible16(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq16 x (Mul16 (Const16 [c])
(Trunc32to16
(Rsh32Ux64
mul:(Mul32
(Const32 [m])
(ZeroExt16to32 x))
(Const64 [s])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(1<<15+umagic16(c).m/2) && s == 16+umagic16(c).s-1
&& x.Op != OpConst16 && udivisibleOK16(c)
=> (Leq16U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int16(udivisible16(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const16 <typ.UInt16> [int16(16-udivisible16(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const16 <typ.UInt16> [int16(udivisible16(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq16 x (Mul16 (Const16 [c])
(Trunc32to16
(Rsh32Ux64
mul:(Mul32
(Const32 [m])
(Rsh32Ux64 (ZeroExt16to32 x) (Const64 [1])))
(Const64 [s])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(1<<15+(umagic16(c).m+1)/2) && s == 16+umagic16(c).s-2
&& x.Op != OpConst16 && udivisibleOK16(c)
=> (Leq16U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int16(udivisible16(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const16 <typ.UInt16> [int16(16-udivisible16(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const16 <typ.UInt16> [int16(udivisible16(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq16 x (Mul16 (Const16 [c])
(Trunc32to16
(Rsh32Ux64
(Avg32u
(Lsh32x64 (ZeroExt16to32 x) (Const64 [16]))
mul:(Mul32
(Const32 [m])
(ZeroExt16to32 x)))
(Const64 [s])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(umagic16(c).m) && s == 16+umagic16(c).s-1
&& x.Op != OpConst16 && udivisibleOK16(c)
=> (Leq16U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int16(udivisible16(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const16 <typ.UInt16> [int16(16-udivisible16(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const16 <typ.UInt16> [int16(udivisible16(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq32 x (Mul32 (Const32 [c])
(Rsh32Ux64
mul:(Hmul32u
(Const32 [m])
x)
(Const64 [s]))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(1<<31+umagic32(c).m/2) && s == umagic32(c).s-1
&& x.Op != OpConst32 && udivisibleOK32(c)
=> (Leq32U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(udivisible32(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const32 <typ.UInt32> [int32(32-udivisible32(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const32 <typ.UInt32> [int32(udivisible32(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq32 x (Mul32 (Const32 [c])
(Rsh32Ux64
mul:(Hmul32u
(Const32 <typ.UInt32> [m])
(Rsh32Ux64 x (Const64 [1])))
(Const64 [s]))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(1<<31+(umagic32(c).m+1)/2) && s == umagic32(c).s-2
&& x.Op != OpConst32 && udivisibleOK32(c)
=> (Leq32U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(udivisible32(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const32 <typ.UInt32> [int32(32-udivisible32(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const32 <typ.UInt32> [int32(udivisible32(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq32 x (Mul32 (Const32 [c])
(Rsh32Ux64
(Avg32u
x
mul:(Hmul32u
(Const32 [m])
x))
(Const64 [s]))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(umagic32(c).m) && s == umagic32(c).s-1
&& x.Op != OpConst32 && udivisibleOK32(c)
=> (Leq32U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(udivisible32(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const32 <typ.UInt32> [int32(32-udivisible32(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const32 <typ.UInt32> [int32(udivisible32(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq32 x (Mul32 (Const32 [c])
(Trunc64to32
(Rsh64Ux64
mul:(Mul64
(Const64 [m])
(ZeroExt32to64 x))
(Const64 [s])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(1<<31+umagic32(c).m/2) && s == 32+umagic32(c).s-1
&& x.Op != OpConst32 && udivisibleOK32(c)
=> (Leq32U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(udivisible32(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const32 <typ.UInt32> [int32(32-udivisible32(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const32 <typ.UInt32> [int32(udivisible32(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq32 x (Mul32 (Const32 [c])
(Trunc64to32
(Rsh64Ux64
mul:(Mul64
(Const64 [m])
(Rsh64Ux64 (ZeroExt32to64 x) (Const64 [1])))
(Const64 [s])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(1<<31+(umagic32(c).m+1)/2) && s == 32+umagic32(c).s-2
&& x.Op != OpConst32 && udivisibleOK32(c)
=> (Leq32U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(udivisible32(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const32 <typ.UInt32> [int32(32-udivisible32(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const32 <typ.UInt32> [int32(udivisible32(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq32 x (Mul32 (Const32 [c])
(Trunc64to32
(Rsh64Ux64
(Avg64u
(Lsh64x64 (ZeroExt32to64 x) (Const64 [32]))
mul:(Mul64
(Const64 [m])
(ZeroExt32to64 x)))
(Const64 [s])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(umagic32(c).m) && s == 32+umagic32(c).s-1
&& x.Op != OpConst32 && udivisibleOK32(c)
=> (Leq32U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(udivisible32(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const32 <typ.UInt32> [int32(32-udivisible32(c).k)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const32 <typ.UInt32> [int32(udivisible32(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq64 x (Mul64 (Const64 [c])
(Rsh64Ux64
mul:(Hmul64u
(Const64 [m])
x)
(Const64 [s]))
)
) && v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(1<<63+umagic64(c).m/2) && s == umagic64(c).s-1
&& x.Op != OpConst64 && udivisibleOK64(c)
=> (Leq64U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(udivisible64(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const64 <typ.UInt64> [64-udivisible64(c).k])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const64 <typ.UInt64> [int64(udivisible64(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq64 x (Mul64 (Const64 [c])
(Rsh64Ux64
mul:(Hmul64u
(Const64 [m])
(Rsh64Ux64 x (Const64 [1])))
(Const64 [s]))
)
) && v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(1<<63+(umagic64(c).m+1)/2) && s == umagic64(c).s-2
&& x.Op != OpConst64 && udivisibleOK64(c)
=> (Leq64U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(udivisible64(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const64 <typ.UInt64> [64-udivisible64(c).k])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const64 <typ.UInt64> [int64(udivisible64(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Eq64 x (Mul64 (Const64 [c])
(Rsh64Ux64
(Avg64u
x
mul:(Hmul64u
(Const64 [m])
x))
(Const64 [s]))
)
) && v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(umagic64(c).m) && s == umagic64(c).s-1
&& x.Op != OpConst64 && udivisibleOK64(c)
=> (Leq64U
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
(RotateLeft64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(udivisible64(c).m)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
x)
(Const64 <typ.UInt64> [64-udivisible64(c).k])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
(Const64 <typ.UInt64> [int64(udivisible64(c).max)])
cmd/compile: add unsigned divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also available. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. Additional rules to lower the generic RotateLeft* ops were also applied. On amd64, the divisibility check is 25-50% faster. name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.08ns ± 1% ~ (p=0.881 n=5+5) DivisibleconstI64-4 2.67ns ± 0% 2.67ns ± 1% ~ (p=1.000 n=5+5) DivisibleWDivconstI64-4 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.683 n=5+5) DivconstU64-4 2.08ns ± 1% 2.08ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstU64-4 2.77ns ± 1% 1.55ns ± 2% -43.90% (p=0.008 n=5+5) DivisibleWDivconstU64-4 2.99ns ± 1% 2.99ns ± 1% ~ (p=1.000 n=5+5) DivconstI32-4 1.53ns ± 2% 1.53ns ± 0% ~ (p=1.000 n=5+5) DivisibleconstI32-4 2.23ns ± 0% 2.25ns ± 3% ~ (p=0.167 n=5+5) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.429 n=5+5) DivconstU32-4 1.78ns ± 0% 1.78ns ± 1% ~ (p=1.000 n=4+5) DivisibleconstU32-4 2.52ns ± 2% 1.26ns ± 0% -49.96% (p=0.000 n=5+4) DivisibleWDivconstU32-4 2.63ns ± 0% 2.85ns ±10% +8.29% (p=0.016 n=4+5) DivconstI16-4 1.54ns ± 0% 1.54ns ± 0% ~ (p=0.333 n=4+5) DivisibleconstI16-4 2.10ns ± 0% 2.10ns ± 1% ~ (p=0.571 n=4+5) DivisibleWDivconstI16-4 2.22ns ± 0% 2.23ns ± 1% ~ (p=0.556 n=4+5) DivconstU16-4 1.09ns ± 0% 1.01ns ± 1% -7.74% (p=0.000 n=4+5) DivisibleconstU16-4 1.83ns ± 0% 1.26ns ± 0% -31.52% (p=0.008 n=5+5) DivisibleWDivconstU16-4 1.88ns ± 0% 1.89ns ± 1% ~ (p=0.365 n=5+5) DivconstI8-4 1.54ns ± 1% 1.54ns ± 1% ~ (p=1.000 n=5+5) DivisibleconstI8-4 2.10ns ± 0% 2.11ns ± 0% ~ (p=0.238 n=5+4) DivisibleWDivconstI8-4 2.22ns ± 0% 2.23ns ± 2% ~ (p=0.762 n=5+5) DivconstU8-4 0.92ns ± 1% 0.94ns ± 1% +2.65% (p=0.008 n=5+5) DivisibleconstU8-4 1.66ns ± 0% 1.26ns ± 1% -24.28% (p=0.008 n=5+5) DivisibleWDivconstU8-4 1.79ns ± 0% 1.80ns ± 1% ~ (p=0.079 n=4+5) A follow-up change will address the signed division case. Updates #30282 Change-Id: I7e995f167179aa5c76bb10fbcbeb49c520943403 Reviewed-on: https://go-review.googlesource.com/c/go/+/168037 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-03-09 21:58:16 -07:00
)
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
// Signed divisibility checks convert to multiply, add and rotate.
(Eq8 x (Mul8 (Const8 [c])
(Sub8
(Rsh32x64
mul:(Mul32
(Const32 [m])
(SignExt8to32 x))
(Const64 [s]))
(Rsh32x64
(SignExt8to32 x)
(Const64 [31])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(smagic8(c).m) && s == 8+smagic8(c).s
&& x.Op != OpConst8 && sdivisibleOK8(c)
=> (Leq8U
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
(RotateLeft8 <typ.UInt8>
(Add8 <typ.UInt8>
(Mul8 <typ.UInt8>
(Const8 <typ.UInt8> [int8(sdivisible8(c).m)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
x)
(Const8 <typ.UInt8> [int8(sdivisible8(c).a)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const8 <typ.UInt8> [int8(8-sdivisible8(c).k)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const8 <typ.UInt8> [int8(sdivisible8(c).max)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Eq16 x (Mul16 (Const16 [c])
(Sub16
(Rsh32x64
mul:(Mul32
(Const32 [m])
(SignExt16to32 x))
(Const64 [s]))
(Rsh32x64
(SignExt16to32 x)
(Const64 [31])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(smagic16(c).m) && s == 16+smagic16(c).s
&& x.Op != OpConst16 && sdivisibleOK16(c)
=> (Leq16U
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
(RotateLeft16 <typ.UInt16>
(Add16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int16(sdivisible16(c).m)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
x)
(Const16 <typ.UInt16> [int16(sdivisible16(c).a)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const16 <typ.UInt16> [int16(16-sdivisible16(c).k)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const16 <typ.UInt16> [int16(sdivisible16(c).max)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Eq32 x (Mul32 (Const32 [c])
(Sub32
(Rsh64x64
mul:(Mul64
(Const64 [m])
(SignExt32to64 x))
(Const64 [s]))
(Rsh64x64
(SignExt32to64 x)
(Const64 [63])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(smagic32(c).m) && s == 32+smagic32(c).s
&& x.Op != OpConst32 && sdivisibleOK32(c)
=> (Leq32U
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
(RotateLeft32 <typ.UInt32>
(Add32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(sdivisible32(c).m)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
x)
(Const32 <typ.UInt32> [int32(sdivisible32(c).a)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const32 <typ.UInt32> [int32(32-sdivisible32(c).k)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const32 <typ.UInt32> [int32(sdivisible32(c).max)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Eq32 x (Mul32 (Const32 [c])
(Sub32
(Rsh32x64
mul:(Hmul32
(Const32 [m])
x)
(Const64 [s]))
(Rsh32x64
x
(Const64 [31])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(smagic32(c).m/2) && s == smagic32(c).s-1
&& x.Op != OpConst32 && sdivisibleOK32(c)
=> (Leq32U
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
(RotateLeft32 <typ.UInt32>
(Add32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(sdivisible32(c).m)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
x)
(Const32 <typ.UInt32> [int32(sdivisible32(c).a)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const32 <typ.UInt32> [int32(32-sdivisible32(c).k)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const32 <typ.UInt32> [int32(sdivisible32(c).max)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Eq32 x (Mul32 (Const32 [c])
(Sub32
(Rsh32x64
(Add32
mul:(Hmul32
(Const32 [m])
x)
x)
(Const64 [s]))
(Rsh32x64
x
(Const64 [31])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int32(smagic32(c).m) && s == smagic32(c).s
&& x.Op != OpConst32 && sdivisibleOK32(c)
=> (Leq32U
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
(RotateLeft32 <typ.UInt32>
(Add32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int32(sdivisible32(c).m)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
x)
(Const32 <typ.UInt32> [int32(sdivisible32(c).a)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const32 <typ.UInt32> [int32(32-sdivisible32(c).k)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const32 <typ.UInt32> [int32(sdivisible32(c).max)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Eq64 x (Mul64 (Const64 [c])
(Sub64
(Rsh64x64
mul:(Hmul64
(Const64 [m])
x)
(Const64 [s]))
(Rsh64x64
x
(Const64 [63])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(smagic64(c).m/2) && s == smagic64(c).s-1
&& x.Op != OpConst64 && sdivisibleOK64(c)
=> (Leq64U
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
(RotateLeft64 <typ.UInt64>
(Add64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(sdivisible64(c).m)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
x)
(Const64 <typ.UInt64> [int64(sdivisible64(c).a)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const64 <typ.UInt64> [64-sdivisible64(c).k])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const64 <typ.UInt64> [int64(sdivisible64(c).max)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Eq64 x (Mul64 (Const64 [c])
(Sub64
(Rsh64x64
(Add64
mul:(Hmul64
(Const64 [m])
x)
x)
(Const64 [s]))
(Rsh64x64
x
(Const64 [63])))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(smagic64(c).m) && s == smagic64(c).s
&& x.Op != OpConst64 && sdivisibleOK64(c)
=> (Leq64U
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
(RotateLeft64 <typ.UInt64>
(Add64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(sdivisible64(c).m)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
x)
(Const64 <typ.UInt64> [int64(sdivisible64(c).a)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const64 <typ.UInt64> [64-sdivisible64(c).k])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
(Const64 <typ.UInt64> [int64(sdivisible64(c).max)])
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
)
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
// Divisibility check for signed integers for power of two constant are simple mask.
cmd/compile: add signed divisibility rules "Division by invariant integers using multiplication" paper by Granlund and Montgomery contains a method for directly computing divisibility (x%c == 0 for c constant) by means of the modular inverse. The method is further elaborated in "Hacker's Delight" by Warren Section 10-17 This general rule can compute divisibilty by one multiplication, and add and a compare for odd divisors and an additional rotate for even divisors. To apply the divisibility rule, we must take into account the rules to rewrite x%c = x-((x/c)*c) and (x/c) for c constant on the first optimization pass "opt". This complicates the matching as we want to match only in the cases where the result of (x/c) is not also needed. So, we must match on the expanded form of (x/c) in the expression x == c*(x/c) in the "late opt" pass after common subexpresion elimination. Note, that if there is an intermediate opt pass introduced in the future we could simplify these rules by delaying the magic division rewrite to "late opt" and matching directly on (x/c) in the intermediate opt pass. On amd64, the divisibility check is 30-45% faster. name old time/op new time/op delta` DivisiblePow2constI64-4 0.83ns ± 1% 0.82ns ± 0% ~ (p=0.079 n=5+4) DivisibleconstI64-4 2.68ns ± 1% 1.87ns ± 0% -30.33% (p=0.000 n=5+4) DivisibleWDivconstI64-4 2.69ns ± 1% 2.71ns ± 3% ~ (p=1.000 n=5+5) DivisiblePow2constI32-4 1.15ns ± 1% 1.15ns ± 0% ~ (p=0.238 n=5+4) DivisibleconstI32-4 2.24ns ± 1% 1.20ns ± 0% -46.48% (p=0.016 n=5+4) DivisibleWDivconstI32-4 2.27ns ± 1% 2.27ns ± 1% ~ (p=0.683 n=5+5) DivisiblePow2constI16-4 0.81ns ± 1% 0.82ns ± 1% ~ (p=0.135 n=5+5) DivisibleconstI16-4 2.11ns ± 2% 1.20ns ± 1% -42.99% (p=0.008 n=5+5) DivisibleWDivconstI16-4 2.23ns ± 0% 2.27ns ± 2% +1.79% (p=0.029 n=4+4) DivisiblePow2constI8-4 0.81ns ± 1% 0.81ns ± 1% ~ (p=0.286 n=5+5) DivisibleconstI8-4 2.13ns ± 3% 1.19ns ± 1% -43.84% (p=0.008 n=5+5) DivisibleWDivconstI8-4 2.23ns ± 1% 2.25ns ± 1% ~ (p=0.183 n=5+5) Fixes #30282 Fixes #15806 Change-Id: Id20d78263a4fdfe0509229ae4dfa2fede83fc1d0 Reviewed-on: https://go-review.googlesource.com/c/go/+/173998 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-05 14:05:07 -06:00
// However, we must match against the rewritten n%c == 0 -> n - c*(n/c) == 0 -> n == c*(n/c)
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
// where n/c contains fixup code to handle signed n.
((Eq8|Neq8) n (Lsh8x64
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
(Rsh8x64
(Add8 <t> n (Rsh8Ux64 <t> (Rsh8x64 <t> n (Const64 <typ.UInt64> [ 7])) (Const64 <typ.UInt64> [kbar])))
(Const64 <typ.UInt64> [k]))
(Const64 <typ.UInt64> [k]))
) && k > 0 && k < 7 && kbar == 8 - k
=> ((Eq8|Neq8) (And8 <t> n (Const8 <t> [1<<uint(k)-1])) (Const8 <t> [0]))
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
((Eq16|Neq16) n (Lsh16x64
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
(Rsh16x64
(Add16 <t> n (Rsh16Ux64 <t> (Rsh16x64 <t> n (Const64 <typ.UInt64> [15])) (Const64 <typ.UInt64> [kbar])))
(Const64 <typ.UInt64> [k]))
(Const64 <typ.UInt64> [k]))
) && k > 0 && k < 15 && kbar == 16 - k
=> ((Eq16|Neq16) (And16 <t> n (Const16 <t> [1<<uint(k)-1])) (Const16 <t> [0]))
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
((Eq32|Neq32) n (Lsh32x64
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
(Rsh32x64
(Add32 <t> n (Rsh32Ux64 <t> (Rsh32x64 <t> n (Const64 <typ.UInt64> [31])) (Const64 <typ.UInt64> [kbar])))
(Const64 <typ.UInt64> [k]))
(Const64 <typ.UInt64> [k]))
) && k > 0 && k < 31 && kbar == 32 - k
=> ((Eq32|Neq32) (And32 <t> n (Const32 <t> [1<<uint(k)-1])) (Const32 <t> [0]))
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
((Eq64|Neq64) n (Lsh64x64
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
(Rsh64x64
(Add64 <t> n (Rsh64Ux64 <t> (Rsh64x64 <t> n (Const64 <typ.UInt64> [63])) (Const64 <typ.UInt64> [kbar])))
(Const64 <typ.UInt64> [k]))
(Const64 <typ.UInt64> [k]))
) && k > 0 && k < 63 && kbar == 64 - k
=> ((Eq64|Neq64) (And64 <t> n (Const64 <t> [1<<uint(k)-1])) (Const64 <t> [0]))
cmd/compile: add signed divisibility by power of 2 rules For powers of two (c=1<<k), the divisibility check x%c == 0 can be made just by checking the trailing zeroes via a mask x&(c-1) == 0 even for signed integers. This avoids division fix-ups when just divisibility check is needed. To apply this rule, we match on the fixed-up version of the division. This is neccessary because the mod and division rewrite rules are already applied during the initial opt pass. The speed up on amd64 due to elimination of unneccessary fix-up code is ~55%: name old time/op new time/op delta DivconstI64-4 2.08ns ± 0% 2.09ns ± 1% ~ (p=0.730 n=5+5) DivisiblePow2constI64-4 1.78ns ± 1% 0.81ns ± 1% -54.66% (p=0.008 n=5+5) DivconstU64-4 2.08ns ± 0% 2.08ns ± 0% ~ (p=0.683 n=5+5) DivconstI32-4 1.53ns ± 0% 1.53ns ± 1% ~ (p=0.968 n=4+5) DivisiblePow2constI32-4 1.79ns ± 1% 0.81ns ± 1% -54.97% (p=0.008 n=5+5) DivconstU32-4 1.78ns ± 1% 1.80ns ± 2% ~ (p=0.206 n=5+5) DivconstI16-4 1.54ns ± 2% 1.54ns ± 0% ~ (p=0.238 n=5+4) DivisiblePow2constI16-4 1.78ns ± 0% 0.81ns ± 1% -54.72% (p=0.000 n=4+5) DivconstU16-4 1.00ns ± 5% 1.01ns ± 1% ~ (p=0.119 n=5+5) DivconstI8-4 1.54ns ± 0% 1.54ns ± 2% ~ (p=0.571 n=4+5) DivisiblePow2constI8-4 1.78ns ± 0% 0.82ns ± 8% -53.71% (p=0.008 n=5+5) DivconstU8-4 0.93ns ± 1% 0.93ns ± 1% ~ (p=0.643 n=5+5) A follow-up CL will address the general case of x%c == 0 for signed integers. Updates #15806 Change-Id: Iabadbbe369b6e0998c8ce85d038ebc236142e42a Reviewed-on: https://go-review.googlesource.com/c/go/+/173557 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-04-23 22:04:38 -06:00
(Eq(8|16|32|64) s:(Sub(8|16|32|64) x y) (Const(8|16|32|64) [0])) && s.Uses == 1 => (Eq(8|16|32|64) x y)
(Neq(8|16|32|64) s:(Sub(8|16|32|64) x y) (Const(8|16|32|64) [0])) && s.Uses == 1 => (Neq(8|16|32|64) x y)
cmd/compile: optimize a - b == 0 into a == b These rules trigger 1141 times during make.bash. Shrinks a few object files a tiny bit: name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% +0.00% (p=0.008 n=5+5) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% ~ (all equal) Compiler 6.25MB ± 0% 6.25MB ± 0% -0.00% (p=0.008 n=5+5) Flate 304kB ± 0% 304kB ± 0% -0.01% (p=0.008 n=5+5) GoParser 370kB ± 0% 370kB ± 0% ~ (all equal) Reflect 1.27MB ± 0% 1.27MB ± 0% ~ (all equal) Tar 421kB ± 0% 421kB ± 0% -0.05% (p=0.008 n=5+5) XML 518kB ± 0% 518kB ± 0% ~ (all equal) archive/tar benchmarks: name old time/op new time/op delta /Writer/USTAR-8 3.97µs ± 1% 3.88µs ± 0% -2.26% (p=0.000 n=26+26) /Writer/GNU-8 4.67µs ± 0% 4.54µs ± 1% -2.72% (p=0.000 n=28+27) /Writer/PAX-8 8.20µs ± 0% 8.01µs ± 0% -2.32% (p=0.000 n=29+29) /Reader/USTAR-8 3.61µs ± 0% 3.54µs ± 1% -2.04% (p=0.000 n=25+28) /Reader/GNU-8 2.27µs ± 2% 2.17µs ± 0% -4.08% (p=0.000 n=30+28) /Reader/PAX-8 7.75µs ± 0% 7.63µs ± 0% -1.60% (p=0.000 n=28+28) [Geo mean] 4.61µs 4.50µs -2.51% Change-Id: Ib4dfade5069a7463ccaba073ea91c8213e9714a0 Reviewed-on: https://go-review.googlesource.com/110235 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ilya Tocar <ilya.tocar@intel.com> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-27 08:50:39 -07:00
// Optimize bitsets
(Eq8 (And8 <t> x (Const8 <t> [y])) (Const8 <t> [y])) && oneBit8(y)
=> (Neq8 (And8 <t> x (Const8 <t> [y])) (Const8 <t> [0]))
(Eq16 (And16 <t> x (Const16 <t> [y])) (Const16 <t> [y])) && oneBit16(y)
=> (Neq16 (And16 <t> x (Const16 <t> [y])) (Const16 <t> [0]))
(Eq32 (And32 <t> x (Const32 <t> [y])) (Const32 <t> [y])) && oneBit32(y)
=> (Neq32 (And32 <t> x (Const32 <t> [y])) (Const32 <t> [0]))
(Eq64 (And64 <t> x (Const64 <t> [y])) (Const64 <t> [y])) && oneBit64(y)
=> (Neq64 (And64 <t> x (Const64 <t> [y])) (Const64 <t> [0]))
(Neq8 (And8 <t> x (Const8 <t> [y])) (Const8 <t> [y])) && oneBit8(y)
=> (Eq8 (And8 <t> x (Const8 <t> [y])) (Const8 <t> [0]))
(Neq16 (And16 <t> x (Const16 <t> [y])) (Const16 <t> [y])) && oneBit16(y)
=> (Eq16 (And16 <t> x (Const16 <t> [y])) (Const16 <t> [0]))
(Neq32 (And32 <t> x (Const32 <t> [y])) (Const32 <t> [y])) && oneBit32(y)
=> (Eq32 (And32 <t> x (Const32 <t> [y])) (Const32 <t> [0]))
(Neq64 (And64 <t> x (Const64 <t> [y])) (Const64 <t> [y])) && oneBit64(y)
=> (Eq64 (And64 <t> x (Const64 <t> [y])) (Const64 <t> [0]))
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// Reassociate expressions involving
// constants such that constants come first,
// exposing obvious constant-folding opportunities.
cmd/compile: automatically handle commuting ops in rewrite rules Note that this is a redo of an undo of the original buggy CL 38666. We have lots of rewrite rules that vary only in the fact that we have 2 versions for the 2 different orderings of various commuting ops. For example: (ADDL x (MOVLconst [c])) -> (ADDLconst [c] x) (ADDL (MOVLconst [c]) x) -> (ADDLconst [c] x) It can get unwieldly quickly, especially when there is more than one commuting op in a rule. Our existing "fix" for this problem is to have rules that canonicalize the operations first. For example: (Eq64 x (Const64 <t> [c])) && x.Op != OpConst64 -> (Eq64 (Const64 <t> [c]) x) Subsequent rules can then assume if there is a constant arg to Eq64, it will be the first one. This fix kinda works, but it is fragile and only works when we remember to include the required extra rules. The fundamental problem is that the rule matcher doesn't know anything about commuting ops. This CL fixes that fact. We already have information about which ops commute. (The register allocator takes advantage of commutivity.) The rule generator now automatically generates multiple rules for a single source rule when there are commutative ops in the rule. We can now drop all of our almost-duplicate source-level rules and the canonicalization rules. I have some CLs in progress that will be a lot less verbose when the rule generator handles commutivity for me. I had to reorganize the load-combining rules a bit. The 8-way OR rules generated 128 different reorderings, which was causing the generator to put too much code in the rewrite*.go files (the big ones were going from 25K lines to 132K lines). Instead I reorganized the rules to combine pairs of loads at a time. The generated rule files are now actually a bit (5%) smaller. Make.bash times are ~unchanged. Compiler benchmarks are not observably different. Probably because we don't spend much compiler time in rule matching anyway. I've also done a pass over all of our ops adding commutative markings for ops which hadn't had them previously. Fixes #18292 Change-Id: Ic1c0e43fbf579539f459971625f69690c9ab8805 Reviewed-on: https://go-review.googlesource.com/38801 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2017-03-30 03:30:22 +00:00
// Reassociate (op (op y C) x) to (op C (op x y)) or similar, where C
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// is constant, which pushes constants to the outside
// of the expression. At that point, any constant-folding
// opportunities should be obvious.
cmd/compile: convert 386 port to use addressing modes pass (take 2) Retrying CL 222782, with a fix that will hopefully stop the random crashing. The issue with the previous CL is that it does pointer arithmetic in a way that may briefly generate an out-of-bounds pointer. If an interrupt happens to occur in that state, the referenced object may be collected incorrectly. Suppose there was code that did s[x+c]. The previous CL had a rule to the effect of ptr + (x + c) -> c + (ptr + x). But ptr+x is not guaranteed to point to the same object as ptr. In contrast, ptr+(x+c) is guaranteed to point to the same object as ptr, because we would have already checked that x+c is in bounds. For example, strconv.trim used to have this code: MOVZX -0x1(BX)(DX*1), BP CMPL $0x30, AL After CL 222782, it had this code: LEAL 0(BX)(DX*1), BP CMPB $0x30, -0x1(BP) An interrupt between those last two instructions could see BP pointing outside the backing store of the slice involved. It's really hard to actually demonstrate a bug. First, you need to have an interrupt occur at exactly the right time. Then, there must be no other pointers to the object in question. Since the interrupted frame will be scanned conservatively, there can't even be a dead pointer in another register or on the stack. (In the example above, a bug can't happen because BX still holds the original pointer.) Then, the object in question needs to be collected (or at least scanned?) before the interrupted code continues. This CL needs to handle load combining somewhat differently than CL 222782 because of the new restriction on arithmetic. That's the only real difference (other than removing the bad rules) from that old CL. This bug is also present in the amd64 rewrite rules, and we haven't seen any crashing as a result. I will fix up that code similarly to this one in a separate CL. Update #37881 Change-Id: I5f0d584d9bef4696bfe89a61ef0a27c8d507329f Reviewed-on: https://go-review.googlesource.com/c/go/+/225798 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2020-03-24 13:39:44 -07:00
// Note: don't include AddPtr here! In order to maintain the
// invariant that pointers must stay within the pointed-to object,
// we can't pull part of a pointer computation above the AddPtr.
// See issue 37881.
// Note: we don't need to handle any (x-C) cases because we already rewrite
// (x-C) to (x+(-C)).
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// x + (C + z) -> C + (x + z)
(Add64 (Add64 i:(Const64 <t>) z) x) && (z.Op != OpConst64 && x.Op != OpConst64) => (Add64 i (Add64 <t> z x))
(Add32 (Add32 i:(Const32 <t>) z) x) && (z.Op != OpConst32 && x.Op != OpConst32) => (Add32 i (Add32 <t> z x))
(Add16 (Add16 i:(Const16 <t>) z) x) && (z.Op != OpConst16 && x.Op != OpConst16) => (Add16 i (Add16 <t> z x))
(Add8 (Add8 i:(Const8 <t>) z) x) && (z.Op != OpConst8 && x.Op != OpConst8) => (Add8 i (Add8 <t> z x))
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// x + (C - z) -> C + (x - z)
(Add64 (Sub64 i:(Const64 <t>) z) x) && (z.Op != OpConst64 && x.Op != OpConst64) => (Add64 i (Sub64 <t> x z))
(Add32 (Sub32 i:(Const32 <t>) z) x) && (z.Op != OpConst32 && x.Op != OpConst32) => (Add32 i (Sub32 <t> x z))
(Add16 (Sub16 i:(Const16 <t>) z) x) && (z.Op != OpConst16 && x.Op != OpConst16) => (Add16 i (Sub16 <t> x z))
(Add8 (Sub8 i:(Const8 <t>) z) x) && (z.Op != OpConst8 && x.Op != OpConst8) => (Add8 i (Sub8 <t> x z))
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// x - (C - z) -> x + (z - C) -> (x + z) - C
(Sub64 x (Sub64 i:(Const64 <t>) z)) && (z.Op != OpConst64 && x.Op != OpConst64) => (Sub64 (Add64 <t> x z) i)
(Sub32 x (Sub32 i:(Const32 <t>) z)) && (z.Op != OpConst32 && x.Op != OpConst32) => (Sub32 (Add32 <t> x z) i)
(Sub16 x (Sub16 i:(Const16 <t>) z)) && (z.Op != OpConst16 && x.Op != OpConst16) => (Sub16 (Add16 <t> x z) i)
(Sub8 x (Sub8 i:(Const8 <t>) z)) && (z.Op != OpConst8 && x.Op != OpConst8) => (Sub8 (Add8 <t> x z) i)
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// x - (z + C) -> x + (-z - C) -> (x - z) - C
(Sub64 x (Add64 z i:(Const64 <t>))) && (z.Op != OpConst64 && x.Op != OpConst64) => (Sub64 (Sub64 <t> x z) i)
(Sub32 x (Add32 z i:(Const32 <t>))) && (z.Op != OpConst32 && x.Op != OpConst32) => (Sub32 (Sub32 <t> x z) i)
(Sub16 x (Add16 z i:(Const16 <t>))) && (z.Op != OpConst16 && x.Op != OpConst16) => (Sub16 (Sub16 <t> x z) i)
(Sub8 x (Add8 z i:(Const8 <t>))) && (z.Op != OpConst8 && x.Op != OpConst8) => (Sub8 (Sub8 <t> x z) i)
// (C - z) - x -> C - (z + x)
(Sub64 (Sub64 i:(Const64 <t>) z) x) && (z.Op != OpConst64 && x.Op != OpConst64) => (Sub64 i (Add64 <t> z x))
(Sub32 (Sub32 i:(Const32 <t>) z) x) && (z.Op != OpConst32 && x.Op != OpConst32) => (Sub32 i (Add32 <t> z x))
(Sub16 (Sub16 i:(Const16 <t>) z) x) && (z.Op != OpConst16 && x.Op != OpConst16) => (Sub16 i (Add16 <t> z x))
(Sub8 (Sub8 i:(Const8 <t>) z) x) && (z.Op != OpConst8 && x.Op != OpConst8) => (Sub8 i (Add8 <t> z x))
// (z + C) -x -> C + (z - x)
(Sub64 (Add64 z i:(Const64 <t>)) x) && (z.Op != OpConst64 && x.Op != OpConst64) => (Add64 i (Sub64 <t> z x))
(Sub32 (Add32 z i:(Const32 <t>)) x) && (z.Op != OpConst32 && x.Op != OpConst32) => (Add32 i (Sub32 <t> z x))
(Sub16 (Add16 z i:(Const16 <t>)) x) && (z.Op != OpConst16 && x.Op != OpConst16) => (Add16 i (Sub16 <t> z x))
(Sub8 (Add8 z i:(Const8 <t>)) x) && (z.Op != OpConst8 && x.Op != OpConst8) => (Add8 i (Sub8 <t> z x))
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// x & (C & z) -> C & (x & z)
(And64 (And64 i:(Const64 <t>) z) x) && (z.Op != OpConst64 && x.Op != OpConst64) => (And64 i (And64 <t> z x))
(And32 (And32 i:(Const32 <t>) z) x) && (z.Op != OpConst32 && x.Op != OpConst32) => (And32 i (And32 <t> z x))
(And16 (And16 i:(Const16 <t>) z) x) && (z.Op != OpConst16 && x.Op != OpConst16) => (And16 i (And16 <t> z x))
(And8 (And8 i:(Const8 <t>) z) x) && (z.Op != OpConst8 && x.Op != OpConst8) => (And8 i (And8 <t> z x))
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// x | (C | z) -> C | (x | z)
(Or64 (Or64 i:(Const64 <t>) z) x) && (z.Op != OpConst64 && x.Op != OpConst64) => (Or64 i (Or64 <t> z x))
(Or32 (Or32 i:(Const32 <t>) z) x) && (z.Op != OpConst32 && x.Op != OpConst32) => (Or32 i (Or32 <t> z x))
(Or16 (Or16 i:(Const16 <t>) z) x) && (z.Op != OpConst16 && x.Op != OpConst16) => (Or16 i (Or16 <t> z x))
(Or8 (Or8 i:(Const8 <t>) z) x) && (z.Op != OpConst8 && x.Op != OpConst8) => (Or8 i (Or8 <t> z x))
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// x ^ (C ^ z) -> C ^ (x ^ z)
(Xor64 (Xor64 i:(Const64 <t>) z) x) && (z.Op != OpConst64 && x.Op != OpConst64) => (Xor64 i (Xor64 <t> z x))
(Xor32 (Xor32 i:(Const32 <t>) z) x) && (z.Op != OpConst32 && x.Op != OpConst32) => (Xor32 i (Xor32 <t> z x))
(Xor16 (Xor16 i:(Const16 <t>) z) x) && (z.Op != OpConst16 && x.Op != OpConst16) => (Xor16 i (Xor16 <t> z x))
(Xor8 (Xor8 i:(Const8 <t>) z) x) && (z.Op != OpConst8 && x.Op != OpConst8) => (Xor8 i (Xor8 <t> z x))
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// x * (D * z) = D * (x * z)
(Mul64 (Mul64 i:(Const64 <t>) z) x) && (z.Op != OpConst64 && x.Op != OpConst64) => (Mul64 i (Mul64 <t> x z))
(Mul32 (Mul32 i:(Const32 <t>) z) x) && (z.Op != OpConst32 && x.Op != OpConst32) => (Mul32 i (Mul32 <t> x z))
(Mul16 (Mul16 i:(Const16 <t>) z) x) && (z.Op != OpConst16 && x.Op != OpConst16) => (Mul16 i (Mul16 <t> x z))
(Mul8 (Mul8 i:(Const8 <t>) z) x) && (z.Op != OpConst8 && x.Op != OpConst8) => (Mul8 i (Mul8 <t> x z))
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// C + (D + x) -> (C + D) + x
(Add64 (Const64 <t> [c]) (Add64 (Const64 <t> [d]) x)) => (Add64 (Const64 <t> [c+d]) x)
(Add32 (Const32 <t> [c]) (Add32 (Const32 <t> [d]) x)) => (Add32 (Const32 <t> [c+d]) x)
(Add16 (Const16 <t> [c]) (Add16 (Const16 <t> [d]) x)) => (Add16 (Const16 <t> [c+d]) x)
(Add8 (Const8 <t> [c]) (Add8 (Const8 <t> [d]) x)) => (Add8 (Const8 <t> [c+d]) x)
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// C + (D - x) -> (C + D) - x
(Add64 (Const64 <t> [c]) (Sub64 (Const64 <t> [d]) x)) => (Sub64 (Const64 <t> [c+d]) x)
(Add32 (Const32 <t> [c]) (Sub32 (Const32 <t> [d]) x)) => (Sub32 (Const32 <t> [c+d]) x)
(Add16 (Const16 <t> [c]) (Sub16 (Const16 <t> [d]) x)) => (Sub16 (Const16 <t> [c+d]) x)
(Add8 (Const8 <t> [c]) (Sub8 (Const8 <t> [d]) x)) => (Sub8 (Const8 <t> [c+d]) x)
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// C - (D - x) -> (C - D) + x
(Sub64 (Const64 <t> [c]) (Sub64 (Const64 <t> [d]) x)) => (Add64 (Const64 <t> [c-d]) x)
(Sub32 (Const32 <t> [c]) (Sub32 (Const32 <t> [d]) x)) => (Add32 (Const32 <t> [c-d]) x)
(Sub16 (Const16 <t> [c]) (Sub16 (Const16 <t> [d]) x)) => (Add16 (Const16 <t> [c-d]) x)
(Sub8 (Const8 <t> [c]) (Sub8 (Const8 <t> [d]) x)) => (Add8 (Const8 <t> [c-d]) x)
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// C - (D + x) -> (C - D) - x
(Sub64 (Const64 <t> [c]) (Add64 (Const64 <t> [d]) x)) => (Sub64 (Const64 <t> [c-d]) x)
(Sub32 (Const32 <t> [c]) (Add32 (Const32 <t> [d]) x)) => (Sub32 (Const32 <t> [c-d]) x)
(Sub16 (Const16 <t> [c]) (Add16 (Const16 <t> [d]) x)) => (Sub16 (Const16 <t> [c-d]) x)
(Sub8 (Const8 <t> [c]) (Add8 (Const8 <t> [d]) x)) => (Sub8 (Const8 <t> [c-d]) x)
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// C & (D & x) -> (C & D) & x
(And64 (Const64 <t> [c]) (And64 (Const64 <t> [d]) x)) => (And64 (Const64 <t> [c&d]) x)
(And32 (Const32 <t> [c]) (And32 (Const32 <t> [d]) x)) => (And32 (Const32 <t> [c&d]) x)
(And16 (Const16 <t> [c]) (And16 (Const16 <t> [d]) x)) => (And16 (Const16 <t> [c&d]) x)
(And8 (Const8 <t> [c]) (And8 (Const8 <t> [d]) x)) => (And8 (Const8 <t> [c&d]) x)
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// C | (D | x) -> (C | D) | x
(Or64 (Const64 <t> [c]) (Or64 (Const64 <t> [d]) x)) => (Or64 (Const64 <t> [c|d]) x)
(Or32 (Const32 <t> [c]) (Or32 (Const32 <t> [d]) x)) => (Or32 (Const32 <t> [c|d]) x)
(Or16 (Const16 <t> [c]) (Or16 (Const16 <t> [d]) x)) => (Or16 (Const16 <t> [c|d]) x)
(Or8 (Const8 <t> [c]) (Or8 (Const8 <t> [d]) x)) => (Or8 (Const8 <t> [c|d]) x)
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// C ^ (D ^ x) -> (C ^ D) ^ x
(Xor64 (Const64 <t> [c]) (Xor64 (Const64 <t> [d]) x)) => (Xor64 (Const64 <t> [c^d]) x)
(Xor32 (Const32 <t> [c]) (Xor32 (Const32 <t> [d]) x)) => (Xor32 (Const32 <t> [c^d]) x)
(Xor16 (Const16 <t> [c]) (Xor16 (Const16 <t> [d]) x)) => (Xor16 (Const16 <t> [c^d]) x)
(Xor8 (Const8 <t> [c]) (Xor8 (Const8 <t> [d]) x)) => (Xor8 (Const8 <t> [c^d]) x)
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// C * (D * x) = (C * D) * x
(Mul64 (Const64 <t> [c]) (Mul64 (Const64 <t> [d]) x)) => (Mul64 (Const64 <t> [c*d]) x)
(Mul32 (Const32 <t> [c]) (Mul32 (Const32 <t> [d]) x)) => (Mul32 (Const32 <t> [c*d]) x)
(Mul16 (Const16 <t> [c]) (Mul16 (Const16 <t> [d]) x)) => (Mul16 (Const16 <t> [c*d]) x)
(Mul8 (Const8 <t> [c]) (Mul8 (Const8 <t> [d]) x)) => (Mul8 (Const8 <t> [c*d]) x)
cmd/compile/ssa: more aggressive constant folding Add rewrite rules that canonicalize the location of constants in expressions, and fold conststants that appear in operations that can be trivially reassociated. After this change, the compiler constant-folds expressions like "4 + x - 1" and "4 & x & 1" Benchmarks affected on darwin/amd64: name old time/op new time/op delta FmtFprintfInt-8 82.1ns ± 1% 81.7ns ± 1% -0.46% (p=0.023 n=8+9) FmtFprintfIntInt-8 122ns ± 2% 120ns ± 2% -1.48% (p=0.047 n=10+10) FmtManyArgs-8 493ns ± 0% 486ns ± 1% -1.37% (p=0.000 n=8+10) Gzip-8 230ms ± 0% 229ms ± 1% -0.46% (p=0.001 n=10+9) HTTPClientServer-8 74.5µs ± 1% 73.7µs ± 1% -1.11% (p=0.000 n=10+10) JSONDecode-8 51.7ms ± 0% 51.9ms ± 1% +0.42% (p=0.017 n=10+9) RegexpMatchEasy0_32-8 82.6ns ± 1% 81.7ns ± 0% -1.02% (p=0.000 n=9+8) RegexpMatchMedium_32-8 121ns ± 1% 120ns ± 1% -1.48% (p=0.001 n=10+10) Revcomp-8 426ms ± 1% 400ms ± 1% -6.16% (p=0.000 n=10+10) TimeFormat-8 330ns ± 1% 327ns ± 0% -0.82% (p=0.000 n=10+10) name old speed new speed delta Gzip-8 84.4MB/s ± 0% 84.8MB/s ± 1% +0.47% (p=0.001 n=10+9) JSONDecode-8 37.6MB/s ± 0% 37.4MB/s ± 1% -0.42% (p=0.016 n=10+9) RegexpMatchEasy0_32-8 387MB/s ± 1% 392MB/s ± 0% +1.06% (p=0.000 n=9+8) RegexpMatchMedium_32-8 8.21MB/s ± 1% 8.34MB/s ± 1% +1.58% (p=0.000 n=10+9) Revcomp-8 597MB/s ± 1% 636MB/s ± 1% +6.57% (p=0.000 n=10+10) Change-Id: Ie37ff91605b76a984a8400dfd1e34f50bf61c864 Reviewed-on: https://go-review.googlesource.com/37290 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-20 08:43:54 -08:00
// floating point optimizations
(Mul(32|64)F x (Const(32|64)F [1])) => x
(Mul32F x (Const32F [-1])) => (Neg32F x)
(Mul64F x (Const64F [-1])) => (Neg64F x)
(Mul32F x (Const32F [2])) => (Add32F x x)
(Mul64F x (Const64F [2])) => (Add64F x x)
(Div32F x (Const32F <t> [c])) && reciprocalExact32(c) => (Mul32F x (Const32F <t> [1/c]))
(Div64F x (Const64F <t> [c])) && reciprocalExact64(c) => (Mul64F x (Const64F <t> [1/c]))
// rewrite single-precision sqrt expression "float32(math.Sqrt(float64(x)))"
(Cvt64Fto32F sqrt0:(Sqrt (Cvt32Fto64F x))) && sqrt0.Uses==1 => (Sqrt32 x)
(Sqrt (Const64F [c])) && !math.IsNaN(math.Sqrt(c)) => (Const64F [math.Sqrt(c)])
// for rewriting results of some late-expanded rewrites (below)
cmd/compile: disallow rewrite rules from declaring reserved names If I change a rule in ARM64.rules to use the variable name "b" in a conflicting way, rulegen would previously not complain, and the compiler would later give a confusing error: $ go run *.go && go build cmd/compile/internal/ssa # cmd/compile/internal/ssa ../rewriteARM64.go:24236:10: b.NewValue0 undefined (type int64 has no field or method NewValue0) Make rulegen complain early about those cases. Sometimes they might happen to be harmless, but in general they can easily cause confusion or unintended effect due to shadowing. After the change, with the same conflicting rule: $ go run *.go && go build cmd/compile/internal/ssa 2021/03/22 11:31:49 rule ARM64.rules:495 uses the reserved name b exit status 1 Note that 24 existing rules were using reserved names. It seems like the shadowing was harmless, as it wasn't causing typechecking issues nor did it seem to cause unintended behavior when the rule rewrite code ran. The bool values "b" were renamed "t", since that seems to have a precedent in other rules and in the fmt package. Sequential values like "a b c" were renamed to "x y z", since "b" is reserved. Finally, "typ" was renamed to "_typ", since there doesn't seem to be an obviously better answer. Passes all three of: $ GOARCH=amd64 go build -toolexec 'toolstash -cmp' -a std $ GOARCH=arm64 go build -toolexec 'toolstash -cmp' -a std $ GOARCH=mips64 go build -toolexec 'toolstash -cmp' -a std Fixes #45154. Change-Id: I1cce194dc7b477886a9c218c17973e996bcedccf Reviewed-on: https://go-review.googlesource.com/c/go/+/303549 Trust: Daniel Martí <mvdan@mvdan.cc> Run-TryBot: Daniel Martí <mvdan@mvdan.cc> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2021-03-22 11:35:02 +01:00
(SelectN [0] (MakeResult x ___)) => x
(SelectN [1] (MakeResult x y ___)) => y
(SelectN [2] (MakeResult x y z ___)) => z
// for late-expanded calls, recognize newobject and remove zeroing and nilchecks
(Zero (SelectN [0] call:(StaticLECall _ _)) mem:(SelectN [1] call))
&& isSameCall(call.Aux, "runtime.newobject")
=> mem
(Store (SelectN [0] call:(StaticLECall _ _)) x mem:(SelectN [1] call))
&& isConstZero(x)
&& isSameCall(call.Aux, "runtime.newobject")
=> mem
(Store (OffPtr (SelectN [0] call:(StaticLECall _ _))) x mem:(SelectN [1] call))
&& isConstZero(x)
&& isSameCall(call.Aux, "runtime.newobject")
=> mem
(NilCheck ptr:(SelectN [0] call:(StaticLECall _ _)) _)
&& isSameCall(call.Aux, "runtime.newobject")
&& warnRule(fe.Debug_checknil(), v, "removed nil check")
=> ptr
(NilCheck ptr:(OffPtr (SelectN [0] call:(StaticLECall _ _))) _)
&& isSameCall(call.Aux, "runtime.newobject")
&& warnRule(fe.Debug_checknil(), v, "removed nil check")
=> ptr
// Addresses of globals are always non-nil.
(NilCheck ptr:(Addr {_} (SB)) _) => ptr
(NilCheck ptr:(Convert (Addr {_} (SB)) _) _) => ptr
// Nil checks of nil checks are redundant.
// See comment at the end of https://go-review.googlesource.com/c/go/+/537775.
(NilCheck ptr:(NilCheck _ _) _ ) => ptr
// for late-expanded calls, recognize memequal applied to a single constant byte
// Support is limited by 1, 2, 4, 8 byte sizes
(StaticLECall {callAux} sptr (Addr {scon} (SB)) (Const64 [1]) mem)
&& isSameCall(callAux, "runtime.memequal")
&& symIsRO(scon)
=> (MakeResult (Eq8 (Load <typ.Int8> sptr mem) (Const8 <typ.Int8> [int8(read8(scon,0))])) mem)
(StaticLECall {callAux} (Addr {scon} (SB)) sptr (Const64 [1]) mem)
&& isSameCall(callAux, "runtime.memequal")
&& symIsRO(scon)
=> (MakeResult (Eq8 (Load <typ.Int8> sptr mem) (Const8 <typ.Int8> [int8(read8(scon,0))])) mem)
(StaticLECall {callAux} sptr (Addr {scon} (SB)) (Const64 [2]) mem)
&& isSameCall(callAux, "runtime.memequal")
&& symIsRO(scon)
&& canLoadUnaligned(config)
=> (MakeResult (Eq16 (Load <typ.Int16> sptr mem) (Const16 <typ.Int16> [int16(read16(scon,0,config.ctxt.Arch.ByteOrder))])) mem)
(StaticLECall {callAux} (Addr {scon} (SB)) sptr (Const64 [2]) mem)
&& isSameCall(callAux, "runtime.memequal")
&& symIsRO(scon)
&& canLoadUnaligned(config)
=> (MakeResult (Eq16 (Load <typ.Int16> sptr mem) (Const16 <typ.Int16> [int16(read16(scon,0,config.ctxt.Arch.ByteOrder))])) mem)
(StaticLECall {callAux} sptr (Addr {scon} (SB)) (Const64 [4]) mem)
&& isSameCall(callAux, "runtime.memequal")
&& symIsRO(scon)
&& canLoadUnaligned(config)
=> (MakeResult (Eq32 (Load <typ.Int32> sptr mem) (Const32 <typ.Int32> [int32(read32(scon,0,config.ctxt.Arch.ByteOrder))])) mem)
(StaticLECall {callAux} (Addr {scon} (SB)) sptr (Const64 [4]) mem)
&& isSameCall(callAux, "runtime.memequal")
&& symIsRO(scon)
&& canLoadUnaligned(config)
=> (MakeResult (Eq32 (Load <typ.Int32> sptr mem) (Const32 <typ.Int32> [int32(read32(scon,0,config.ctxt.Arch.ByteOrder))])) mem)
(StaticLECall {callAux} sptr (Addr {scon} (SB)) (Const64 [8]) mem)
&& isSameCall(callAux, "runtime.memequal")
&& symIsRO(scon)
&& canLoadUnaligned(config) && config.PtrSize == 8
=> (MakeResult (Eq64 (Load <typ.Int64> sptr mem) (Const64 <typ.Int64> [int64(read64(scon,0,config.ctxt.Arch.ByteOrder))])) mem)
(StaticLECall {callAux} (Addr {scon} (SB)) sptr (Const64 [8]) mem)
&& isSameCall(callAux, "runtime.memequal")
&& symIsRO(scon)
&& canLoadUnaligned(config) && config.PtrSize == 8
=> (MakeResult (Eq64 (Load <typ.Int64> sptr mem) (Const64 <typ.Int64> [int64(read64(scon,0,config.ctxt.Arch.ByteOrder))])) mem)
(StaticLECall {callAux} _ _ (Const64 [0]) mem)
&& isSameCall(callAux, "runtime.memequal")
=> (MakeResult (ConstBool <typ.Bool> [true]) mem)
(Static(Call|LECall) {callAux} p q _ mem)
&& isSameCall(callAux, "runtime.memequal")
&& isSamePtr(p, q)
=> (MakeResult (ConstBool <typ.Bool> [true]) mem)
cmd/compile: inline known-size memclrNoHeapPointers calls This patch rewrites known-size calls to memclrNoHeapPointers with an OpZero. This significantly improves performance and lets some clears get DSE'd. One of the cases where this applies is zeroing a known-size array, example: var x [256]int8 ... for a := range x { x[a] = 0 } Other cases can be found in the runtime itself where memclrNoHeapPointers is sometimes directly invoked with a constant. It seems that for some sized-clears on some architectures (AMD64, maybe others) the memcrlNoHeapPointers is more performant than OpZero. See the issue #56997 for more details. Benches ARM (M1 Pro): name old time/op new time/op delta MemclrKnownSize1-10 2.03ns ± 0% 0.31ns ± 0% -84.69% (p=0.000 n=18+19) MemclrKnownSize2-10 1.97ns ± 0% 0.31ns ± 0% -84.19% (p=0.000 n=12+19) MemclrKnownSize4-10 2.02ns ± 0% 0.31ns ± 0% -84.56% (p=0.000 n=12+20) MemclrKnownSize8-10 2.02ns ± 0% 0.31ns ± 0% -84.59% (p=0.000 n=14+19) MemclrKnownSize16-10 2.15ns ± 0% 0.31ns ± 0% -85.50% (p=0.000 n=18+19) MemclrKnownSize32-10 2.48ns ± 0% 0.31ns ± 0% -87.48% (p=0.000 n=20+19) MemclrKnownSize64-10 1.93ns ± 0% 0.62ns ± 0% -67.88% (p=0.000 n=20+19) MemclrKnownSize112-10 2.48ns ± 0% 1.80ns ± 0% -27.74% (p=0.000 n=19+20) MemclrKnownSize128-10 10.0ns ±112% 2.0ns ± 0% -79.76% (p=0.000 n=18+17) MemclrKnownSize192-10 27.4ns ±103% 2.6ns ± 0% -90.38% (p=0.000 n=16+19) MemclrKnownSize248-10 9.67ns ±43% 3.26ns ± 0% -66.29% (p=0.000 n=19+19) MemclrKnownSize256-10 85.4ns ±148% 3.3ns ± 0% -96.18% (p=0.000 n=20+20) MemclrKnownSize512-10 223ns ±54% 6ns ± 0% -97.42% (p=0.000 n=18+20) MemclrKnownSize1024-10 216ns ±26% 11ns ± 0% -95.00% (p=0.000 n=18+15) MemclrKnownSize4096-10 265ns ± 2% 88ns ± 0% -66.84% (p=0.000 n=19+17) MemclrKnownSize512KiB-10 9.91µs ± 1% 10.23µs ± 2% +3.14% (p=0.000 n=19+19) [Geo mean] 15.6ns 2.5ns -83.62% name old speed new speed delta MemclrKnownSize1-10 493MB/s ± 0% 3216MB/s ± 0% +553.04% (p=0.000 n=18+19) MemclrKnownSize2-10 1.02GB/s ± 0% 6.43GB/s ± 0% +532.33% (p=0.000 n=16+19) MemclrKnownSize4-10 1.99GB/s ± 0% 12.86GB/s ± 0% +547.67% (p=0.000 n=18+20) MemclrKnownSize8-10 3.96GB/s ± 0% 25.72GB/s ± 0% +548.81% (p=0.000 n=19+19) MemclrKnownSize16-10 7.46GB/s ± 0% 51.43GB/s ± 0% +589.72% (p=0.000 n=20+19) MemclrKnownSize32-10 12.9GB/s ± 0% 102.9GB/s ± 0% +698.60% (p=0.000 n=20+18) MemclrKnownSize64-10 33.1GB/s ± 0% 103.0GB/s ± 0% +211.34% (p=0.000 n=19+19) MemclrKnownSize112-10 45.1GB/s ± 0% 62.4GB/s ± 0% +38.38% (p=0.000 n=19+20) MemclrKnownSize128-10 13.3GB/s ±107% 63.5GB/s ± 0% +378.03% (p=0.000 n=19+18) MemclrKnownSize192-10 6.97GB/s ±139% 72.72GB/s ± 0% +943.44% (p=0.000 n=19+19) MemclrKnownSize248-10 25.9GB/s ±46% 76.1GB/s ± 0% +194.16% (p=0.000 n=20+17) MemclrKnownSize256-10 8.64GB/s ±196% 78.51GB/s ± 0% +808.19% (p=0.000 n=20+20) MemclrKnownSize512-10 2.33GB/s ±86% 89.13GB/s ± 0% +3719.50% (p=0.000 n=17+20) MemclrKnownSize1024-10 4.85GB/s ±32% 94.93GB/s ± 0% +1856.74% (p=0.000 n=18+19) MemclrKnownSize4096-10 15.4GB/s ± 2% 46.6GB/s ± 0% +201.55% (p=0.000 n=19+18) MemclrKnownSize512KiB-10 52.9GB/s ± 1% 51.3GB/s ± 2% -3.04% (p=0.000 n=19+19) [Geo mean] 7.54GB/s 42.86GB/s +468.76% Intel Alder Lake 12600k: name old time/op new time/op delta MemclrKnownSize1-16 0.59ns ± 3% 0.38ns ± 6% -36.00% (p=0.000 n=19+18) MemclrKnownSize2-16 0.57ns ± 1% 0.19ns ± 5% -66.27% (p=0.000 n=19+19) MemclrKnownSize4-16 0.66ns ± 2% 0.36ns ±21% -45.12% (p=0.000 n=19+20) MemclrKnownSize8-16 0.74ns ± 1% 0.30ns ±26% -59.81% (p=0.000 n=18+20) MemclrKnownSize16-16 1.00ns ± 7% 0.21ns ± 8% -79.51% (p=0.000 n=20+19) MemclrKnownSize32-16 0.95ns ± 1% 0.40ns ± 1% -57.61% (p=0.000 n=20+18) MemclrKnownSize64-16 1.20ns ± 2% 0.41ns ± 0% -65.82% (p=0.000 n=20+18) MemclrKnownSize112-16 1.27ns ± 2% 1.03ns ± 0% -19.35% (p=0.000 n=20+18) MemclrKnownSize128-16 1.34ns ± 2% 1.03ns ± 0% -23.02% (p=0.000 n=20+20) MemclrKnownSize192-16 1.92ns ± 2% 1.44ns ± 0% -24.89% (p=0.000 n=20+16) MemclrKnownSize248-16 2.77ns ± 1% 3.29ns ± 0% +18.81% (p=0.000 n=20+16) MemclrKnownSize256-16 1.92ns ± 1% 1.86ns ± 0% -3.49% (p=0.000 n=19+15) MemclrKnownSize512-16 2.81ns ± 2% 3.49ns ± 0% +24.15% (p=0.000 n=20+17) MemclrKnownSize1024-16 4.02ns ± 1% 6.78ns ± 0% +68.44% (p=0.000 n=20+18) MemclrKnownSize4096-16 17.2ns ± 2% 14.4ns ± 0% -16.73% (p=0.000 n=20+17) MemclrKnownSize512KiB-16 6.71µs ± 1% 6.52µs ± 0% -2.85% (p=0.000 n=20+18) [Geo mean] 2.60ns 1.71ns -34.06% name old speed new speed delta MemclrKnownSize1-16 1.71GB/s ± 3% 2.67GB/s ± 6% +56.39% (p=0.000 n=19+18) MemclrKnownSize2-16 3.52GB/s ± 2% 10.43GB/s ± 6% +196.04% (p=0.000 n=20+20) MemclrKnownSize4-16 6.06GB/s ± 1% 10.83GB/s ±11% +78.63% (p=0.000 n=19+18) MemclrKnownSize8-16 10.7GB/s ± 1% 27.0GB/s ±21% +151.49% (p=0.000 n=18+20) MemclrKnownSize16-16 16.0GB/s ± 8% 78.1GB/s ± 7% +387.24% (p=0.000 n=20+19) MemclrKnownSize32-16 33.6GB/s ± 1% 79.4GB/s ± 1% +135.89% (p=0.000 n=20+18) MemclrKnownSize64-16 53.3GB/s ± 2% 155.9GB/s ± 0% +192.58% (p=0.000 n=20+18) MemclrKnownSize112-16 88.0GB/s ± 2% 109.1GB/s ± 0% +23.97% (p=0.000 n=20+18) MemclrKnownSize128-16 95.3GB/s ± 2% 123.8GB/s ± 0% +29.88% (p=0.000 n=20+20) MemclrKnownSize192-16 100GB/s ± 2% 133GB/s ± 0% +33.12% (p=0.000 n=20+17) MemclrKnownSize248-16 89.7GB/s ± 1% 75.5GB/s ± 0% -15.84% (p=0.000 n=20+19) MemclrKnownSize256-16 133GB/s ± 1% 138GB/s ± 0% +3.61% (p=0.000 n=19+14) MemclrKnownSize512-16 182GB/s ± 2% 147GB/s ± 0% -19.46% (p=0.000 n=20+17) MemclrKnownSize1024-16 254GB/s ± 1% 151GB/s ± 0% -40.64% (p=0.000 n=20+18) MemclrKnownSize4096-16 237GB/s ± 2% 285GB/s ± 0% +20.09% (p=0.000 n=20+17) MemclrKnownSize512KiB-16 78.2GB/s ± 1% 80.4GB/s ± 0% +2.93% (p=0.000 n=20+18) [Geo mean] 42.1GB/s 63.8GB/s +51.53% compilecmp linux/amd64: runtime runtime.(*pallocData).allocAll 85 -> 45 (-47.06%) runtime.(*pageAlloc).allocRange 942 -> 923 (-2.02%) runtime.(*pageAlloc).free 798 -> 774 (-3.01%) runtime.(*pageBits).clearAll 66 -> 20 (-69.70%) runtime.startCheckmarks 255 -> 246 (-3.53%) runtime.(*pallocData).freeAll 86 -> 46 (-46.51%) runtime.(*pallocBits).freeAll 66 -> 20 (-69.70%) runtime.(*consistentHeapStats).unsafeClear 66 -> 19 (-71.21%) runtime.newproc1 965 -> 933 (-3.32%) crypto/rc4 crypto/rc4.(*Cipher).Reset 78 -> 69 (-11.54%) compress/bzip2 compress/bzip2.(*reader).readBlock 2973 -> 2941 (-1.08%) image/jpeg image/jpeg.(*decoder).processDHT 1179 -> 1166 (-1.10%) index/suffixarray index/suffixarray.bucketMax_8_32 394 -> 241 (-38.83%) index/suffixarray.freq_8_32 317 -> 185 (-41.64%) index/suffixarray.freq_8_64 317 -> 178 (-43.85%) index/suffixarray.bucketMin_8_32 394 -> 243 (-38.32%) index/suffixarray.bucketMin_8_64 398 -> 234 (-41.21%) index/suffixarray.bucketMax_8_64 398 -> 234 (-41.21%) compress/flate compress/flate.(*huffmanBitWriter).generateCodegen 965 -> 838 (-13.16%) compress/flate.(*compressor).reset 429 -> 409 (-4.66%) cmd/vendor/golang.org/x/sys/unix cmd/vendor/golang.org/x/sys/unix.(*FdSet).Zero 66 -> 60 (-9.09%) cmd/vendor/golang.org/x/sys/unix.(*Ifreq).SetInet4Addr 211 -> 129 (-38.86%) cmd/vendor/golang.org/x/sys/unix.(*Ifreq).SetUint32 98 -> 14 (-85.71%) cmd/vendor/golang.org/x/sys/unix.(*Ifreq).clear 66 -> 11 (-83.33%) cmd/vendor/golang.org/x/sys/unix.(*Ifreq).SetUint16 101 -> 15 (-85.15%) cmd/vendor/golang.org/x/sys/unix.(*CPUSet).Zero 66 -> 60 (-9.09%) internal/coverage/decodemeta internal/coverage/decodemeta.(*CoverageMetaFileReader).rdUint64 325 -> 293 (-9.85%) crypto/tls crypto/tls.(*halfConn).setTrafficSecret 253 -> 247 (-2.37%) crypto/tls.(*Conn).readRecordOrCCS 10315 -> 10283 (-0.31%) crypto/tls.(*halfConn).changeCipherSpec 271 -> 261 (-3.69%) crypto/tls.(*Conn).writeRecordLocked 1765 -> 1748 (-0.96%) file before after Δ % runtime.s 512467 512164 -303 -0.059% crypto/rc4.s 955 946 -9 -0.942% compress/bzip2.s 9586 9554 -32 -0.334% image/jpeg.s 32122 32109 -13 -0.040% index/suffixarray.s 38547 37644 -903 -2.343% compress/flate.s 46668 46521 -147 -0.315% cmd/vendor/golang.org/x/sys/unix.s 118620 118301 -319 -0.269% internal/coverage/decodemeta.s 7224 7192 -32 -0.443% crypto/tls.s 288762 288697 -65 -0.023% cmd/compile/internal/ssa.s 3639799 3640727 +928 +0.025% total 20790248 20789353 -895 -0.004% src/runtime benchmarks (Linux Alder Lake 12600k): name old time/op new time/op delta MakeChan/Byte-16 26.2ns ± 2% 25.6ns ± 3% -2.05% (p=0.003 n=9+10) MakeChan/Int-16 33.9ns ± 2% 33.3ns ± 4% -1.99% (p=0.015 n=10+10) MakeChan/Ptr-16 54.2ns ± 2% 53.7ns ± 1% -0.90% (p=0.016 n=10+9) MakeChan/Struct/0-16 23.8ns ± 3% 23.4ns ± 1% -1.72% (p=0.009 n=10+8) MakeChan/Struct/32-16 55.9ns ± 2% 53.9ns ± 1% -3.48% (p=0.000 n=10+10) MakeChan/Struct/40-16 63.5ns ± 1% 61.1ns ± 2% -3.79% (p=0.000 n=10+9) ChanNonblocking-16 0.22ns ± 0% 0.22ns ± 0% +0.40% (p=0.011 n=9+8) SelectUncontended-16 4.63ns ± 1% 4.62ns ± 0% -0.35% (p=0.001 n=10+8) SelectSyncContended-16 1.58µs ± 2% 1.59µs ± 1% ~ (p=0.540 n=10+10) SelectAsyncContended-16 290ns ± 0% 291ns ± 0% +0.14% (p=0.012 n=8+9) SelectNonblock-16 0.95ns ± 1% 0.95ns ± 1% ~ (p=0.546 n=9+9) ChanUncontended-16 239ns ± 3% 242ns ± 6% ~ (p=0.886 n=9+10) ChanContended-16 17.7µs ± 1% 18.2µs ± 1% +2.87% (p=0.000 n=10+9) ChanSync-16 109ns ± 2% 109ns ± 1% ~ (p=0.342 n=10+10) ChanSyncWork-16 6.55µs ± 1% 6.53µs ± 1% ~ (p=0.101 n=10+10) ChanProdCons0-16 502ns ± 1% 499ns ± 0% -0.55% (p=0.001 n=10+9) ChanProdCons10-16 373ns ± 2% 377ns ± 1% ~ (p=0.095 n=10+9) ChanProdCons100-16 224ns ± 2% 223ns ± 3% ~ (p=0.150 n=9+10) ChanProdConsWork0-16 491ns ± 1% 484ns ± 0% -1.26% (p=0.000 n=10+9) ChanProdConsWork10-16 451ns ± 2% 448ns ± 2% ~ (p=0.210 n=8+10) ChanProdConsWork100-16 406ns ± 0% 407ns ± 1% ~ (p=0.138 n=8+8) SelectProdCons-16 509ns ± 0% 509ns ± 0% ~ (p=0.917 n=9+9) ReceiveDataFromClosedChan-16 12.1ns ± 0% 12.1ns ± 0% ~ (p=0.780 n=10+10) ChanCreation-16 22.6ns ± 1% 22.4ns ± 0% -0.72% (p=0.001 n=10+8) ChanSem-16 165ns ± 1% 166ns ± 1% +0.72% (p=0.002 n=10+10) ChanPopular-16 500µs ± 2% 498µs ± 1% ~ (p=0.218 n=10+10) ChanClosed-16 0.29ns ± 0% 0.29ns ± 0% +0.09% (p=0.019 n=9+8) CallClosure-16 1.28ns ± 0% 1.27ns ± 0% -0.51% (p=0.000 n=9+9) CallClosure1-16 1.50ns ± 0% 1.50ns ± 0% ~ (p=0.123 n=9+9) CallClosure2-16 8.86ns ± 1% 8.86ns ± 3% ~ (p=0.590 n=9+10) CallClosure3-16 8.75ns ± 2% 8.69ns ± 2% ~ (p=0.247 n=10+10) CallClosure4-16 8.65ns ± 2% 8.56ns ± 2% ~ (p=0.105 n=10+10) Complex128DivNormal-16 2.47ns ± 0% 2.47ns ± 0% ~ (p=0.790 n=10+9) Complex128DivNisNaN-16 4.44ns ± 0% 4.43ns ± 0% ~ (p=0.564 n=10+10) Complex128DivDisNaN-16 4.48ns ± 0% 4.48ns ± 0% ~ (p=0.101 n=10+10) Complex128DivNisInf-16 2.58ns ± 0% 2.58ns ± 0% ~ (p=0.808 n=10+10) Complex128DivDisInf-16 6.30ns ± 0% 6.31ns ± 0% ~ (p=0.305 n=10+10) SetTypePtr-16 0.73ns ± 1% 0.73ns ± 3% ~ (p=0.644 n=10+10) SetTypePtr8-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.127 n=10+10) SetTypePtr16-16 4.13ns ± 1% 4.12ns ± 0% ~ (p=0.109 n=10+10) SetTypePtr32-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.203 n=9+10) SetTypePtr64-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.696 n=10+10) SetTypePtr126-16 6.91ns ± 0% 6.91ns ± 0% ~ (p=0.469 n=10+10) SetTypePtr128-16 6.66ns ± 0% 6.67ns ± 0% ~ (p=0.246 n=9+10) SetTypePtrSlice-16 54.1ns ± 1% 54.1ns ± 1% ~ (p=0.509 n=9+10) SetTypeNode1-16 4.13ns ± 1% 4.12ns ± 0% ~ (p=0.342 n=10+10) SetTypeNode1Slice-16 10.1ns ± 1% 10.0ns ± 1% -1.18% (p=0.000 n=10+10) SetTypeNode8-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.137 n=8+8) SetTypeNode8Slice-16 22.6ns ± 0% 22.6ns ± 0% ~ (p=0.423 n=10+10) SetTypeNode64-16 6.90ns ± 0% 6.91ns ± 0% ~ (p=0.275 n=10+10) SetTypeNode64Slice-16 173ns ± 0% 173ns ± 0% ~ (p=0.610 n=9+10) SetTypeNode64Dead-16 5.53ns ± 0% 5.52ns ± 0% ~ (p=0.123 n=10+6) SetTypeNode64DeadSlice-16 150ns ± 0% 150ns ± 0% ~ (p=0.398 n=10+10) SetTypeNode124-16 6.90ns ± 0% 6.90ns ± 0% ~ (p=0.779 n=10+10) SetTypeNode124Slice-16 222ns ± 5% 217ns ± 0% ~ (p=0.302 n=10+10) SetTypeNode126-16 6.66ns ± 0% 6.66ns ± 0% ~ (p=0.324 n=10+9) SetTypeNode126Slice-16 218ns ± 0% 218ns ± 0% ~ (p=0.119 n=9+10) SetTypeNode128-16 9.76ns ± 0% 9.73ns ± 0% -0.31% (p=0.003 n=9+10) SetTypeNode128Slice-16 279ns ± 0% 278ns ± 0% ~ (p=0.112 n=10+9) SetTypeNode130-16 9.77ns ± 0% 9.73ns ± 0% -0.33% (p=0.002 n=10+10) SetTypeNode130Slice-16 284ns ± 0% 284ns ± 0% ~ (p=0.668 n=10+10) SetTypeNode1024-16 51.2ns ± 0% 51.6ns ± 1% ~ (p=0.080 n=9+9) SetTypeNode1024Slice-16 1.83µs ± 0% 1.82µs ± 0% ~ (p=0.115 n=10+10) Allocation-16 4.64µs ± 1% 4.37µs ± 1% -5.69% (p=0.000 n=9+9) ReadMemStats-16 5.62µs ± 2% 5.55µs ± 5% -1.36% (p=0.050 n=10+10) WriteBarrier-16 4.95ns ± 3% 4.99ns ± 3% ~ (p=0.255 n=10+10) BulkWriteBarrier-16 1.69ns ± 2% 1.63ns ± 4% -3.77% (p=0.001 n=10+10) ScanStackNoLocals-16 12.8ms ± 2% 12.9ms ± 1% +0.72% (p=0.019 n=10+10) MSpanCountAlloc/bits=64-16 1.65ns ± 0% 1.65ns ± 0% ~ (p=0.124 n=10+10) MSpanCountAlloc/bits=128-16 2.08ns ± 1% 2.06ns ± 1% -0.87% (p=0.000 n=10+10) MSpanCountAlloc/bits=256-16 2.71ns ± 1% 2.69ns ± 1% -0.74% (p=0.001 n=10+9) MSpanCountAlloc/bits=512-16 4.15ns ± 0% 4.23ns ± 2% +2.15% (p=0.000 n=10+10) MSpanCountAlloc/bits=1024-16 7.89ns ± 1% 7.89ns ± 1% ~ (p=0.867 n=10+10) Hash5-16 1.93ns ± 1% 2.01ns ± 0% +3.99% (p=0.000 n=10+8) Hash16-16 2.04ns ± 1% 2.21ns ± 1% +8.61% (p=0.000 n=10+10) Hash64-16 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.154 n=9+9) Hash1024-16 16.4ns ± 0% 16.4ns ± 0% +0.17% (p=0.020 n=9+10) Hash65536-16 886ns ± 0% 885ns ± 0% ~ (p=0.725 n=10+10) AlignedLoad-16 0.96ns ± 2% 0.95ns ± 3% ~ (p=0.123 n=10+10) UnalignedLoad-16 0.95ns ± 2% 1.01ns ± 2% +6.31% (p=0.000 n=10+10) EqEfaceConcrete-16 0.31ns ± 3% 0.33ns ± 5% +8.10% (p=0.000 n=10+10) EqIfaceConcrete-16 0.31ns ±13% 0.28ns ± 2% -9.23% (p=0.001 n=10+10) NeEfaceConcrete-16 0.29ns ± 1% 0.31ns ± 7% +5.59% (p=0.010 n=8+8) NeIfaceConcrete-16 0.28ns ± 2% 0.29ns ± 1% +4.49% (p=0.000 n=9+8) ConvT2EByteSized/bool-16 0.53ns ± 1% 0.52ns ± 1% -2.18% (p=0.000 n=10+10) ConvT2EByteSized/uint8-16 0.53ns ± 1% 0.53ns ± 0% +1.22% (p=0.000 n=10+10) ConvT2ESmall-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.774 n=9+9) ConvT2EUintptr-16 1.03ns ± 0% 1.04ns ± 0% +0.50% (p=0.000 n=10+8) ConvT2ELarge-16 14.4ns ± 2% 14.4ns ± 1% ~ (p=0.726 n=10+10) ConvT2ISmall-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.693 n=9+10) ConvT2IUintptr-16 1.03ns ± 0% 1.03ns ± 0% +0.44% (p=0.000 n=10+10) ConvT2ILarge-16 14.2ns ± 1% 14.4ns ± 1% +0.85% (p=0.007 n=9+10) ConvI2E-16 0.54ns ± 1% 0.54ns ± 0% -0.39% (p=0.037 n=10+8) ConvI2I-16 2.68ns ± 0% 2.70ns ± 1% +0.73% (p=0.000 n=9+10) AssertE2T-16 0.28ns ± 1% 0.39ns ± 5% +37.38% (p=0.000 n=10+10) AssertE2TLarge-16 0.42ns ± 2% 0.48ns ± 1% +14.92% (p=0.000 n=9+10) AssertE2I-16 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.352 n=9+9) AssertI2T-16 0.37ns ± 3% 0.34ns ± 1% -6.16% (p=0.000 n=10+10) AssertI2I-16 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.286 n=10+10) AssertI2E-16 0.54ns ± 1% 0.54ns ± 0% -0.94% (p=0.000 n=10+10) AssertE2E-16 0.41ns ± 0% 0.41ns ± 1% ~ (p=0.880 n=9+9) AssertE2T2-16 0.41ns ± 1% 0.41ns ± 1% ~ (p=0.725 n=10+10) AssertE2T2Blank-16 0.24ns ± 5% 0.21ns ± 1% -14.79% (p=0.000 n=10+9) AssertI2E2-16 0.69ns ± 0% 0.69ns ± 1% ~ (p=0.541 n=10+10) AssertI2E2Blank-16 0.26ns ± 9% 0.21ns ± 1% -18.86% (p=0.000 n=10+9) AssertE2E2-16 0.53ns ± 1% 0.53ns ± 1% +0.72% (p=0.004 n=10+10) AssertE2E2Blank-16 0.23ns ± 4% 0.21ns ± 1% -8.42% (p=0.000 n=10+10) ConvT2Ezero/zero/16-16 1.13ns ± 0% 1.14ns ± 1% ~ (p=0.583 n=9+10) ConvT2Ezero/zero/32-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.417 n=10+10) ConvT2Ezero/zero/64-16 1.03ns ± 1% 1.03ns ± 0% ~ (p=0.051 n=10+10) ConvT2Ezero/zero/str-16 1.03ns ± 0% 1.03ns ± 0% ~ (p=0.132 n=10+10) ConvT2Ezero/zero/slice-16 1.14ns ± 0% 1.15ns ± 0% +0.49% (p=0.001 n=10+10) ConvT2Ezero/zero/big-16 123ns ± 1% 123ns ± 1% ~ (p=0.171 n=10+10) ConvT2Ezero/nonzero/str-16 19.4ns ± 1% 19.3ns ± 3% ~ (p=0.548 n=9+10) ConvT2Ezero/nonzero/slice-16 22.2ns ± 2% 22.0ns ± 2% ~ (p=0.109 n=10+10) ConvT2Ezero/nonzero/big-16 123ns ± 1% 123ns ± 1% ~ (p=0.446 n=10+8) ConvT2Ezero/smallint/16-16 1.13ns ± 0% 1.14ns ± 1% ~ (p=0.362 n=10+10) ConvT2Ezero/smallint/32-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.907 n=10+9) ConvT2Ezero/smallint/64-16 1.04ns ± 0% 1.03ns ± 0% -0.38% (p=0.002 n=10+10) ConvT2Ezero/largeint/16-16 6.65ns ± 1% 6.65ns ± 2% ~ (p=0.618 n=10+9) ConvT2Ezero/largeint/32-16 6.75ns ± 3% 6.63ns ± 2% -1.77% (p=0.015 n=10+10) ConvT2Ezero/largeint/64-16 9.19ns ± 1% 9.26ns ± 2% ~ (p=0.123 n=10+10) Malloc8-16 8.66ns ± 1% 8.89ns ± 2% +2.74% (p=0.000 n=10+10) Malloc16-16 13.7ns ± 1% 13.8ns ± 1% +0.71% (p=0.022 n=10+8) MallocTypeInfo8-16 11.7ns ± 3% 11.6ns ± 2% ~ (p=0.469 n=10+10) MallocTypeInfo16-16 18.3ns ± 1% 18.2ns ± 2% ~ (p=0.251 n=9+10) MallocLargeStruct-16 195ns ± 1% 198ns ± 1% +1.65% (p=0.000 n=9+10) GoroutineSelect-16 1.10ms ± 1% 1.12ms ± 1% +1.36% (p=0.000 n=10+8) GoroutineBlocking-16 986µs ± 1% 998µs ± 1% +1.23% (p=0.002 n=10+10) GoroutineForRange-16 985µs ± 1% 1001µs ± 1% +1.68% (p=0.000 n=10+10) GoroutineIdle-16 679µs ± 1% 691µs ± 0% +1.74% (p=0.000 n=10+9) HashStringSpeed-16 5.33ns ± 5% 5.19ns ± 4% ~ (p=0.113 n=9+9) HashBytesSpeed-16 8.20ns ± 3% 8.24ns ± 1% ~ (p=0.497 n=10+9) HashInt32Speed-16 4.01ns ± 2% 3.90ns ± 4% -2.63% (p=0.011 n=9+10) HashInt64Speed-16 3.94ns ± 4% 3.79ns ± 1% -3.74% (p=0.003 n=10+9) HashStringArraySpeed-16 12.5ns ± 4% 12.3ns ± 1% ~ (p=0.055 n=10+10) MegMap-16 3.72ns ± 1% 3.73ns ± 1% ~ (p=0.484 n=9+10) MegOneMap-16 2.28ns ± 1% 2.27ns ± 1% ~ (p=0.287 n=10+10) MegEqMap-16 22.0µs ± 3% 22.3µs ± 2% +1.48% (p=0.028 n=10+9) MegEmptyMap-16 0.93ns ± 1% 0.92ns ± 1% -0.52% (p=0.030 n=10+10) SmallStrMap-16 3.77ns ± 0% 3.77ns ± 0% ~ (p=0.324 n=10+10) MapStringKeysEight_16-16 3.91ns ± 0% 3.91ns ± 0% ~ (p=0.088 n=9+9) MapStringKeysEight_32-16 3.58ns ± 1% 3.50ns ± 0% -2.11% (p=0.000 n=10+10) MapStringKeysEight_64-16 3.58ns ± 1% 3.50ns ± 0% -2.23% (p=0.000 n=10+10) MapStringKeysEight_1M-16 3.57ns ± 1% 3.50ns ± 0% -1.92% (p=0.000 n=10+10) IntMap-16 2.89ns ± 1% 2.89ns ± 0% ~ (p=0.381 n=10+10) MapFirst/1-16 1.60ns ± 1% 1.59ns ± 2% -0.49% (p=0.020 n=10+9) MapFirst/2-16 1.61ns ± 0% 1.59ns ± 1% -1.17% (p=0.001 n=10+10) MapFirst/3-16 1.61ns ± 1% 1.59ns ± 1% -1.45% (p=0.000 n=10+10) MapFirst/4-16 1.60ns ± 1% 1.59ns ± 1% -1.16% (p=0.000 n=10+10) MapFirst/5-16 1.60ns ± 1% 1.58ns ± 1% -0.98% (p=0.000 n=10+10) MapFirst/6-16 1.60ns ± 1% 1.59ns ± 1% -0.87% (p=0.001 n=10+10) MapFirst/7-16 1.60ns ± 1% 1.59ns ± 1% -0.79% (p=0.002 n=10+10) MapFirst/8-16 1.60ns ± 1% 1.59ns ± 1% -0.67% (p=0.017 n=9+10) MapFirst/9-16 2.83ns ± 0% 2.83ns ± 0% ~ (p=0.492 n=10+10) MapFirst/10-16 2.83ns ± 0% 2.84ns ± 0% +0.24% (p=0.017 n=10+10) MapFirst/11-16 2.83ns ± 0% 2.83ns ± 0% ~ (p=0.445 n=10+10) MapFirst/12-16 2.83ns ± 0% 2.83ns ± 0% ~ (p=0.564 n=10+10) MapFirst/13-16 2.83ns ± 0% 2.84ns ± 0% ~ (p=0.175 n=9+10) MapFirst/14-16 2.83ns ± 0% 2.83ns ± 0% ~ (p=0.322 n=10+9) MapFirst/15-16 2.83ns ± 0% 2.84ns ± 1% ~ (p=0.209 n=10+10) MapFirst/16-16 2.83ns ± 1% 2.84ns ± 0% ~ (p=0.238 n=10+10) MapMid/1-16 1.64ns ± 0% 1.64ns ± 0% ~ (p=0.453 n=10+9) MapMid/2-16 1.86ns ± 1% 1.86ns ± 0% ~ (p=0.764 n=10+9) MapMid/3-16 1.86ns ± 0% 1.86ns ± 1% ~ (p=1.000 n=10+10) MapMid/4-16 2.06ns ± 0% 2.06ns ± 0% -0.27% (p=0.014 n=10+9) MapMid/5-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.075 n=9+10) MapMid/6-16 2.27ns ± 0% 2.27ns ± 1% ~ (p=0.898 n=10+10) MapMid/7-16 2.27ns ± 1% 2.26ns ± 0% -0.23% (p=0.049 n=10+10) MapMid/8-16 2.47ns ± 0% 2.47ns ± 1% ~ (p=0.840 n=10+10) MapMid/9-16 4.21ns ± 7% 4.13ns ±19% ~ (p=0.315 n=10+10) MapMid/10-16 4.17ns ± 7% 4.31ns ± 5% +3.37% (p=0.021 n=10+9) MapMid/11-16 4.18ns ± 7% 4.32ns ± 6% +3.50% (p=0.015 n=10+10) MapMid/12-16 4.34ns ± 7% 4.30ns ± 5% ~ (p=0.858 n=9+10) MapMid/13-16 4.25ns ± 6% 4.28ns ± 6% ~ (p=0.489 n=9+9) MapMid/14-16 3.75ns ±23% 3.90ns ±16% ~ (p=0.353 n=10+10) MapMid/15-16 3.87ns ±25% 3.95ns ±26% ~ (p=0.315 n=10+10) MapMid/16-16 4.06ns ±19% 3.94ns ±16% ~ (p=0.796 n=10+10) MapLast/1-16 1.65ns ± 0% 1.65ns ± 0% ~ (p=0.607 n=10+10) MapLast/2-16 1.86ns ± 0% 1.86ns ± 0% +0.26% (p=0.029 n=10+10) MapLast/3-16 2.06ns ± 1% 2.06ns ± 0% ~ (p=0.689 n=8+9) MapLast/4-16 2.27ns ± 1% 2.26ns ± 0% ~ (p=0.148 n=10+9) MapLast/5-16 2.47ns ± 0% 2.47ns ± 0% ~ (p=0.385 n=9+10) MapLast/6-16 2.67ns ± 0% 2.68ns ± 0% ~ (p=0.202 n=9+10) MapLast/7-16 2.88ns ± 0% 2.88ns ± 0% ~ (p=0.751 n=10+10) MapLast/8-16 3.08ns ± 0% 3.08ns ± 0% ~ (p=0.826 n=10+9) MapLast/9-16 4.31ns ± 6% 4.54ns ± 5% ~ (p=0.070 n=9+8) MapLast/10-16 4.25ns ± 5% 4.42ns ± 6% ~ (p=0.321 n=9+8) MapLast/11-16 4.59ns ±16% 5.42ns ±44% +17.99% (p=0.019 n=10+10) MapLast/12-16 5.04ns ±19% 6.11ns ±28% +21.35% (p=0.005 n=9+10) MapLast/13-16 6.00ns ±35% 5.76ns ± 3% ~ (p=0.173 n=10+8) MapLast/14-16 4.27ns ± 5% 4.53ns ± 6% +6.14% (p=0.007 n=10+10) MapLast/15-16 4.41ns ± 1% 4.44ns ± 7% ~ (p=0.515 n=8+10) MapLast/16-16 4.18ns ± 6% 4.99ns ±18% +19.48% (p=0.000 n=10+10) MapCycle-16 7.48ns ± 2% 7.46ns ± 1% ~ (p=0.699 n=10+10) RepeatedLookupStrMapKey32-16 6.98ns ± 3% 6.73ns ± 2% -3.63% (p=0.000 n=10+10) RepeatedLookupStrMapKey1M-16 14.7µs ± 5% 14.7µs ± 4% ~ (p=0.604 n=9+10) MakeMap/[Byte]Byte-16 58.5ns ± 1% 58.5ns ± 1% ~ (p=0.780 n=10+9) MakeMap/[Int]Int-16 113ns ± 0% 113ns ± 1% ~ (p=0.100 n=8+10) NewEmptyMap-16 2.47ns ± 0% 2.47ns ± 0% ~ (p=0.638 n=10+10) NewSmallMap-16 11.5ns ± 1% 11.6ns ± 0% +1.18% (p=0.000 n=10+10) MapIter-16 42.2ns ± 0% 42.8ns ± 1% +1.50% (p=0.000 n=10+10) MapIterEmpty-16 1.85ns ± 0% 1.85ns ± 0% ~ (p=0.651 n=10+10) SameLengthMap-16 1.85ns ± 1% 1.85ns ± 0% ~ (p=0.247 n=10+10) BigKeyMap-16 7.18ns ± 1% 7.42ns ± 4% +3.33% (p=0.004 n=10+10) BigValMap-16 7.03ns ± 2% 7.19ns ± 1% +2.33% (p=0.000 n=10+9) SmallKeyMap-16 5.32ns ± 1% 5.24ns ± 1% -1.41% (p=0.000 n=10+10) MapPopulate/1-16 6.30ns ± 0% 6.41ns ± 1% +1.81% (p=0.000 n=8+10) MapPopulate/10-16 239ns ± 2% 234ns ± 2% -2.05% (p=0.001 n=9+10) MapPopulate/100-16 4.19µs ± 2% 4.22µs ± 2% ~ (p=0.171 n=10+10) MapPopulate/1000-16 52.3µs ± 1% 52.5µs ± 1% ~ (p=0.133 n=9+10) MapPopulate/10000-16 459µs ± 1% 466µs ± 2% +1.45% (p=0.005 n=10+10) MapPopulate/100000-16 4.22ms ± 2% 4.25ms ± 2% ~ (p=0.393 n=10+10) ComplexAlgMap-16 12.5ns ± 1% 12.4ns ± 1% -0.95% (p=0.022 n=10+10) GoMapClear/Reflexive/1-16 9.61ns ± 1% 9.58ns ± 0% -0.27% (p=0.027 n=10+10) GoMapClear/Reflexive/10-16 10.0ns ± 1% 10.0ns ± 1% ~ (p=0.648 n=9+9) GoMapClear/Reflexive/100-16 31.4ns ± 0% 31.4ns ± 1% ~ (p=0.305 n=9+10) GoMapClear/Reflexive/1000-16 147ns ± 0% 149ns ± 2% +1.21% (p=0.000 n=10+10) GoMapClear/Reflexive/10000-16 3.99µs ± 0% 4.00µs ± 0% +0.21% (p=0.018 n=9+10) GoMapClear/NonReflexive/1-16 41.4ns ± 2% 41.7ns ± 1% +0.55% (p=0.043 n=9+10) GoMapClear/NonReflexive/10-16 50.3ns ± 1% 50.9ns ± 1% +1.16% (p=0.000 n=10+10) GoMapClear/NonReflexive/100-16 125ns ± 0% 126ns ± 0% +0.96% (p=0.000 n=8+10) GoMapClear/NonReflexive/1000-16 1.08µs ± 0% 1.08µs ± 1% ~ (p=0.097 n=10+10) GoMapClear/NonReflexive/10000-16 8.18µs ± 2% 8.10µs ± 0% -0.91% (p=0.019 n=10+8) MapStringConversion/32/simple-16 4.66ns ± 1% 4.69ns ± 3% ~ (p=0.905 n=9+10) MapStringConversion/32/struct-16 4.65ns ± 3% 4.94ns ± 2% +6.23% (p=0.000 n=10+10) MapStringConversion/32/array-16 4.69ns ± 3% 4.72ns ± 3% ~ (p=0.631 n=10+10) MapStringConversion/64/simple-16 4.14ns ± 0% 4.14ns ± 1% ~ (p=0.342 n=10+10) MapStringConversion/64/struct-16 4.13ns ± 0% 4.13ns ± 0% ~ (p=0.809 n=10+10) MapStringConversion/64/array-16 4.13ns ± 1% 4.13ns ± 1% ~ (p=0.752 n=10+10) MapInterfaceString-16 7.90ns ±23% 8.51ns ±33% ~ (p=0.604 n=9+10) MapInterfacePtr-16 7.68ns ±29% 7.10ns ±36% ~ (p=0.353 n=10+10) NewEmptyMapHintLessThan8-16 3.70ns ± 0% 3.70ns ± 0% ~ (p=0.209 n=10+10) NewEmptyMapHintGreaterThan8-16 270ns ± 1% 272ns ± 1% +0.71% (p=0.005 n=10+9) MapPop100-16 6.45µs ± 0% 6.50µs ± 1% +0.77% (p=0.000 n=10+10) MapPop1000-16 114µs ± 1% 114µs ± 1% ~ (p=0.190 n=10+10) MapPop10000-16 2.28ms ± 2% 2.28ms ± 2% ~ (p=0.912 n=10+10) MapAssign/Int32/256-16 4.75ns ± 2% 4.82ns ± 4% ~ (p=0.101 n=10+10) MapAssign/Int32/65536-16 16.4ns ± 1% 16.7ns ± 0% +1.44% (p=0.000 n=10+9) MapAssign/Int64/256-16 4.79ns ± 5% 4.79ns ± 1% ~ (p=0.616 n=10+8) MapAssign/Int64/65536-16 17.1ns ± 1% 16.8ns ± 0% -1.28% (p=0.000 n=10+8) MapAssign/Str/256-16 6.07ns ± 6% 6.24ns ± 2% +2.84% (p=0.035 n=10+9) MapAssign/Str/65536-16 21.4ns ± 0% 21.4ns ± 3% ~ (p=0.300 n=7+10) MapOperatorAssign/Int32/256-16 4.82ns ± 3% 4.81ns ± 3% ~ (p=0.684 n=10+10) MapOperatorAssign/Int32/65536-16 16.8ns ± 1% 16.5ns ± 1% -1.68% (p=0.000 n=9+10) MapOperatorAssign/Int64/256-16 4.74ns ± 1% 4.77ns ± 3% ~ (p=0.563 n=10+9) MapOperatorAssign/Int64/65536-16 16.9ns ± 1% 17.2ns ± 1% +1.88% (p=0.000 n=10+10) MapOperatorAssign/Str/256-16 1.09µs ± 1% 1.10µs ± 2% ~ (p=0.210 n=10+10) MapOperatorAssign/Str/65536-16 184ns ± 9% 184ns ± 8% ~ (p=0.922 n=10+9) MapAppendAssign/Int32/256-16 13.8ns ±10% 14.4ns ±11% ~ (p=0.190 n=10+10) MapAppendAssign/Int32/65536-16 28.9ns ± 5% 30.7ns ± 6% +6.13% (p=0.001 n=9+10) MapAppendAssign/Int64/256-16 14.5ns ±12% 13.8ns ± 8% -5.02% (p=0.037 n=10+10) MapAppendAssign/Int64/65536-16 30.9ns ± 1% 30.4ns ± 2% -1.56% (p=0.001 n=10+10) MapAppendAssign/Str/256-16 30.2ns ± 6% 30.0ns ±10% ~ (p=0.645 n=10+10) MapAppendAssign/Str/65536-16 44.5ns ± 4% 46.8ns ± 3% +5.17% (p=0.001 n=8+9) MapDelete/Int32/100-16 18.7ns ± 0% 18.7ns ± 0% -0.27% (p=0.017 n=10+10) MapDelete/Int32/1000-16 17.6ns ± 1% 17.5ns ± 1% -0.85% (p=0.000 n=9+10) MapDelete/Int32/10000-16 18.7ns ± 0% 18.3ns ± 1% -1.92% (p=0.000 n=10+10) MapDelete/Int64/100-16 19.1ns ± 0% 19.2ns ± 0% +0.68% (p=0.000 n=10+9) MapDelete/Int64/1000-16 17.7ns ± 2% 18.3ns ± 1% +3.00% (p=0.000 n=10+10) MapDelete/Int64/10000-16 18.8ns ± 1% 19.2ns ± 0% +2.01% (p=0.000 n=10+9) MapDelete/Str/100-16 26.5ns ± 0% 26.4ns ± 1% -0.73% (p=0.000 n=10+10) MapDelete/Str/1000-16 23.5ns ± 2% 23.4ns ± 1% ~ (p=0.425 n=10+10) MapDelete/Str/10000-16 25.1ns ± 0% 25.1ns ± 1% +0.28% (p=0.037 n=10+10) MapDelete/Pointer/100-16 20.6ns ± 1% 20.6ns ± 0% ~ (p=0.117 n=10+10) MapDelete/Pointer/1000-16 19.2ns ± 1% 19.4ns ± 1% +0.97% (p=0.004 n=10+10) MapDelete/Pointer/10000-16 20.0ns ± 0% 20.1ns ± 1% +0.52% (p=0.022 n=10+10) Memmove/0-16 0.21ns ± 2% 0.21ns ± 1% ~ (p=0.671 n=10+10) Memmove/1-16 0.93ns ± 0% 0.93ns ± 0% +0.21% (p=0.034 n=10+10) Memmove/2-16 0.93ns ± 0% 0.93ns ± 0% ~ (p=0.101 n=10+10) Memmove/3-16 0.93ns ± 1% 0.93ns ± 1% +0.49% (p=0.004 n=10+10) Memmove/4-16 1.03ns ± 0% 1.03ns ± 0% ~ (p=0.260 n=10+10) Memmove/5-16 1.13ns ± 0% 1.13ns ± 0% +0.20% (p=0.034 n=10+10) Memmove/6-16 1.13ns ± 0% 1.13ns ± 1% ~ (p=0.126 n=10+10) Memmove/7-16 1.13ns ± 0% 1.13ns ± 1% +0.22% (p=0.028 n=10+10) Memmove/8-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.545 n=9+10) Memmove/9-16 1.25ns ± 0% 1.35ns ± 0% +7.98% (p=0.000 n=10+10) Memmove/10-16 1.25ns ± 0% 1.35ns ± 0% +7.96% (p=0.000 n=9+9) Memmove/11-16 1.25ns ± 0% 1.35ns ± 0% +8.53% (p=0.000 n=10+9) Memmove/12-16 1.25ns ± 0% 1.35ns ± 1% +8.24% (p=0.000 n=10+10) Memmove/13-16 1.25ns ± 0% 1.34ns ± 0% +7.75% (p=0.000 n=10+10) Memmove/14-16 1.25ns ± 0% 1.35ns ± 1% +8.28% (p=0.000 n=10+9) Memmove/15-16 1.25ns ± 0% 1.35ns ± 0% +8.07% (p=0.000 n=10+9) Memmove/16-16 1.25ns ± 0% 1.35ns ± 1% +8.35% (p=0.000 n=9+10) Memmove/32-16 1.34ns ± 0% 1.36ns ± 1% +1.22% (p=0.000 n=10+10) Memmove/64-16 1.45ns ± 0% 1.64ns ± 0% +13.07% (p=0.000 n=10+9) Memmove/128-16 1.86ns ± 0% 2.02ns ± 0% +8.64% (p=0.000 n=10+10) Memmove/256-16 2.47ns ± 0% 2.49ns ± 1% +1.14% (p=0.000 n=10+10) Memmove/512-16 3.96ns ± 1% 3.96ns ± 0% ~ (p=0.182 n=10+10) Memmove/1024-16 5.90ns ± 1% 5.87ns ± 1% ~ (p=0.258 n=9+9) Memmove/2048-16 9.62ns ± 1% 9.62ns ± 2% ~ (p=0.963 n=8+9) Memmove/4096-16 16.4ns ± 0% 17.1ns ± 4% +4.19% (p=0.003 n=8+9) MemmoveOverlap/32-16 1.62ns ± 1% 1.68ns ± 1% +3.53% (p=0.000 n=10+10) MemmoveOverlap/64-16 1.64ns ± 0% 1.65ns ± 0% +0.29% (p=0.002 n=9+9) MemmoveOverlap/128-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.070 n=10+10) MemmoveOverlap/256-16 2.67ns ± 0% 2.67ns ± 0% +0.26% (p=0.012 n=10+10) MemmoveOverlap/512-16 6.20ns ±18% 5.74ns ± 0% ~ (p=0.645 n=10+8) MemmoveOverlap/1024-16 7.28ns ± 0% 7.30ns ± 0% +0.28% (p=0.006 n=8+10) MemmoveOverlap/2048-16 11.9ns ± 0% 12.0ns ± 1% +0.37% (p=0.014 n=9+9) MemmoveOverlap/4096-16 23.3ns ± 1% 23.1ns ± 1% -0.84% (p=0.000 n=8+10) MemmoveUnalignedDst/0-16 1.03ns ± 0% 1.03ns ± 0% +0.19% (p=0.007 n=10+10) MemmoveUnalignedDst/1-16 1.24ns ± 0% 1.25ns ± 1% +0.52% (p=0.022 n=10+10) MemmoveUnalignedDst/2-16 1.23ns ± 0% 1.23ns ± 0% ~ (p=0.051 n=10+10) MemmoveUnalignedDst/3-16 1.23ns ± 0% 1.23ns ± 0% +0.14% (p=0.006 n=9+9) MemmoveUnalignedDst/4-16 1.23ns ± 0% 1.24ns ± 1% +0.37% (p=0.004 n=10+10) MemmoveUnalignedDst/5-16 1.35ns ± 0% 1.35ns ± 0% ~ (p=0.075 n=10+10) MemmoveUnalignedDst/6-16 1.34ns ± 0% 1.34ns ± 0% ~ (p=0.779 n=10+10) MemmoveUnalignedDst/7-16 1.34ns ± 0% 1.34ns ± 0% ~ (p=1.000 n=10+10) MemmoveUnalignedDst/8-16 1.34ns ± 0% 1.35ns ± 1% +0.39% (p=0.024 n=10+10) MemmoveUnalignedDst/9-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.849 n=10+10) MemmoveUnalignedDst/10-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.255 n=10+10) MemmoveUnalignedDst/11-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.304 n=10+10) MemmoveUnalignedDst/12-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.672 n=10+10) MemmoveUnalignedDst/13-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.435 n=10+10) MemmoveUnalignedDst/14-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.340 n=10+10) MemmoveUnalignedDst/15-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.911 n=10+9) MemmoveUnalignedDst/16-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.074 n=10+10) MemmoveUnalignedDst/32-16 1.62ns ± 0% 1.63ns ± 0% ~ (p=0.059 n=10+10) MemmoveUnalignedDst/64-16 1.65ns ± 0% 1.65ns ± 0% ~ (p=0.234 n=10+10) MemmoveUnalignedDst/128-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.709 n=10+9) MemmoveUnalignedDst/256-16 3.69ns ± 0% 3.70ns ± 0% ~ (p=0.144 n=10+10) MemmoveUnalignedDst/512-16 4.15ns ± 1% 4.14ns ± 0% ~ (p=0.778 n=10+8) MemmoveUnalignedDst/1024-16 7.52ns ± 0% 7.53ns ± 1% ~ (p=0.650 n=9+9) MemmoveUnalignedDst/2048-16 12.9ns ± 0% 12.9ns ± 1% ~ (p=0.548 n=8+8) MemmoveUnalignedDst/4096-16 25.4ns ± 0% 25.4ns ± 0% ~ (p=0.947 n=9+9) MemmoveUnalignedDstOverlap/32-16 4.08ns ± 0% 4.09ns ± 0% ~ (p=0.360 n=10+10) MemmoveUnalignedDstOverlap/64-16 4.56ns ± 0% 4.56ns ± 0% ~ (p=0.705 n=10+9) MemmoveUnalignedDstOverlap/128-16 4.67ns ± 0% 4.67ns ± 0% ~ (p=0.397 n=10+10) MemmoveUnalignedDstOverlap/256-16 5.08ns ± 0% 5.08ns ± 0% ~ (p=0.159 n=10+9) MemmoveUnalignedDstOverlap/512-16 8.45ns ± 5% 8.19ns ± 0% -3.10% (p=0.021 n=10+9) MemmoveUnalignedDstOverlap/1024-16 9.55ns ± 0% 9.56ns ± 0% ~ (p=0.221 n=8+8) MemmoveUnalignedDstOverlap/2048-16 14.0ns ± 0% 14.0ns ± 1% ~ (p=0.200 n=10+9) MemmoveUnalignedDstOverlap/4096-16 26.5ns ± 0% 26.5ns ± 0% ~ (p=0.458 n=10+9) MemmoveUnalignedSrc/0-16 1.02ns ± 1% 0.99ns ± 1% -2.67% (p=0.000 n=10+9) MemmoveUnalignedSrc/1-16 1.13ns ± 0% 1.13ns ± 1% -0.25% (p=0.027 n=10+9) MemmoveUnalignedSrc/2-16 1.13ns ± 1% 1.13ns ± 0% -0.28% (p=0.012 n=10+9) MemmoveUnalignedSrc/3-16 1.24ns ± 1% 1.23ns ± 0% -0.25% (p=0.022 n=9+10) MemmoveUnalignedSrc/4-16 1.24ns ± 0% 1.23ns ± 1% ~ (p=0.118 n=9+10) MemmoveUnalignedSrc/5-16 1.34ns ± 0% 1.34ns ± 1% ~ (p=0.564 n=8+10) MemmoveUnalignedSrc/6-16 1.34ns ± 0% 1.34ns ± 0% -0.39% (p=0.000 n=10+10) MemmoveUnalignedSrc/7-16 1.34ns ± 0% 1.34ns ± 0% ~ (p=0.235 n=10+10) MemmoveUnalignedSrc/8-16 1.34ns ± 0% 1.34ns ± 0% -0.37% (p=0.002 n=10+9) MemmoveUnalignedSrc/9-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.579 n=10+9) MemmoveUnalignedSrc/10-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.534 n=10+9) MemmoveUnalignedSrc/11-16 1.44ns ± 0% 1.44ns ± 1% ~ (p=0.415 n=10+10) MemmoveUnalignedSrc/12-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.218 n=10+10) MemmoveUnalignedSrc/13-16 1.44ns ± 0% 1.44ns ± 1% ~ (p=0.693 n=10+10) MemmoveUnalignedSrc/14-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.901 n=10+10) MemmoveUnalignedSrc/15-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.379 n=10+10) MemmoveUnalignedSrc/16-16 1.44ns ± 1% 1.44ns ± 0% ~ (p=0.538 n=10+10) MemmoveUnalignedSrc/32-16 1.60ns ± 1% 1.60ns ± 0% ~ (p=0.491 n=10+10) MemmoveUnalignedSrc/64-16 1.65ns ± 0% 1.65ns ± 0% ~ (p=0.564 n=10+10) MemmoveUnalignedSrc/128-16 2.09ns ± 0% 2.09ns ± 0% ~ (p=0.497 n=10+9) MemmoveUnalignedSrc/256-16 2.70ns ± 0% 2.78ns ± 1% +2.81% (p=0.000 n=10+10) MemmoveUnalignedSrc/512-16 4.31ns ± 0% 4.30ns ± 0% -0.26% (p=0.031 n=8+9) MemmoveUnalignedSrc/1024-16 7.28ns ± 0% 7.21ns ± 1% -1.05% (p=0.000 n=8+10) MemmoveUnalignedSrc/2048-16 13.0ns ± 0% 13.0ns ± 0% ~ (p=0.180 n=9+8) MemmoveUnalignedSrc/4096-16 25.4ns ± 0% 25.3ns ± 1% ~ (p=0.054 n=10+10) MemmoveUnalignedSrcOverlap/32-16 4.04ns ± 0% 4.06ns ± 0% +0.62% (p=0.000 n=9+10) MemmoveUnalignedSrcOverlap/64-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.421 n=10+10) MemmoveUnalignedSrcOverlap/128-16 4.53ns ± 0% 4.52ns ± 0% ~ (p=0.251 n=10+10) MemmoveUnalignedSrcOverlap/256-16 6.17ns ± 0% 6.15ns ± 0% -0.35% (p=0.000 n=10+9) MemmoveUnalignedSrcOverlap/512-16 7.43ns ± 0% 7.44ns ± 0% ~ (p=0.524 n=9+8) MemmoveUnalignedSrcOverlap/1024-16 8.94ns ± 0% 8.94ns ± 0% ~ (p=0.419 n=8+8) MemmoveUnalignedSrcOverlap/2048-16 13.2ns ± 0% 14.5ns ±21% ~ (p=0.107 n=8+10) MemmoveUnalignedSrcOverlap/4096-16 25.6ns ± 0% 25.6ns ± 1% ~ (p=0.650 n=9+9) Memclr/5-16 0.86ns ± 1% 0.86ns ± 2% ~ (p=0.531 n=9+9) Memclr/16-16 1.04ns ± 0% 1.04ns ± 0% +0.32% (p=0.013 n=9+10) Memclr/64-16 1.23ns ± 0% 1.26ns ± 0% +2.28% (p=0.000 n=10+10) Memclr/256-16 2.27ns ± 0% 2.27ns ± 0% ~ (p=0.127 n=10+10) Memclr/4096-16 17.1ns ± 1% 17.3ns ± 0% +0.88% (p=0.000 n=10+10) Memclr/65536-16 821ns ± 0% 822ns ± 0% ~ (p=0.516 n=10+10) Memclr/1M-16 14.1µs ± 1% 14.0µs ± 1% ~ (p=0.516 n=10+10) Memclr/4M-16 86.1µs ± 1% 85.9µs ± 0% ~ (p=0.123 n=10+10) Memclr/8M-16 174µs ± 2% 173µs ± 0% ~ (p=0.408 n=10+8) Memclr/16M-16 385µs ± 4% 387µs ± 0% ~ (p=0.173 n=10+8) Memclr/64M-16 2.18ms ± 0% 2.19ms ± 0% ~ (p=0.113 n=10+9) GoMemclr/5-16 0.82ns ± 0% 0.82ns ± 0% ~ (p=0.346 n=9+10) GoMemclr/16-16 1.02ns ± 0% 1.02ns ± 0% +0.22% (p=0.003 n=10+8) GoMemclr/64-16 1.14ns ± 0% 1.14ns ± 0% ~ (p=0.948 n=10+9) GoMemclr/256-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.868 n=10+10) MemclrRange/1K_2K-16 457ns ± 0% 428ns ± 1% -6.38% (p=0.000 n=10+10) MemclrRange/2K_8K-16 1.46µs ± 0% 1.46µs ± 0% ~ (p=0.700 n=10+10) MemclrRange/4K_16K-16 1.16µs ± 0% 1.16µs ± 0% ~ (p=0.567 n=9+10) MemclrRange/160K_228K-16 20.7µs ± 0% 20.7µs ± 0% ~ (p=0.160 n=10+10) ClearFat7-16 0.38ns ± 5% 0.21ns ± 1% -45.79% (p=0.000 n=9+10) ClearFat8-16 0.21ns ± 3% 0.12ns ± 2% -44.16% (p=0.000 n=8+9) ClearFat11-16 0.35ns ± 3% 0.21ns ± 1% -40.46% (p=0.000 n=9+9) ClearFat12-16 0.23ns ± 9% 0.21ns ± 1% -10.23% (p=0.000 n=10+9) ClearFat13-16 0.22ns ± 6% 0.21ns ± 2% -6.53% (p=0.000 n=10+10) ClearFat14-16 0.22ns ± 4% 0.21ns ± 1% -5.97% (p=0.000 n=10+10) ClearFat15-16 0.22ns ± 4% 0.21ns ± 1% -6.96% (p=0.000 n=10+9) ClearFat16-16 0.19ns ± 9% 0.12ns ± 6% -34.89% (p=0.000 n=9+10) ClearFat24-16 0.23ns ± 6% 0.21ns ± 1% -10.26% (p=0.000 n=10+9) ClearFat32-16 0.22ns ± 5% 0.21ns ± 2% -5.31% (p=0.000 n=10+10) ClearFat40-16 0.34ns ± 4% 0.62ns ± 1% +83.00% (p=0.000 n=10+10) ClearFat48-16 0.33ns ± 2% 0.41ns ± 0% +26.71% (p=0.000 n=10+10) ClearFat56-16 0.41ns ± 1% 0.41ns ± 0% ~ (p=0.838 n=10+10) ClearFat64-16 0.41ns ± 0% 0.41ns ± 0% ~ (p=0.178 n=10+8) ClearFat72-16 0.82ns ± 0% 0.82ns ± 0% ~ (p=0.669 n=10+10) ClearFat128-16 1.04ns ± 0% 1.04ns ± 0% ~ (p=0.679 n=10+10) ClearFat256-16 1.86ns ± 0% 1.86ns ± 0% ~ (p=0.066 n=9+10) ClearFat512-16 3.50ns ± 0% 3.50ns ± 0% ~ (p=0.626 n=10+10) ClearFat1024-16 6.79ns ± 0% 6.79ns ± 0% ~ (p=0.986 n=10+10) ClearFat1032-16 13.6ns ± 0% 13.6ns ± 0% +0.13% (p=0.044 n=10+10) ClearFat1040-16 10.3ns ± 0% 10.3ns ± 0% ~ (p=0.175 n=10+9) CopyFat7-16 0.37ns ±13% 0.25ns ± 1% -31.74% (p=0.000 n=10+9) CopyFat8-16 0.17ns ± 1% 0.17ns ± 2% +1.35% (p=0.004 n=9+9) CopyFat11-16 0.26ns ± 1% 0.30ns ± 3% +12.58% (p=0.000 n=9+10) CopyFat12-16 0.28ns ± 2% 0.26ns ± 1% -5.66% (p=0.000 n=9+9) CopyFat13-16 0.26ns ± 0% 0.28ns ± 4% +7.35% (p=0.000 n=8+10) CopyFat14-16 0.29ns ± 6% 0.26ns ± 2% -10.46% (p=0.000 n=10+9) CopyFat15-16 0.26ns ± 1% 0.30ns ± 6% +14.12% (p=0.000 n=8+10) CopyFat16-16 0.21ns ± 1% 0.21ns ± 0% ~ (p=0.426 n=8+8) CopyFat24-16 0.29ns ± 3% 0.25ns ± 1% -12.27% (p=0.000 n=9+10) CopyFat32-16 0.26ns ± 4% 0.29ns ± 4% +11.71% (p=0.000 n=10+10) CopyFat64-16 0.46ns ± 8% 0.42ns ± 1% -8.37% (p=0.002 n=10+10) CopyFat72-16 0.82ns ± 0% 0.82ns ± 0% ~ (p=0.563 n=10+10) CopyFat128-16 1.53ns ± 0% 1.54ns ± 0% +0.62% (p=0.000 n=10+10) CopyFat256-16 2.68ns ± 0% 2.65ns ± 1% -1.23% (p=0.000 n=10+10) CopyFat512-16 4.93ns ± 1% 5.19ns ± 3% +5.16% (p=0.000 n=9+9) CopyFat520-16 6.99ns ± 0% 6.99ns ± 0% ~ (p=0.539 n=10+10) CopyFat1024-16 11.5ns ± 1% 9.8ns ± 1% -14.98% (p=0.000 n=9+10) CopyFat1032-16 13.6ns ± 0% 13.6ns ± 0% ~ (p=0.728 n=10+10) CopyFat1040-16 11.0ns ± 0% 11.1ns ± 0% +0.53% (p=0.000 n=10+10) Issue18740/2byte-16 10.1µs ± 0% 10.1µs ± 0% ~ (p=0.342 n=10+10) Issue18740/4byte-16 2.34µs ± 0% 2.35µs ± 0% +0.30% (p=0.002 n=10+8) Issue18740/8byte-16 1.28µs ± 0% 1.28µs ± 0% +0.32% (p=0.000 n=9+10) Finalizer-16 345µs ± 1% 336µs ± 0% -2.55% (p=0.000 n=10+9) FinalizerRun-16 450ns ± 3% 420ns ± 1% -6.65% (p=0.000 n=10+10) PallocBitsSummarize/Unpacked00-16 2.88ns ± 0% 2.88ns ± 0% ~ (p=0.358 n=10+10) PallocBitsSummarize/UnpackedFFFFFFFFFFFFFFFF-16 15.2ns ± 0% 15.2ns ± 0% ~ (p=0.925 n=10+10) PallocBitsSummarize/UnpackedAA-16 16.4ns ± 0% 16.3ns ± 0% ~ (p=0.113 n=9+9) PallocBitsSummarize/UnpackedAAAAAAAAAAAAAAAA-16 16.5ns ± 0% 16.6ns ± 0% ~ (p=0.238 n=10+10) PallocBitsSummarize/Unpacked80000000AAAAAAAA-16 37.8ns ± 1% 36.4ns ± 0% -3.70% (p=0.000 n=10+9) PallocBitsSummarize/UnpackedAAAAAAAA00000001-16 41.8ns ± 1% 39.9ns ± 0% -4.68% (p=0.000 n=9+10) PallocBitsSummarize/UnpackedBBBBBBBBBBBBBBBB-16 18.3ns ± 0% 18.3ns ± 0% ~ (p=0.781 n=10+10) PallocBitsSummarize/Unpacked80000000BBBBBBBB-16 38.8ns ± 1% 38.1ns ± 0% -1.78% (p=0.000 n=9+10) PallocBitsSummarize/UnpackedBBBBBBBB00000001-16 37.5ns ± 0% 36.1ns ± 1% -3.88% (p=0.000 n=8+10) PallocBitsSummarize/UnpackedCCCCCCCCCCCCCCCC-16 21.8ns ± 0% 21.9ns ± 0% +0.20% (p=0.018 n=10+9) PallocBitsSummarize/Unpacked4444444444444444-16 21.8ns ± 0% 21.9ns ± 0% +0.20% (p=0.029 n=10+9) PallocBitsSummarize/Unpacked4040404040404040-16 26.5ns ± 0% 26.5ns ± 0% -0.24% (p=0.001 n=9+10) PallocBitsSummarize/Unpacked4000400040004000-16 33.4ns ± 1% 31.3ns ± 0% -6.20% (p=0.000 n=9+10) PallocBitsSummarize/Unpacked1000404044CCAAFF-16 36.4ns ± 1% 35.9ns ± 0% -1.50% (p=0.000 n=10+10) FindBitRange64/Pattern00Size2-16 0.34ns ± 1% 0.35ns ± 1% +3.80% (p=0.000 n=10+9) FindBitRange64/Pattern00Size8-16 0.70ns ± 1% 0.70ns ± 0% -0.68% (p=0.000 n=10+10) FindBitRange64/Pattern00Size32-16 0.70ns ± 1% 0.69ns ± 0% -0.86% (p=0.001 n=10+8) FindBitRange64/PatternFFFFFFFFFFFFFFFFSize2-16 0.34ns ± 1% 0.35ns ± 1% +4.45% (p=0.000 n=9+8) FindBitRange64/PatternFFFFFFFFFFFFFFFFSize8-16 1.54ns ± 0% 1.54ns ± 1% ~ (p=0.914 n=9+9) FindBitRange64/PatternFFFFFFFFFFFFFFFFSize32-16 2.78ns ± 0% 2.78ns ± 0% ~ (p=0.295 n=9+10) FindBitRange64/PatternAASize2-16 0.34ns ± 2% 0.35ns ± 2% +4.61% (p=0.000 n=10+10) FindBitRange64/PatternAASize8-16 0.70ns ± 1% 0.70ns ± 1% -0.82% (p=0.005 n=10+10) FindBitRange64/PatternAASize32-16 0.70ns ± 1% 0.70ns ± 0% -0.73% (p=0.003 n=10+9) FindBitRange64/PatternAAAAAAAAAAAAAAAASize2-16 0.34ns ± 2% 0.35ns ± 2% +3.94% (p=0.000 n=10+10) FindBitRange64/PatternAAAAAAAAAAAAAAAASize8-16 0.70ns ± 1% 0.70ns ± 1% -0.67% (p=0.025 n=10+10) FindBitRange64/PatternAAAAAAAAAAAAAAAASize32-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.118 n=9+10) FindBitRange64/Pattern80000000AAAAAAAASize2-16 0.34ns ± 1% 0.35ns ± 2% +3.72% (p=0.000 n=10+9) FindBitRange64/Pattern80000000AAAAAAAASize8-16 0.70ns ± 1% 0.70ns ± 0% ~ (p=0.102 n=10+10) FindBitRange64/Pattern80000000AAAAAAAASize32-16 0.70ns ± 1% 0.70ns ± 1% -0.55% (p=0.011 n=10+10) FindBitRange64/PatternAAAAAAAA00000001Size2-16 0.34ns ± 2% 0.35ns ± 1% +3.83% (p=0.000 n=10+9) FindBitRange64/PatternAAAAAAAA00000001Size8-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.065 n=10+10) FindBitRange64/PatternAAAAAAAA00000001Size32-16 0.70ns ± 1% 0.70ns ± 1% -0.95% (p=0.002 n=10+10) FindBitRange64/PatternBBBBBBBBBBBBBBBBSize2-16 0.34ns ± 0% 0.35ns ± 1% +4.12% (p=0.000 n=8+10) FindBitRange64/PatternBBBBBBBBBBBBBBBBSize8-16 1.24ns ± 0% 1.23ns ± 0% -0.30% (p=0.002 n=10+9) FindBitRange64/PatternBBBBBBBBBBBBBBBBSize32-16 1.24ns ± 0% 1.24ns ± 0% -0.17% (p=0.023 n=9+10) FindBitRange64/Pattern80000000BBBBBBBBSize2-16 0.34ns ± 1% 0.35ns ± 2% +4.82% (p=0.000 n=9+10) FindBitRange64/Pattern80000000BBBBBBBBSize8-16 1.24ns ± 1% 1.24ns ± 0% ~ (p=0.063 n=10+10) FindBitRange64/Pattern80000000BBBBBBBBSize32-16 1.24ns ± 0% 1.24ns ± 0% ~ (p=0.164 n=9+10) FindBitRange64/PatternBBBBBBBB00000001Size2-16 0.34ns ± 1% 0.35ns ± 1% +4.38% (p=0.000 n=8+10) FindBitRange64/PatternBBBBBBBB00000001Size8-16 1.24ns ± 1% 1.24ns ± 0% ~ (p=0.052 n=10+10) FindBitRange64/PatternBBBBBBBB00000001Size32-16 1.24ns ± 0% 1.23ns ± 0% -0.40% (p=0.000 n=10+10) FindBitRange64/PatternCCCCCCCCCCCCCCCCSize2-16 0.34ns ± 0% 0.35ns ± 2% +3.96% (p=0.000 n=9+10) FindBitRange64/PatternCCCCCCCCCCCCCCCCSize8-16 1.24ns ± 0% 1.23ns ± 0% -0.30% (p=0.000 n=10+9) FindBitRange64/PatternCCCCCCCCCCCCCCCCSize32-16 1.24ns ± 0% 1.24ns ± 1% ~ (p=0.284 n=10+10) FindBitRange64/Pattern4444444444444444Size2-16 0.34ns ± 1% 0.35ns ± 1% +3.91% (p=0.000 n=9+9) FindBitRange64/Pattern4444444444444444Size8-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.617 n=10+10) FindBitRange64/Pattern4444444444444444Size32-16 0.70ns ± 1% 0.70ns ± 1% -0.60% (p=0.006 n=10+10) FindBitRange64/Pattern4040404040404040Size2-16 0.34ns ± 2% 0.35ns ± 2% +3.67% (p=0.000 n=10+10) FindBitRange64/Pattern4040404040404040Size8-16 0.70ns ± 2% 0.70ns ± 1% -0.87% (p=0.014 n=10+10) FindBitRange64/Pattern4040404040404040Size32-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.256 n=10+10) FindBitRange64/Pattern4000400040004000Size2-16 0.34ns ± 2% 0.35ns ± 3% +4.71% (p=0.000 n=10+10) FindBitRange64/Pattern4000400040004000Size8-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.393 n=10+10) FindBitRange64/Pattern4000400040004000Size32-16 0.70ns ± 1% 0.70ns ± 1% -0.86% (p=0.014 n=10+10) NetpollBreak-16 1.49µs ± 1% 1.50µs ± 3% ~ (p=0.181 n=8+10) Syscall-16 3.68ns ± 1% 3.66ns ± 2% ~ (p=0.148 n=10+10) SyscallWork-16 5.15ns ± 1% 5.13ns ± 0% ~ (p=0.188 n=10+9) SyscallExcess-16 3.89ns ± 2% 3.83ns ± 1% -1.52% (p=0.001 n=10+10) SyscallExcessWork-16 5.34ns ± 1% 5.31ns ± 0% -0.64% (p=0.000 n=10+9) PingPongHog-16 397ns ± 7% 394ns ±11% ~ (p=0.912 n=10+10) StackGrowth-16 67.9ns ± 0% 68.8ns ± 0% +1.28% (p=0.000 n=10+8) StackGrowthDeep-16 7.70µs ± 1% 8.48µs ± 2% +10.06% (p=0.000 n=9+10) CreateGoroutines-16 124ns ± 1% 124ns ± 1% ~ (p=0.254 n=10+10) CreateGoroutinesParallel-16 25.7ns ± 1% 27.6ns ± 2% +7.51% (p=0.000 n=10+10) CreateGoroutinesCapture-16 823ns ± 1% 821ns ± 2% ~ (p=0.699 n=10+10) CreateGoroutinesSingle-16 175ns ± 3% 172ns ± 3% -1.90% (p=0.011 n=10+10) ClosureCall-16 0.11ns ± 7% 0.12ns ± 3% ~ (p=0.842 n=9+10) WakeupParallelSpinning/0s-16 11.4µs ± 0% 11.4µs ± 0% ~ (p=0.325 n=9+10) WakeupParallelSpinning/1µs-16 15.4µs ± 0% 15.4µs ± 1% ~ (p=0.955 n=10+10) WakeupParallelSpinning/2µs-16 18.7µs ± 2% 18.9µs ± 2% ~ (p=0.052 n=10+10) WakeupParallelSpinning/5µs-16 30.7µs ± 0% 30.7µs ± 0% -0.03% (p=0.003 n=10+10) WakeupParallelSpinning/10µs-16 48.8µs ± 0% 48.8µs ± 0% ~ (p=0.670 n=10+10) WakeupParallelSpinning/20µs-16 90.8µs ± 0% 90.8µs ± 0% -0.02% (p=0.004 n=10+10) WakeupParallelSpinning/50µs-16 211µs ± 0% 211µs ± 0% ~ (p=0.194 n=10+10) WakeupParallelSpinning/100µs-16 323µs ± 0% 323µs ± 0% ~ (p=1.000 n=10+9) WakeupParallelSyscall/0s-16 118µs ± 0% 118µs ± 0% ~ (p=0.447 n=10+9) WakeupParallelSyscall/1µs-16 119µs ± 2% 119µs ± 1% ~ (p=0.604 n=10+9) WakeupParallelSyscall/2µs-16 120µs ± 1% 121µs ± 3% ~ (p=0.263 n=8+10) WakeupParallelSyscall/5µs-16 126µs ± 2% 126µs ± 2% ~ (p=0.510 n=10+9) WakeupParallelSyscall/10µs-16 136µs ± 1% 137µs ± 1% ~ (p=0.095 n=9+10) WakeupParallelSyscall/20µs-16 156µs ± 2% 157µs ± 3% ~ (p=0.604 n=10+9) WakeupParallelSyscall/50µs-16 221µs ± 1% 220µs ± 1% ~ (p=0.063 n=10+10) WakeupParallelSyscall/100µs-16 326µs ± 0% 325µs ± 0% -0.26% (p=0.003 n=9+10) Matmult-16 0.67ns ± 2% 0.66ns ± 2% ~ (p=0.256 n=10+10) Fastrand-16 0.08ns ±11% 0.08ns ±13% ~ (p=0.661 n=9+10) Fastrand64-16 0.08ns ±11% 0.08ns ± 6% ~ (p=0.631 n=10+10) FastrandHashiter-16 1.76ns ± 1% 1.76ns ± 1% ~ (p=0.854 n=8+8) Fastrandn/2-16 0.86ns ± 1% 0.86ns ± 1% +1.09% (p=0.000 n=10+9) Fastrandn/3-16 0.85ns ± 1% 0.86ns ± 1% +1.23% (p=0.001 n=10+10) Fastrandn/4-16 0.85ns ± 1% 0.87ns ± 2% +1.60% (p=0.000 n=10+10) Fastrandn/5-16 0.85ns ± 1% 0.86ns ± 1% +1.05% (p=0.000 n=10+10) IfaceCmp100-16 46.6ns ± 0% 46.1ns ± 0% -1.18% (p=0.000 n=10+10) IfaceCmpNil100-16 26.8ns ± 0% 26.8ns ± 0% ~ (p=0.777 n=10+8) EfaceCmpDiff-16 132ns ± 0% 130ns ± 0% -0.95% (p=0.000 n=10+9) EfaceCmpDiffIndirect-16 209ns ± 0% 211ns ± 0% +1.14% (p=0.000 n=10+9) Defer-16 3.40ns ± 1% 3.04ns ± 0% -10.67% (p=0.000 n=10+10) Defer10-16 29.4ns ± 2% 27.2ns ± 3% -7.26% (p=0.000 n=10+10) DeferMany-16 110ns ± 6% 113ns ± 2% +3.45% (p=0.017 n=9+9) PanicRecover-16 67.6ns ± 0% 67.7ns ± 2% ~ (p=0.436 n=9+9) GoroutineProfile/small-nil/idle-16 3.90µs ± 4% 3.86µs ± 2% ~ (p=0.305 n=10+9) GoroutineProfile/small-nil/loaded-16 4.82µs ± 6% 4.82µs ± 4% ~ (p=0.905 n=10+9) GoroutineProfile/small/idle-16 103µs ± 3% 102µs ± 3% ~ (p=0.113 n=9+9) GoroutineProfile/small/loaded-16 432µs ± 5% 440µs ±13% ~ (p=0.604 n=9+10) GoroutineProfile/large-nil/idle-16 3.86µs ± 3% 3.82µs ± 3% ~ (p=0.210 n=10+10) GoroutineProfile/large-nil/loaded-16 4.90µs ± 2% 4.90µs ± 5% ~ (p=0.780 n=10+9) GoroutineProfile/large/idle-16 2.58ms ± 1% 2.52ms ± 1% -2.38% (p=0.000 n=10+10) GoroutineProfile/large/loaded-16 8.62ms ± 9% 8.90ms ±11% ~ (p=0.400 n=9+10) GoroutineProfile/sparse-nil/idle-16 3.85µs ± 4% 3.81µs ± 3% ~ (p=0.470 n=10+10) GoroutineProfile/sparse-nil/loaded-16 4.82µs ± 4% 4.69µs ± 5% ~ (p=0.052 n=10+10) GoroutineProfile/sparse/idle-16 102µs ± 4% 102µs ± 2% ~ (p=0.497 n=10+9) GoroutineProfile/sparse/loaded-16 438µs ± 7% 437µs ± 6% ~ (p=0.796 n=10+10) RWMutexUncontended-16 6.79ns ± 0% 6.78ns ± 0% ~ (p=0.228 n=10+8) RWMutexWrite100-16 85.4ns ± 0% 87.1ns ± 0% +2.00% (p=0.000 n=10+8) RWMutexWrite10-16 168ns ±25% 152ns ±11% ~ (p=0.063 n=10+10) RWMutexWorkWrite100-16 106ns ± 0% 106ns ± 3% ~ (p=0.136 n=10+10) RWMutexWorkWrite10-16 567ns ± 3% 571ns ± 1% ~ (p=0.326 n=10+9) SemTable/OneAddrCollision/n=1000-16 15.9µs ± 1% 16.0µs ± 1% +0.50% (p=0.031 n=9+9) SemTable/ManyAddrCollision/n=1000-16 56.2µs ± 1% 56.8µs ± 1% +1.06% (p=0.000 n=10+10) SemTable/OneAddrCollision/n=2000-16 32.6µs ± 2% 32.9µs ± 4% ~ (p=0.156 n=10+9) SemTable/ManyAddrCollision/n=2000-16 118µs ± 0% 119µs ± 0% +0.75% (p=0.000 n=9+10) SemTable/OneAddrCollision/n=4000-16 65.3µs ± 1% 65.6µs ± 3% ~ (p=0.497 n=9+10) SemTable/ManyAddrCollision/n=4000-16 245µs ± 0% 248µs ± 2% +1.36% (p=0.000 n=9+10) SemTable/OneAddrCollision/n=8000-16 131µs ± 1% 130µs ± 1% -1.01% (p=0.002 n=9+10) SemTable/ManyAddrCollision/n=8000-16 503µs ± 1% 508µs ± 0% +0.97% (p=0.000 n=10+10) MakeSliceCopy/mallocmove/Byte-16 67.6ns ± 1% 64.1ns ± 2% -5.20% (p=0.000 n=10+10) MakeSliceCopy/mallocmove/Int-16 65.0ns ± 7% 61.7ns ± 4% -5.08% (p=0.009 n=10+10) MakeSliceCopy/mallocmove/Ptr-16 88.1ns ± 1% 79.9ns ± 1% -9.29% (p=0.000 n=10+10) MakeSliceCopy/makecopy/Byte-16 65.2ns ± 6% 63.4ns ± 0% ~ (p=0.500 n=10+8) MakeSliceCopy/makecopy/Int-16 63.2ns ± 1% 64.1ns ± 1% +1.34% (p=0.001 n=9+9) MakeSliceCopy/makecopy/Ptr-16 88.1ns ± 1% 80.1ns ± 1% -9.09% (p=0.000 n=10+10) MakeSliceCopy/nilappend/Byte-16 69.8ns ± 1% 65.7ns ± 3% -5.80% (p=0.000 n=10+10) MakeSliceCopy/nilappend/Int-16 69.6ns ± 2% 67.2ns ± 1% -3.50% (p=0.000 n=10+9) MakeSliceCopy/nilappend/Ptr-16 91.5ns ± 1% 83.8ns ± 1% -8.42% (p=0.000 n=9+10) MakeSlice/Byte-16 6.64ns ± 3% 6.58ns ± 2% ~ (p=0.393 n=10+10) MakeSlice/Int16-16 8.60ns ± 1% 8.38ns ± 3% -2.48% (p=0.001 n=9+10) MakeSlice/Int-16 17.7ns ± 3% 16.9ns ± 1% -4.67% (p=0.000 n=10+9) MakeSlice/Ptr-16 24.0ns ± 3% 23.3ns ± 2% -3.25% (p=0.000 n=10+9) MakeSlice/Struct/24-16 34.1ns ± 1% 32.0ns ± 1% -6.11% (p=0.000 n=10+10) MakeSlice/Struct/32-16 39.1ns ± 4% 38.2ns ± 1% ~ (p=0.829 n=10+8) MakeSlice/Struct/40-16 47.0ns ± 5% 43.0ns ± 2% -8.55% (p=0.000 n=10+9) GrowSlice/Byte-16 15.3ns ± 3% 15.0ns ± 2% -1.75% (p=0.005 n=9+9) GrowSlice/Int16-16 18.9ns ± 2% 18.4ns ± 2% -2.71% (p=0.000 n=10+9) GrowSlice/Int-16 33.9ns ± 1% 32.2ns ± 1% -4.89% (p=0.000 n=10+9) GrowSlice/Ptr-16 45.3ns ± 2% 43.5ns ± 1% -4.12% (p=0.000 n=10+10) GrowSlice/Struct/24-16 61.9ns ± 2% 60.0ns ± 4% -3.10% (p=0.002 n=10+10) GrowSlice/Struct/32-16 79.9ns ± 2% 72.3ns ± 3% -9.58% (p=0.000 n=8+10) GrowSlice/Struct/40-16 97.1ns ± 7% 88.8ns ± 5% -8.49% (p=0.000 n=10+10) ExtendSlice/IntSlice-16 21.1ns ± 2% 20.3ns ± 2% -3.71% (p=0.000 n=10+10) ExtendSlice/PointerSlice-16 26.8ns ± 2% 26.3ns ± 2% -1.86% (p=0.004 n=10+10) ExtendSlice/NoGrow-16 1.23ns ± 0% 1.30ns ± 1% +5.03% (p=0.000 n=10+10) Append-16 4.58ns ± 1% 4.53ns ± 0% -1.11% (p=0.000 n=10+10) AppendGrowByte-16 1.46ms ± 8% 1.42ms ± 7% -3.24% (p=0.035 n=10+10) AppendGrowString-16 27.8ms ± 4% 27.2ms ± 5% ~ (p=0.052 n=10+10) AppendSlice/1Bytes-16 1.03ns ± 1% 1.04ns ± 1% ~ (p=0.303 n=10+10) AppendSlice/4Bytes-16 1.04ns ± 0% 1.05ns ± 0% +0.79% (p=0.000 n=9+10) AppendSlice/7Bytes-16 1.23ns ± 1% 1.24ns ± 0% +0.45% (p=0.001 n=10+10) AppendSlice/8Bytes-16 1.24ns ± 0% 1.24ns ± 0% ~ (p=0.183 n=10+10) AppendSlice/15Bytes-16 1.37ns ± 1% 1.43ns ± 1% +3.88% (p=0.000 n=10+10) AppendSlice/16Bytes-16 1.37ns ± 1% 1.42ns ± 1% +3.63% (p=0.000 n=9+10) AppendSlice/32Bytes-16 1.44ns ± 0% 1.47ns ± 1% +1.83% (p=0.000 n=10+10) AppendSliceLarge/1024Bytes-16 257ns ± 2% 234ns ± 1% -8.96% (p=0.000 n=8+9) AppendSliceLarge/4096Bytes-16 871ns ± 6% 812ns ± 1% -6.80% (p=0.000 n=10+10) AppendSliceLarge/16384Bytes-16 3.15µs ± 6% 3.04µs ± 5% ~ (p=0.052 n=10+10) AppendSliceLarge/65536Bytes-16 10.7µs ± 7% 10.8µs ± 2% ~ (p=0.278 n=10+9) AppendSliceLarge/262144Bytes-16 42.9µs ± 1% 39.6µs ± 5% -7.75% (p=0.000 n=9+10) AppendSliceLarge/1048576Bytes-16 147µs ± 4% 144µs ± 4% -2.21% (p=0.035 n=10+10) AppendStr/1Bytes-16 1.20ns ± 0% 1.20ns ± 0% ~ (p=0.755 n=10+10) AppendStr/4Bytes-16 1.13ns ± 0% 1.14ns ± 1% +1.20% (p=0.000 n=10+10) AppendStr/8Bytes-16 1.24ns ± 0% 1.25ns ± 0% +0.93% (p=0.000 n=10+10) AppendStr/16Bytes-16 1.40ns ± 0% 1.42ns ± 0% +2.10% (p=0.000 n=9+10) AppendStr/32Bytes-16 1.44ns ± 0% 1.45ns ± 0% +0.99% (p=0.000 n=10+10) AppendSpecialCase-16 8.64ns ± 1% 8.89ns ± 2% +2.90% (p=0.000 n=10+10) Copy/1Byte-16 1.24ns ± 1% 1.24ns ± 0% -0.28% (p=0.000 n=10+6) Copy/1String-16 1.24ns ± 0% 1.23ns ± 0% ~ (p=0.160 n=10+10) Copy/2Byte-16 1.24ns ± 0% 1.24ns ± 0% ~ (p=0.115 n=10+10) Copy/2String-16 1.24ns ± 0% 1.24ns ± 1% ~ (p=0.954 n=10+10) Copy/4Byte-16 1.24ns ± 0% 1.24ns ± 0% -0.44% (p=0.001 n=10+10) Copy/4String-16 1.23ns ± 0% 1.23ns ± 0% ~ (p=0.081 n=10+10) Copy/8Byte-16 1.37ns ± 0% 1.34ns ± 0% -1.79% (p=0.000 n=9+9) Copy/8String-16 1.34ns ± 0% 1.34ns ± 0% -0.58% (p=0.000 n=9+10) Copy/12Byte-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.149 n=9+9) Copy/12String-16 1.44ns ± 0% 1.45ns ± 0% ~ (p=0.124 n=9+9) Copy/16Byte-16 1.44ns ± 0% 1.44ns ± 0% -0.19% (p=0.004 n=10+9) Copy/16String-16 1.44ns ± 0% 1.45ns ± 0% +0.30% (p=0.008 n=10+10) Copy/32Byte-16 1.63ns ± 1% 1.62ns ± 1% -0.72% (p=0.002 n=10+10) Copy/32String-16 1.60ns ± 1% 1.64ns ± 0% +2.23% (p=0.000 n=10+10) Copy/128Byte-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.757 n=9+10) Copy/128String-16 2.07ns ± 0% 2.07ns ± 0% +0.36% (p=0.004 n=10+10) Copy/1024Byte-16 6.07ns ± 2% 6.00ns ± 1% -1.20% (p=0.000 n=9+10) Copy/1024String-16 6.05ns ± 0% 5.95ns ± 1% -1.54% (p=0.000 n=10+9) AppendInPlace/NoGrow/Byte-16 288ns ± 1% 284ns ± 1% -1.58% (p=0.000 n=10+10) AppendInPlace/NoGrow/1Ptr-16 844ns ± 1% 809ns ± 3% -4.13% (p=0.000 n=9+10) AppendInPlace/NoGrow/2Ptr-16 1.47µs ± 1% 1.46µs ± 1% ~ (p=0.388 n=9+10) AppendInPlace/NoGrow/3Ptr-16 1.87µs ± 7% 1.91µs ± 1% ~ (p=0.166 n=10+8) AppendInPlace/NoGrow/4Ptr-16 2.66µs ± 1% 2.67µs ± 3% ~ (p=0.968 n=9+10) AppendInPlace/Grow/Byte-16 126ns ± 2% 121ns ± 2% -4.06% (p=0.000 n=10+10) AppendInPlace/Grow/1Ptr-16 132ns ± 2% 127ns ± 2% -4.28% (p=0.000 n=10+9) AppendInPlace/Grow/2Ptr-16 196ns ± 2% 188ns ± 1% -4.20% (p=0.000 n=10+8) AppendInPlace/Grow/3Ptr-16 264ns ± 1% 260ns ± 1% -1.51% (p=0.000 n=9+10) AppendInPlace/Grow/4Ptr-16 297ns ± 2% 294ns ± 2% ~ (p=0.085 n=10+10) StackCopyPtr-16 36.4ms ± 2% 36.7ms ± 2% ~ (p=0.481 n=10+10) StackCopy-16 33.9ms ± 3% 32.6ms ± 1% -3.87% (p=0.000 n=10+8) StackCopyNoCache-16 1.00ms ± 5% 1.01ms ± 5% ~ (p=0.143 n=10+10) StackCopyWithStkobj-16 11.0ms ± 3% 10.9ms ± 4% ~ (p=0.579 n=10+10) Issue18138-16 49.2µs ± 5% 49.0µs ± 4% ~ (p=1.000 n=10+9) CompareStringEqual-16 1.39ns ± 1% 1.45ns ± 2% +3.80% (p=0.000 n=8+10) CompareStringIdentical-16 0.55ns ± 1% 0.55ns ± 0% +0.42% (p=0.007 n=10+10) CompareStringSameLength-16 1.03ns ± 0% 1.03ns ± 0% ~ (p=0.430 n=9+10) CompareStringDifferentLength-16 0.11ns ± 2% 0.11ns ± 3% ~ (p=0.139 n=9+10) CompareStringBigUnaligned-16 23.9µs ± 1% 24.0µs ± 1% ~ (p=0.370 n=9+8) CompareStringBig-16 22.0µs ± 3% 22.2µs ± 3% ~ (p=0.243 n=9+10) ConcatStringAndBytes-16 10.7ns ± 1% 10.0ns ± 2% -6.33% (p=0.000 n=10+10) SliceByteToString/1-16 1.34ns ± 0% 1.34ns ± 0% ~ (p=0.057 n=10+10) SliceByteToString/2-16 6.67ns ± 2% 6.60ns ± 3% ~ (p=0.101 n=10+10) SliceByteToString/4-16 7.76ns ± 2% 7.56ns ± 3% -2.59% (p=0.001 n=10+10) SliceByteToString/8-16 9.81ns ± 4% 9.57ns ± 2% -2.48% (p=0.005 n=10+10) SliceByteToString/16-16 14.0ns ± 3% 13.7ns ± 2% -2.31% (p=0.009 n=10+10) SliceByteToString/32-16 17.3ns ± 1% 16.7ns ± 2% -3.41% (p=0.000 n=10+10) SliceByteToString/64-16 25.1ns ± 1% 24.1ns ± 2% -3.93% (p=0.000 n=9+10) SliceByteToString/128-16 38.6ns ± 1% 36.5ns ± 1% -5.60% (p=0.000 n=10+10) RuneCount/lenruneslice/ASCII-16 4.12ns ± 0% 4.11ns ± 0% ~ (p=0.382 n=10+10) RuneCount/lenruneslice/Japanese-16 25.4ns ± 2% 25.6ns ± 2% ~ (p=0.138 n=9+10) RuneCount/lenruneslice/MixedLength-16 17.1ns ± 0% 17.2ns ± 0% +0.59% (p=0.000 n=9+9) RuneCount/rangeloop/ASCII-16 3.30ns ± 1% 3.29ns ± 0% ~ (p=0.267 n=10+10) RuneCount/rangeloop/Japanese-16 20.1ns ± 1% 24.9ns ± 1% +24.31% (p=0.000 n=9+9) RuneCount/rangeloop/MixedLength-16 16.5ns ± 1% 16.7ns ± 1% +1.34% (p=0.000 n=10+10) RuneCount/utf8.RuneCountInString/ASCII-16 5.71ns ± 1% 5.73ns ± 2% ~ (p=0.579 n=10+10) RuneCount/utf8.RuneCountInString/Japanese-16 22.0ns ± 6% 18.4ns ± 3% -16.41% (p=0.000 n=9+10) RuneCount/utf8.RuneCountInString/MixedLength-16 15.0ns ± 1% 14.9ns ± 1% -1.01% (p=0.004 n=9+10) RuneIterate/range/ASCII-16 2.69ns ± 1% 2.72ns ± 0% +0.94% (p=0.026 n=10+9) RuneIterate/range/Japanese-16 24.5ns ± 2% 25.3ns ± 2% +3.23% (p=0.000 n=9+10) RuneIterate/range/MixedLength-16 17.0ns ± 1% 17.1ns ± 1% +0.85% (p=0.000 n=10+10) RuneIterate/range1/ASCII-16 2.70ns ± 1% 2.72ns ± 0% ~ (p=0.058 n=9+9) RuneIterate/range1/Japanese-16 24.1ns ± 2% 25.2ns ± 3% +4.30% (p=0.000 n=10+10) RuneIterate/range1/MixedLength-16 16.9ns ± 1% 17.7ns ± 0% +5.04% (p=0.000 n=10+8) RuneIterate/range2/ASCII-16 2.84ns ± 8% 2.72ns ± 1% -4.28% (p=0.003 n=10+9) RuneIterate/range2/Japanese-16 22.7ns ± 4% 25.2ns ± 3% +10.97% (p=0.000 n=10+10) RuneIterate/range2/MixedLength-16 17.0ns ± 1% 17.2ns ± 0% +0.95% (p=0.000 n=10+10) ArrayEqual-16 0.40ns ± 5% 0.35ns ± 2% -11.83% (p=0.000 n=10+10) Func/Name-16 8.05ns ± 1% 8.09ns ± 1% +0.40% (p=0.025 n=8+10) Func/Entry-16 1.73ns ± 1% 1.66ns ± 1% -3.93% (p=0.000 n=10+10) Func/FileLine-16 27.5ns ± 2% 26.0ns ± 0% -5.50% (p=0.000 n=10+10) [Geo mean] 16.7ns 15.7ns -6.08% name old speed new speed delta SetTypePtr-16 11.0GB/s ± 1% 11.0GB/s ± 3% ~ (p=0.684 n=10+10) SetTypePtr8-16 15.5GB/s ± 0% 15.5GB/s ± 0% ~ (p=0.123 n=10+10) SetTypePtr16-16 31.0GB/s ± 1% 31.1GB/s ± 0% ~ (p=0.123 n=10+10) SetTypePtr32-16 62.1GB/s ± 0% 62.2GB/s ± 0% ~ (p=0.123 n=10+10) SetTypePtr64-16 124GB/s ± 0% 124GB/s ± 0% ~ (p=0.684 n=10+10) SetTypePtr126-16 146GB/s ± 0% 146GB/s ± 0% ~ (p=0.481 n=10+10) SetTypePtr128-16 154GB/s ± 0% 154GB/s ± 0% ~ (p=0.243 n=9+10) SetTypePtrSlice-16 151GB/s ± 1% 151GB/s ± 1% ~ (p=0.497 n=9+10) SetTypeNode1-16 5.82GB/s ± 1% 5.82GB/s ± 0% ~ (p=0.353 n=10+10) SetTypeNode1Slice-16 76.1GB/s ± 1% 77.0GB/s ± 1% +1.19% (p=0.000 n=10+10) SetTypeNode8-16 19.4GB/s ± 0% 19.4GB/s ± 0% ~ (p=0.130 n=8+8) SetTypeNode8Slice-16 113GB/s ± 0% 113GB/s ± 0% ~ (p=0.604 n=10+9) SetTypeNode64-16 76.5GB/s ± 0% 76.5GB/s ± 0% ~ (p=0.190 n=10+10) SetTypeNode64Slice-16 97.8GB/s ± 0% 97.7GB/s ± 0% ~ (p=0.549 n=9+10) SetTypeNode64Dead-16 95.5GB/s ± 0% 95.7GB/s ± 0% ~ (p=0.118 n=10+6) SetTypeNode64DeadSlice-16 112GB/s ± 0% 112GB/s ± 0% ~ (p=0.353 n=10+10) SetTypeNode124-16 146GB/s ± 0% 146GB/s ± 0% ~ (p=0.853 n=10+10) SetTypeNode124Slice-16 146GB/s ± 5% 149GB/s ± 0% ~ (p=0.315 n=10+10) SetTypeNode126-16 154GB/s ± 0% 154GB/s ± 0% ~ (p=0.356 n=10+9) SetTypeNode126Slice-16 150GB/s ± 0% 150GB/s ± 0% ~ (p=0.095 n=9+10) SetTypeNode128-16 107GB/s ± 0% 107GB/s ± 0% +0.31% (p=0.003 n=9+10) SetTypeNode128Slice-16 119GB/s ± 0% 120GB/s ± 0% ~ (p=0.156 n=10+9) SetTypeNode130-16 108GB/s ± 0% 108GB/s ± 0% +0.33% (p=0.002 n=10+10) SetTypeNode130Slice-16 119GB/s ± 0% 119GB/s ± 0% ~ (p=0.739 n=10+10) SetTypeNode1024-16 160GB/s ± 0% 159GB/s ± 1% ~ (p=0.113 n=9+9) SetTypeNode1024Slice-16 144GB/s ± 0% 144GB/s ± 0% ~ (p=0.063 n=10+10) Hash5-16 2.59GB/s ± 1% 2.49GB/s ± 0% -3.90% (p=0.000 n=10+9) Hash16-16 7.85GB/s ± 1% 7.23GB/s ± 1% -7.92% (p=0.000 n=10+10) Hash64-16 24.0GB/s ± 0% 23.9GB/s ± 0% ~ (p=0.190 n=9+9) Hash1024-16 62.4GB/s ± 0% 62.3GB/s ± 0% -0.16% (p=0.017 n=9+10) Hash65536-16 74.0GB/s ± 0% 74.0GB/s ± 0% ~ (p=0.796 n=10+10) Memmove/1-16 1.08GB/s ± 0% 1.08GB/s ± 0% -0.21% (p=0.035 n=10+10) Memmove/2-16 2.16GB/s ± 0% 2.15GB/s ± 0% ~ (p=0.105 n=10+10) Memmove/3-16 3.24GB/s ± 1% 3.22GB/s ± 1% -0.49% (p=0.004 n=10+10) Memmove/4-16 3.89GB/s ± 0% 3.89GB/s ± 0% ~ (p=0.218 n=10+10) Memmove/5-16 4.42GB/s ± 0% 4.42GB/s ± 0% ~ (p=0.075 n=10+10) Memmove/6-16 5.31GB/s ± 0% 5.29GB/s ± 1% ~ (p=0.218 n=10+10) Memmove/7-16 6.19GB/s ± 0% 6.18GB/s ± 0% -0.15% (p=0.035 n=10+9) Memmove/8-16 7.07GB/s ± 0% 7.07GB/s ± 0% ~ (p=0.684 n=10+10) Memmove/9-16 7.22GB/s ± 0% 6.68GB/s ± 0% -7.37% (p=0.000 n=10+10) Memmove/10-16 8.02GB/s ± 0% 7.43GB/s ± 0% -7.38% (p=0.000 n=9+9) Memmove/11-16 8.83GB/s ± 0% 8.13GB/s ± 0% -7.87% (p=0.000 n=10+9) Memmove/12-16 9.62GB/s ± 0% 8.89GB/s ± 1% -7.61% (p=0.000 n=10+10) Memmove/13-16 10.4GB/s ± 0% 9.7GB/s ± 0% -7.20% (p=0.000 n=10+10) Memmove/14-16 11.2GB/s ± 0% 10.4GB/s ± 1% -7.64% (p=0.000 n=10+9) Memmove/15-16 12.0GB/s ± 0% 11.1GB/s ± 0% -7.46% (p=0.000 n=10+9) Memmove/16-16 12.8GB/s ± 0% 11.8GB/s ± 1% -7.67% (p=0.000 n=10+10) Memmove/32-16 23.8GB/s ± 0% 23.5GB/s ± 1% -1.20% (p=0.000 n=10+10) Memmove/64-16 44.2GB/s ± 0% 39.1GB/s ± 0% -11.56% (p=0.000 n=10+9) Memmove/128-16 68.7GB/s ± 0% 63.2GB/s ± 0% -7.95% (p=0.000 n=10+10) Memmove/256-16 104GB/s ± 0% 103GB/s ± 0% -1.13% (p=0.000 n=10+10) Memmove/512-16 129GB/s ± 1% 129GB/s ± 0% ~ (p=0.165 n=10+10) Memmove/1024-16 174GB/s ± 1% 174GB/s ± 1% ~ (p=0.258 n=9+9) Memmove/2048-16 213GB/s ± 1% 213GB/s ± 2% ~ (p=0.963 n=8+9) Memmove/4096-16 250GB/s ± 1% 240GB/s ± 4% -3.83% (p=0.006 n=9+9) MemmoveOverlap/32-16 19.8GB/s ± 1% 19.1GB/s ± 1% -3.40% (p=0.000 n=10+10) MemmoveOverlap/64-16 39.0GB/s ± 0% 38.8GB/s ± 0% -0.28% (p=0.001 n=9+9) MemmoveOverlap/128-16 62.2GB/s ± 0% 62.1GB/s ± 0% ~ (p=0.063 n=10+10) MemmoveOverlap/256-16 96.0GB/s ± 0% 95.8GB/s ± 0% -0.26% (p=0.009 n=10+10) MemmoveOverlap/512-16 83.6GB/s ±16% 89.2GB/s ± 0% ~ (p=0.696 n=10+8) MemmoveOverlap/1024-16 141GB/s ± 0% 140GB/s ± 0% -0.28% (p=0.006 n=8+10) MemmoveOverlap/2048-16 172GB/s ± 0% 171GB/s ± 1% -0.38% (p=0.008 n=9+9) MemmoveOverlap/4096-16 176GB/s ± 1% 177GB/s ± 1% +0.84% (p=0.001 n=8+10) MemmoveUnalignedDst/1-16 806MB/s ± 0% 802MB/s ± 1% -0.52% (p=0.023 n=10+10) MemmoveUnalignedDst/2-16 1.62GB/s ± 0% 1.62GB/s ± 0% -0.11% (p=0.041 n=10+10) MemmoveUnalignedDst/3-16 2.43GB/s ± 0% 2.43GB/s ± 0% -0.14% (p=0.006 n=9+9) MemmoveUnalignedDst/4-16 3.24GB/s ± 0% 3.23GB/s ± 1% -0.36% (p=0.007 n=10+10) MemmoveUnalignedDst/5-16 3.71GB/s ± 0% 3.71GB/s ± 0% ~ (p=0.063 n=10+10) MemmoveUnalignedDst/6-16 4.48GB/s ± 0% 4.47GB/s ± 0% ~ (p=0.912 n=10+10) MemmoveUnalignedDst/7-16 5.22GB/s ± 0% 5.22GB/s ± 0% ~ (p=1.000 n=10+10) MemmoveUnalignedDst/8-16 5.95GB/s ± 0% 5.93GB/s ± 1% -0.40% (p=0.023 n=10+10) MemmoveUnalignedDst/9-16 6.24GB/s ± 0% 6.24GB/s ± 0% ~ (p=0.912 n=10+10) MemmoveUnalignedDst/10-16 6.94GB/s ± 0% 6.94GB/s ± 0% ~ (p=0.353 n=10+10) MemmoveUnalignedDst/11-16 7.64GB/s ± 0% 7.63GB/s ± 0% ~ (p=0.393 n=10+10) MemmoveUnalignedDst/12-16 8.33GB/s ± 0% 8.33GB/s ± 0% ~ (p=0.971 n=10+10) MemmoveUnalignedDst/13-16 9.02GB/s ± 0% 9.01GB/s ± 0% ~ (p=0.436 n=10+10) MemmoveUnalignedDst/14-16 9.71GB/s ± 0% 9.71GB/s ± 0% ~ (p=0.280 n=10+10) MemmoveUnalignedDst/15-16 10.4GB/s ± 0% 10.4GB/s ± 1% ~ (p=0.853 n=10+10) MemmoveUnalignedDst/16-16 11.1GB/s ± 0% 11.1GB/s ± 0% ~ (p=0.089 n=10+10) MemmoveUnalignedDst/32-16 19.7GB/s ± 1% 19.6GB/s ± 0% ~ (p=0.075 n=10+10) MemmoveUnalignedDst/64-16 38.9GB/s ± 0% 38.8GB/s ± 0% ~ (p=0.218 n=10+10) MemmoveUnalignedDst/128-16 62.1GB/s ± 0% 62.1GB/s ± 0% ~ (p=0.549 n=10+9) MemmoveUnalignedDst/256-16 69.4GB/s ± 0% 69.3GB/s ± 0% ~ (p=0.105 n=10+10) MemmoveUnalignedDst/512-16 124GB/s ± 1% 124GB/s ± 0% ~ (p=0.762 n=10+8) MemmoveUnalignedDst/1024-16 136GB/s ± 0% 136GB/s ± 1% ~ (p=0.666 n=9+9) MemmoveUnalignedDst/2048-16 159GB/s ± 0% 159GB/s ± 0% ~ (p=0.574 n=8+8) MemmoveUnalignedDst/4096-16 161GB/s ± 0% 161GB/s ± 0% ~ (p=1.000 n=9+9) MemmoveUnalignedDstOverlap/32-16 7.84GB/s ± 0% 7.83GB/s ± 0% ~ (p=0.353 n=10+10) MemmoveUnalignedDstOverlap/64-16 14.0GB/s ± 0% 14.0GB/s ± 0% ~ (p=0.661 n=10+9) MemmoveUnalignedDstOverlap/128-16 27.4GB/s ± 0% 27.4GB/s ± 0% ~ (p=0.353 n=10+10) MemmoveUnalignedDstOverlap/256-16 50.4GB/s ± 0% 50.4GB/s ± 0% ~ (p=0.156 n=10+9) MemmoveUnalignedDstOverlap/512-16 60.7GB/s ± 4% 62.5GB/s ± 0% +3.07% (p=0.022 n=10+9) MemmoveUnalignedDstOverlap/1024-16 107GB/s ± 0% 107GB/s ± 0% ~ (p=0.234 n=8+8) MemmoveUnalignedDstOverlap/2048-16 146GB/s ± 0% 146GB/s ± 1% ~ (p=0.182 n=10+9) MemmoveUnalignedDstOverlap/4096-16 155GB/s ± 0% 155GB/s ± 0% ~ (p=0.400 n=10+9) MemmoveUnalignedSrc/1-16 882MB/s ± 0% 884MB/s ± 1% +0.24% (p=0.033 n=10+9) MemmoveUnalignedSrc/2-16 1.76GB/s ± 1% 1.77GB/s ± 0% +0.27% (p=0.028 n=10+9) MemmoveUnalignedSrc/3-16 2.43GB/s ± 0% 2.43GB/s ± 0% +0.26% (p=0.027 n=9+10) MemmoveUnalignedSrc/4-16 3.24GB/s ± 0% 3.24GB/s ± 1% ~ (p=0.079 n=9+10) MemmoveUnalignedSrc/5-16 3.73GB/s ± 0% 3.73GB/s ± 1% ~ (p=0.829 n=8+10) MemmoveUnalignedSrc/6-16 4.47GB/s ± 0% 4.49GB/s ± 0% +0.39% (p=0.000 n=10+10) MemmoveUnalignedSrc/7-16 5.22GB/s ± 0% 5.23GB/s ± 0% ~ (p=0.280 n=10+10) MemmoveUnalignedSrc/8-16 5.95GB/s ± 0% 5.98GB/s ± 0% +0.39% (p=0.001 n=10+9) MemmoveUnalignedSrc/9-16 6.24GB/s ± 0% 6.25GB/s ± 0% ~ (p=0.549 n=10+9) MemmoveUnalignedSrc/10-16 6.93GB/s ± 0% 6.94GB/s ± 0% ~ (p=0.604 n=10+9) MemmoveUnalignedSrc/11-16 7.63GB/s ± 0% 7.63GB/s ± 1% ~ (p=0.353 n=10+10) MemmoveUnalignedSrc/12-16 8.32GB/s ± 0% 8.32GB/s ± 0% ~ (p=0.218 n=10+10) MemmoveUnalignedSrc/13-16 9.02GB/s ± 0% 9.00GB/s ± 1% ~ (p=0.684 n=10+10) MemmoveUnalignedSrc/14-16 9.71GB/s ± 0% 9.71GB/s ± 0% ~ (p=0.739 n=10+10) MemmoveUnalignedSrc/15-16 10.4GB/s ± 0% 10.4GB/s ± 0% ~ (p=0.353 n=10+10) MemmoveUnalignedSrc/16-16 11.1GB/s ± 1% 11.1GB/s ± 0% ~ (p=0.579 n=10+10) MemmoveUnalignedSrc/32-16 20.0GB/s ± 1% 20.0GB/s ± 0% ~ (p=0.631 n=10+10) MemmoveUnalignedSrc/64-16 38.8GB/s ± 0% 38.8GB/s ± 0% ~ (p=0.579 n=10+10) MemmoveUnalignedSrc/128-16 61.2GB/s ± 0% 61.2GB/s ± 0% ~ (p=0.780 n=10+9) MemmoveUnalignedSrc/256-16 94.8GB/s ± 0% 92.2GB/s ± 1% -2.73% (p=0.000 n=10+10) MemmoveUnalignedSrc/512-16 119GB/s ± 0% 119GB/s ± 0% +0.26% (p=0.027 n=8+9) MemmoveUnalignedSrc/1024-16 141GB/s ± 0% 142GB/s ± 1% +1.07% (p=0.000 n=8+10) MemmoveUnalignedSrc/2048-16 157GB/s ± 0% 157GB/s ± 0% ~ (p=0.167 n=9+8) MemmoveUnalignedSrc/4096-16 161GB/s ± 0% 162GB/s ± 1% ~ (p=0.063 n=10+10) MemmoveUnalignedSrcOverlap/32-16 7.93GB/s ± 0% 7.88GB/s ± 0% -0.63% (p=0.000 n=9+10) MemmoveUnalignedSrcOverlap/64-16 15.5GB/s ± 0% 15.5GB/s ± 0% ~ (p=0.529 n=10+10) MemmoveUnalignedSrcOverlap/128-16 28.3GB/s ± 0% 28.3GB/s ± 0% ~ (p=0.218 n=10+10) MemmoveUnalignedSrcOverlap/256-16 41.5GB/s ± 0% 41.6GB/s ± 0% +0.35% (p=0.000 n=10+9) MemmoveUnalignedSrcOverlap/512-16 68.9GB/s ± 0% 68.8GB/s ± 0% ~ (p=0.541 n=9+8) MemmoveUnalignedSrcOverlap/1024-16 115GB/s ± 0% 115GB/s ± 0% ~ (p=0.382 n=8+8) MemmoveUnalignedSrcOverlap/2048-16 155GB/s ± 0% 144GB/s ±18% ~ (p=0.101 n=8+10) MemmoveUnalignedSrcOverlap/4096-16 160GB/s ± 0% 160GB/s ± 1% ~ (p=0.605 n=9+9) Memclr/5-16 5.81GB/s ± 1% 5.80GB/s ± 2% ~ (p=0.546 n=9+9) Memclr/16-16 15.5GB/s ± 0% 15.4GB/s ± 0% -0.32% (p=0.008 n=9+10) Memclr/64-16 51.8GB/s ± 0% 50.7GB/s ± 0% -2.22% (p=0.000 n=10+10) Memclr/256-16 113GB/s ± 0% 113GB/s ± 0% ~ (p=0.143 n=10+10) Memclr/4096-16 239GB/s ± 1% 237GB/s ± 0% -0.87% (p=0.000 n=10+10) Memclr/65536-16 79.8GB/s ± 0% 79.7GB/s ± 0% ~ (p=0.529 n=10+10) Memclr/1M-16 74.6GB/s ± 1% 74.7GB/s ± 1% ~ (p=0.529 n=10+10) Memclr/4M-16 48.7GB/s ± 1% 48.8GB/s ± 0% ~ (p=0.123 n=10+10) Memclr/8M-16 48.2GB/s ± 2% 48.6GB/s ± 0% ~ (p=0.408 n=10+8) Memclr/16M-16 43.6GB/s ± 4% 43.3GB/s ± 0% ~ (p=0.173 n=10+8) Memclr/64M-16 30.7GB/s ± 0% 30.7GB/s ± 0% ~ (p=0.113 n=10+9) GoMemclr/5-16 6.07GB/s ± 0% 6.08GB/s ± 0% ~ (p=0.367 n=9+10) GoMemclr/16-16 15.6GB/s ± 0% 15.6GB/s ± 0% -0.22% (p=0.004 n=9+9) GoMemclr/64-16 56.1GB/s ± 0% 56.1GB/s ± 0% ~ (p=0.968 n=10+9) GoMemclr/256-16 125GB/s ± 0% 124GB/s ± 0% ~ (p=0.912 n=10+10) MemclrRange/1K_2K-16 210GB/s ± 0% 224GB/s ± 1% +6.81% (p=0.000 n=10+10) MemclrRange/2K_8K-16 228GB/s ± 0% 228GB/s ± 0% ~ (p=0.684 n=10+10) MemclrRange/4K_16K-16 279GB/s ± 0% 279GB/s ± 0% ~ (p=0.780 n=9+10) MemclrRange/160K_228K-16 80.3GB/s ± 0% 80.2GB/s ± 0% ~ (p=0.165 n=10+10) Copy/1Byte-16 808MB/s ± 1% 810MB/s ± 0% +0.28% (p=0.000 n=10+8) Copy/1String-16 810MB/s ± 0% 811MB/s ± 0% ~ (p=0.105 n=10+10) Copy/2Byte-16 1.62GB/s ± 0% 1.62GB/s ± 0% ~ (p=0.182 n=10+9) Copy/2String-16 1.62GB/s ± 1% 1.62GB/s ± 1% ~ (p=1.000 n=10+10) Copy/4Byte-16 3.22GB/s ± 0% 3.24GB/s ± 0% +0.46% (p=0.000 n=10+10) Copy/4String-16 3.24GB/s ± 0% 3.24GB/s ± 0% ~ (p=0.075 n=10+10) Copy/8Byte-16 5.86GB/s ± 0% 5.96GB/s ± 0% +1.82% (p=0.000 n=9+9) Copy/8String-16 5.95GB/s ± 0% 5.99GB/s ± 0% +0.59% (p=0.000 n=9+10) Copy/12Byte-16 8.32GB/s ± 0% 8.32GB/s ± 0% ~ (p=0.190 n=9+9) Copy/12String-16 8.31GB/s ± 0% 8.29GB/s ± 0% ~ (p=0.068 n=9+10) Copy/16Byte-16 11.1GB/s ± 0% 11.1GB/s ± 0% +0.18% (p=0.003 n=10+9) Copy/16String-16 11.1GB/s ± 0% 11.1GB/s ± 0% -0.31% (p=0.009 n=10+10) Copy/32Byte-16 19.6GB/s ± 1% 19.8GB/s ± 1% +0.72% (p=0.002 n=10+10) Copy/32String-16 20.0GB/s ± 0% 19.5GB/s ± 0% -2.19% (p=0.000 n=10+10) Copy/128Byte-16 62.2GB/s ± 0% 62.1GB/s ± 0% ~ (p=0.661 n=9+10) Copy/128String-16 61.9GB/s ± 0% 61.7GB/s ± 0% -0.35% (p=0.005 n=10+10) Copy/1024Byte-16 169GB/s ± 2% 171GB/s ± 1% +1.21% (p=0.000 n=9+10) Copy/1024String-16 169GB/s ± 0% 172GB/s ± 1% +1.57% (p=0.000 n=10+9) CompareStringBigUnaligned-16 43.8GB/s ± 1% 43.7GB/s ± 1% ~ (p=0.370 n=9+8) CompareStringBig-16 47.6GB/s ± 3% 47.3GB/s ± 3% ~ (p=0.243 n=9+10) [Geo mean] 25.3GB/s 28.0GB/s +10.66% name old p50-ns new p50-ns delta ReadMemStatsLatency-16 98.4k ±37% 112.7k ±64% ~ (p=0.436 n=10+10) ReadMetricsLatency-16 1.69k ± 3% 1.70k ± 2% ~ (p=0.646 n=9+10) GoroutineProfile/small-nil/idle-16 3.75k ± 4% 3.72k ± 1% ~ (p=0.447 n=10+9) GoroutineProfile/small-nil/loaded-16 4.33k ± 3% 4.31k ± 5% ~ (p=0.931 n=9+9) GoroutineProfile/small/idle-16 102k ± 3% 101k ± 4% ~ (p=0.113 n=9+9) GoroutineProfile/small/loaded-16 214k ± 3% 215k ± 6% ~ (p=0.842 n=9+10) GoroutineProfile/large-nil/idle-16 3.70k ± 2% 3.65k ± 2% ~ (p=0.075 n=10+10) GoroutineProfile/large-nil/loaded-16 4.36k ± 3% 4.31k ±10% ~ (p=0.631 n=10+10) GoroutineProfile/large/idle-16 2.56M ± 1% 2.51M ± 1% -2.28% (p=0.000 n=10+10) GoroutineProfile/large/loaded-16 6.77M ± 4% 6.85M ±19% ~ (p=0.536 n=7+10) GoroutineProfile/sparse-nil/idle-16 3.66k ± 1% 3.64k ± 2% ~ (p=0.136 n=9+9) GoroutineProfile/sparse-nil/loaded-16 4.25k ± 5% 4.15k ± 4% ~ (p=0.190 n=10+10) GoroutineProfile/sparse/idle-16 102k ± 4% 101k ± 3% ~ (p=0.447 n=10+9) GoroutineProfile/sparse/loaded-16 216k ± 4% 218k ± 4% ~ (p=0.549 n=9+10) [Geo mean] 35.8k 35.9k +0.35% name old p90-ns new p90-ns delta ReadMemStatsLatency-16 983k ±310% 200k ±34% -79.62% (p=0.034 n=10+8) ReadMetricsLatency-16 4.01k ±35% 3.75k ±17% ~ (p=0.315 n=10+10) GoroutineProfile/small-nil/idle-16 4.21k ± 4% 4.27k ± 8% ~ (p=0.968 n=9+10) GoroutineProfile/small-nil/loaded-16 5.58k ± 8% 5.35k ±12% ~ (p=0.190 n=10+10) GoroutineProfile/small/idle-16 108k ± 6% 107k ± 7% ~ (p=0.497 n=9+10) GoroutineProfile/small/loaded-16 450k ± 5% 432k ± 3% -3.92% (p=0.002 n=9+9) GoroutineProfile/large-nil/idle-16 4.13k ± 7% 4.04k ± 2% ~ (p=0.181 n=10+8) GoroutineProfile/large-nil/loaded-16 5.76k ± 4% 5.67k ± 5% ~ (p=0.190 n=10+10) GoroutineProfile/large/idle-16 2.63M ± 2% 2.58M ± 1% -1.97% (p=0.000 n=10+10) GoroutineProfile/large/loaded-16 16.9M ± 4% 17.0M ± 6% ~ (p=0.661 n=9+10) GoroutineProfile/sparse-nil/idle-16 4.21k ±10% 4.07k ± 7% ~ (p=0.128 n=10+10) GoroutineProfile/sparse-nil/loaded-16 5.55k ± 8% 5.38k ± 6% ~ (p=0.089 n=10+10) GoroutineProfile/sparse/idle-16 106k ± 4% 106k ± 3% ~ (p=0.661 n=10+9) GoroutineProfile/sparse/loaded-16 454k ± 6% 441k ± 6% -2.86% (p=0.043 n=10+10) [Geo mean] 58.4k 51.0k -12.61% name old p99-ns new p99-ns delta ReadMemStatsLatency-16 983k ±310% 200k ±34% -79.62% (p=0.034 n=10+8) ReadMetricsLatency-16 26.6k ±22% 26.6k ±17% ~ (p=0.971 n=10+10) GoroutineProfile/small-nil/idle-16 5.27k ±12% 5.19k ±12% ~ (p=0.579 n=10+10) GoroutineProfile/small-nil/loaded-16 7.28k ± 3% 7.04k ± 9% ~ (p=0.113 n=9+10) GoroutineProfile/small/idle-16 114k ± 6% 113k ± 6% ~ (p=0.604 n=9+10) GoroutineProfile/small/loaded-16 4.84M ±70% 6.06M ±131% ~ (p=0.842 n=9+10) GoroutineProfile/large-nil/idle-16 5.38k ±19% 5.26k ±13% ~ (p=0.912 n=10+10) GoroutineProfile/large-nil/loaded-16 7.38k ± 3% 7.23k ± 4% ~ (p=0.143 n=10+10) GoroutineProfile/large/idle-16 2.79M ± 5% 2.72M ± 3% ~ (p=0.089 n=10+10) GoroutineProfile/large/loaded-16 24.0M ±24% 24.9M ±29% ~ (p=0.684 n=10+10) GoroutineProfile/sparse-nil/idle-16 5.32k ±17% 5.49k ±17% ~ (p=0.684 n=10+10) GoroutineProfile/sparse-nil/loaded-16 7.25k ± 4% 6.97k ± 5% -3.90% (p=0.005 n=9+10) GoroutineProfile/sparse/idle-16 113k ± 5% 112k ± 6% ~ (p=0.631 n=10+10) GoroutineProfile/sparse/loaded-16 4.00M ±66% 4.26M ±65% ~ (p=0.489 n=9+9) [Geo mean] 107k 97k -9.55% name old alloc/op new alloc/op delta NewEmptyMap-16 0.00B 0.00B ~ (all equal) NewSmallMap-16 0.00B 0.00B ~ (all equal) MapPopulate/1-16 0.00B 0.00B ~ (all equal) MapPopulate/10-16 179B ± 0% 179B ± 0% ~ (all equal) MapPopulate/100-16 3.35kB ± 0% 3.35kB ± 0% ~ (p=0.294 n=10+8) MapPopulate/1000-16 53.3kB ± 0% 53.3kB ± 0% ~ (p=1.000 n=8+10) MapPopulate/10000-16 428kB ± 0% 428kB ± 0% ~ (p=0.469 n=10+10) MapPopulate/100000-16 3.62MB ± 0% 3.62MB ± 0% ~ (p=0.888 n=9+10) MapStringConversion/32/simple-16 0.00B 0.00B ~ (all equal) MapStringConversion/32/struct-16 0.00B 0.00B ~ (all equal) MapStringConversion/32/array-16 0.00B 0.00B ~ (all equal) MapStringConversion/64/simple-16 0.00B 0.00B ~ (all equal) MapStringConversion/64/struct-16 0.00B 0.00B ~ (all equal) MapStringConversion/64/array-16 0.00B 0.00B ~ (all equal) NewEmptyMapHintLessThan8-16 0.00B 0.00B ~ (all equal) NewEmptyMapHintGreaterThan8-16 1.15kB ± 0% 1.15kB ± 0% ~ (all equal) MapAppendAssign/Int32/256-16 41.7B ±15% 44.3B ±12% ~ (p=0.106 n=10+10) MapAppendAssign/Int32/65536-16 22.6B ± 6% 23.5B ± 6% +4.19% (p=0.025 n=9+10) MapAppendAssign/Int64/256-16 43.5B ±10% 42.9B ± 7% ~ (p=0.757 n=10+10) MapAppendAssign/Int64/65536-16 24.7B ± 7% 21.8B ± 6% -11.74% (p=0.000 n=10+10) MapAppendAssign/Str/256-16 87.6B ±10% 89.2B ± 9% ~ (p=0.379 n=10+10) MapAppendAssign/Str/65536-16 45.1B ±14% 47.2B ± 8% ~ (p=0.150 n=10+9) CreateGoroutinesCapture-16 144B ± 0% 144B ± 0% ~ (all equal) [Geo mean] 769B 770B +0.21% name old allocs/op new allocs/op delta NewEmptyMap-16 0.00 0.00 ~ (all equal) NewSmallMap-16 0.00 0.00 ~ (all equal) MapPopulate/1-16 0.00 0.00 ~ (all equal) MapPopulate/10-16 1.00 ± 0% 1.00 ± 0% ~ (all equal) MapPopulate/100-16 17.0 ± 0% 17.0 ± 0% ~ (all equal) MapPopulate/1000-16 73.0 ± 0% 73.0 ± 0% ~ (all equal) MapPopulate/10000-16 320 ± 0% 320 ± 0% ~ (p=1.000 n=10+10) MapPopulate/100000-16 4.00k ± 0% 4.00k ± 0% ~ (p=0.753 n=10+10) MapStringConversion/32/simple-16 0.00 0.00 ~ (all equal) MapStringConversion/32/struct-16 0.00 0.00 ~ (all equal) MapStringConversion/32/array-16 0.00 0.00 ~ (all equal) MapStringConversion/64/simple-16 0.00 0.00 ~ (all equal) MapStringConversion/64/struct-16 0.00 0.00 ~ (all equal) MapStringConversion/64/array-16 0.00 0.00 ~ (all equal) NewEmptyMapHintLessThan8-16 0.00 0.00 ~ (all equal) NewEmptyMapHintGreaterThan8-16 1.00 ± 0% 1.00 ± 0% ~ (all equal) MapAppendAssign/Int32/256-16 0.00 0.00 ~ (all equal) MapAppendAssign/Int32/65536-16 0.00 0.00 ~ (all equal) MapAppendAssign/Int64/256-16 0.00 0.00 ~ (all equal) MapAppendAssign/Int64/65536-16 0.00 0.00 ~ (all equal) MapAppendAssign/Str/256-16 0.00 0.00 ~ (all equal) MapAppendAssign/Str/65536-16 0.00 0.00 ~ (all equal) CreateGoroutinesCapture-16 5.00 ± 0% 5.00 ± 0% ~ (all equal) [Geo mean] 26.0 26.0 +0.00% Change-Id: I5fb03e93df8b380e04795afbdcd1c94aeeecacc6 Reviewed-on: https://go-review.googlesource.com/c/go/+/454255 Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com> Run-TryBot: Jakub Ciolek <jakub@ciolek.dev> TryBot-Result: Gopher Robot <gobot@golang.org>
2022-11-30 15:31:26 +01:00
// Turn known-size calls to memclrNoHeapPointers into a Zero.
// Note that we are using types.Types[types.TUINT8] instead of sptr.Type.Elem() - see issue 55122 and CL 431496 for more details.
(SelectN [0] call:(StaticCall {sym} sptr (Const(64|32) [c]) mem))
cmd/compile: inline constant sized memclrNoHeapPointers calls on PPC64 Update the function isInlinableMemclr for ppc64 and ppc64le to enable inlining for the constant sized cases < 512. Larger cases can use dcbz which performs better but requires alignment checking so it is best to continue using memclrNoHeapPointers for those cases. Results on p10: MemclrKnownSize1 2.07ns ± 0% 0.57ns ± 0% -72.59% MemclrKnownSize2 2.56ns ± 5% 0.57ns ± 0% -77.82% MemclrKnownSize4 5.15ns ± 0% 0.57ns ± 0% -89.00% MemclrKnownSize8 2.23ns ± 0% 0.57ns ± 0% -74.57% MemclrKnownSize16 2.23ns ± 0% 0.50ns ± 0% -77.74% MemclrKnownSize32 2.28ns ± 0% 0.56ns ± 0% -75.28% MemclrKnownSize64 2.49ns ± 0% 0.72ns ± 0% -70.95% MemclrKnownSize112 2.97ns ± 2% 1.14ns ± 0% -61.72% MemclrKnownSize128 4.64ns ± 6% 2.45ns ± 1% -47.17% MemclrKnownSize192 5.45ns ± 5% 2.79ns ± 0% -48.87% MemclrKnownSize248 4.51ns ± 0% 2.83ns ± 0% -37.12% MemclrKnownSize256 6.34ns ± 1% 3.58ns ± 0% -43.53% MemclrKnownSize512 3.64ns ± 0% 3.64ns ± 0% -0.03% MemclrKnownSize1024 4.73ns ± 0% 4.73ns ± 0% +0.01% MemclrKnownSize4096 17.1ns ± 0% 17.1ns ± 0% +0.07% MemclrKnownSize512KiB 2.12µs ± 0% 2.12µs ± 0% ~ (all equal) Change-Id: If1abf5749f4802c64523a41fe0058bd144d0ea46 Reviewed-on: https://go-review.googlesource.com/c/go/+/464340 Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com> Reviewed-by: Jakub Ciolek <jakub@ciolek.dev> Reviewed-by: Archana Ravindar <aravind5@in.ibm.com> Reviewed-by: Ian Lance Taylor <iant@google.com> Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org> TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Than McIntosh <thanm@google.com>
2023-01-31 14:27:30 -06:00
&& isInlinableMemclr(config, int64(c))
cmd/compile: inline known-size memclrNoHeapPointers calls This patch rewrites known-size calls to memclrNoHeapPointers with an OpZero. This significantly improves performance and lets some clears get DSE'd. One of the cases where this applies is zeroing a known-size array, example: var x [256]int8 ... for a := range x { x[a] = 0 } Other cases can be found in the runtime itself where memclrNoHeapPointers is sometimes directly invoked with a constant. It seems that for some sized-clears on some architectures (AMD64, maybe others) the memcrlNoHeapPointers is more performant than OpZero. See the issue #56997 for more details. Benches ARM (M1 Pro): name old time/op new time/op delta MemclrKnownSize1-10 2.03ns ± 0% 0.31ns ± 0% -84.69% (p=0.000 n=18+19) MemclrKnownSize2-10 1.97ns ± 0% 0.31ns ± 0% -84.19% (p=0.000 n=12+19) MemclrKnownSize4-10 2.02ns ± 0% 0.31ns ± 0% -84.56% (p=0.000 n=12+20) MemclrKnownSize8-10 2.02ns ± 0% 0.31ns ± 0% -84.59% (p=0.000 n=14+19) MemclrKnownSize16-10 2.15ns ± 0% 0.31ns ± 0% -85.50% (p=0.000 n=18+19) MemclrKnownSize32-10 2.48ns ± 0% 0.31ns ± 0% -87.48% (p=0.000 n=20+19) MemclrKnownSize64-10 1.93ns ± 0% 0.62ns ± 0% -67.88% (p=0.000 n=20+19) MemclrKnownSize112-10 2.48ns ± 0% 1.80ns ± 0% -27.74% (p=0.000 n=19+20) MemclrKnownSize128-10 10.0ns ±112% 2.0ns ± 0% -79.76% (p=0.000 n=18+17) MemclrKnownSize192-10 27.4ns ±103% 2.6ns ± 0% -90.38% (p=0.000 n=16+19) MemclrKnownSize248-10 9.67ns ±43% 3.26ns ± 0% -66.29% (p=0.000 n=19+19) MemclrKnownSize256-10 85.4ns ±148% 3.3ns ± 0% -96.18% (p=0.000 n=20+20) MemclrKnownSize512-10 223ns ±54% 6ns ± 0% -97.42% (p=0.000 n=18+20) MemclrKnownSize1024-10 216ns ±26% 11ns ± 0% -95.00% (p=0.000 n=18+15) MemclrKnownSize4096-10 265ns ± 2% 88ns ± 0% -66.84% (p=0.000 n=19+17) MemclrKnownSize512KiB-10 9.91µs ± 1% 10.23µs ± 2% +3.14% (p=0.000 n=19+19) [Geo mean] 15.6ns 2.5ns -83.62% name old speed new speed delta MemclrKnownSize1-10 493MB/s ± 0% 3216MB/s ± 0% +553.04% (p=0.000 n=18+19) MemclrKnownSize2-10 1.02GB/s ± 0% 6.43GB/s ± 0% +532.33% (p=0.000 n=16+19) MemclrKnownSize4-10 1.99GB/s ± 0% 12.86GB/s ± 0% +547.67% (p=0.000 n=18+20) MemclrKnownSize8-10 3.96GB/s ± 0% 25.72GB/s ± 0% +548.81% (p=0.000 n=19+19) MemclrKnownSize16-10 7.46GB/s ± 0% 51.43GB/s ± 0% +589.72% (p=0.000 n=20+19) MemclrKnownSize32-10 12.9GB/s ± 0% 102.9GB/s ± 0% +698.60% (p=0.000 n=20+18) MemclrKnownSize64-10 33.1GB/s ± 0% 103.0GB/s ± 0% +211.34% (p=0.000 n=19+19) MemclrKnownSize112-10 45.1GB/s ± 0% 62.4GB/s ± 0% +38.38% (p=0.000 n=19+20) MemclrKnownSize128-10 13.3GB/s ±107% 63.5GB/s ± 0% +378.03% (p=0.000 n=19+18) MemclrKnownSize192-10 6.97GB/s ±139% 72.72GB/s ± 0% +943.44% (p=0.000 n=19+19) MemclrKnownSize248-10 25.9GB/s ±46% 76.1GB/s ± 0% +194.16% (p=0.000 n=20+17) MemclrKnownSize256-10 8.64GB/s ±196% 78.51GB/s ± 0% +808.19% (p=0.000 n=20+20) MemclrKnownSize512-10 2.33GB/s ±86% 89.13GB/s ± 0% +3719.50% (p=0.000 n=17+20) MemclrKnownSize1024-10 4.85GB/s ±32% 94.93GB/s ± 0% +1856.74% (p=0.000 n=18+19) MemclrKnownSize4096-10 15.4GB/s ± 2% 46.6GB/s ± 0% +201.55% (p=0.000 n=19+18) MemclrKnownSize512KiB-10 52.9GB/s ± 1% 51.3GB/s ± 2% -3.04% (p=0.000 n=19+19) [Geo mean] 7.54GB/s 42.86GB/s +468.76% Intel Alder Lake 12600k: name old time/op new time/op delta MemclrKnownSize1-16 0.59ns ± 3% 0.38ns ± 6% -36.00% (p=0.000 n=19+18) MemclrKnownSize2-16 0.57ns ± 1% 0.19ns ± 5% -66.27% (p=0.000 n=19+19) MemclrKnownSize4-16 0.66ns ± 2% 0.36ns ±21% -45.12% (p=0.000 n=19+20) MemclrKnownSize8-16 0.74ns ± 1% 0.30ns ±26% -59.81% (p=0.000 n=18+20) MemclrKnownSize16-16 1.00ns ± 7% 0.21ns ± 8% -79.51% (p=0.000 n=20+19) MemclrKnownSize32-16 0.95ns ± 1% 0.40ns ± 1% -57.61% (p=0.000 n=20+18) MemclrKnownSize64-16 1.20ns ± 2% 0.41ns ± 0% -65.82% (p=0.000 n=20+18) MemclrKnownSize112-16 1.27ns ± 2% 1.03ns ± 0% -19.35% (p=0.000 n=20+18) MemclrKnownSize128-16 1.34ns ± 2% 1.03ns ± 0% -23.02% (p=0.000 n=20+20) MemclrKnownSize192-16 1.92ns ± 2% 1.44ns ± 0% -24.89% (p=0.000 n=20+16) MemclrKnownSize248-16 2.77ns ± 1% 3.29ns ± 0% +18.81% (p=0.000 n=20+16) MemclrKnownSize256-16 1.92ns ± 1% 1.86ns ± 0% -3.49% (p=0.000 n=19+15) MemclrKnownSize512-16 2.81ns ± 2% 3.49ns ± 0% +24.15% (p=0.000 n=20+17) MemclrKnownSize1024-16 4.02ns ± 1% 6.78ns ± 0% +68.44% (p=0.000 n=20+18) MemclrKnownSize4096-16 17.2ns ± 2% 14.4ns ± 0% -16.73% (p=0.000 n=20+17) MemclrKnownSize512KiB-16 6.71µs ± 1% 6.52µs ± 0% -2.85% (p=0.000 n=20+18) [Geo mean] 2.60ns 1.71ns -34.06% name old speed new speed delta MemclrKnownSize1-16 1.71GB/s ± 3% 2.67GB/s ± 6% +56.39% (p=0.000 n=19+18) MemclrKnownSize2-16 3.52GB/s ± 2% 10.43GB/s ± 6% +196.04% (p=0.000 n=20+20) MemclrKnownSize4-16 6.06GB/s ± 1% 10.83GB/s ±11% +78.63% (p=0.000 n=19+18) MemclrKnownSize8-16 10.7GB/s ± 1% 27.0GB/s ±21% +151.49% (p=0.000 n=18+20) MemclrKnownSize16-16 16.0GB/s ± 8% 78.1GB/s ± 7% +387.24% (p=0.000 n=20+19) MemclrKnownSize32-16 33.6GB/s ± 1% 79.4GB/s ± 1% +135.89% (p=0.000 n=20+18) MemclrKnownSize64-16 53.3GB/s ± 2% 155.9GB/s ± 0% +192.58% (p=0.000 n=20+18) MemclrKnownSize112-16 88.0GB/s ± 2% 109.1GB/s ± 0% +23.97% (p=0.000 n=20+18) MemclrKnownSize128-16 95.3GB/s ± 2% 123.8GB/s ± 0% +29.88% (p=0.000 n=20+20) MemclrKnownSize192-16 100GB/s ± 2% 133GB/s ± 0% +33.12% (p=0.000 n=20+17) MemclrKnownSize248-16 89.7GB/s ± 1% 75.5GB/s ± 0% -15.84% (p=0.000 n=20+19) MemclrKnownSize256-16 133GB/s ± 1% 138GB/s ± 0% +3.61% (p=0.000 n=19+14) MemclrKnownSize512-16 182GB/s ± 2% 147GB/s ± 0% -19.46% (p=0.000 n=20+17) MemclrKnownSize1024-16 254GB/s ± 1% 151GB/s ± 0% -40.64% (p=0.000 n=20+18) MemclrKnownSize4096-16 237GB/s ± 2% 285GB/s ± 0% +20.09% (p=0.000 n=20+17) MemclrKnownSize512KiB-16 78.2GB/s ± 1% 80.4GB/s ± 0% +2.93% (p=0.000 n=20+18) [Geo mean] 42.1GB/s 63.8GB/s +51.53% compilecmp linux/amd64: runtime runtime.(*pallocData).allocAll 85 -> 45 (-47.06%) runtime.(*pageAlloc).allocRange 942 -> 923 (-2.02%) runtime.(*pageAlloc).free 798 -> 774 (-3.01%) runtime.(*pageBits).clearAll 66 -> 20 (-69.70%) runtime.startCheckmarks 255 -> 246 (-3.53%) runtime.(*pallocData).freeAll 86 -> 46 (-46.51%) runtime.(*pallocBits).freeAll 66 -> 20 (-69.70%) runtime.(*consistentHeapStats).unsafeClear 66 -> 19 (-71.21%) runtime.newproc1 965 -> 933 (-3.32%) crypto/rc4 crypto/rc4.(*Cipher).Reset 78 -> 69 (-11.54%) compress/bzip2 compress/bzip2.(*reader).readBlock 2973 -> 2941 (-1.08%) image/jpeg image/jpeg.(*decoder).processDHT 1179 -> 1166 (-1.10%) index/suffixarray index/suffixarray.bucketMax_8_32 394 -> 241 (-38.83%) index/suffixarray.freq_8_32 317 -> 185 (-41.64%) index/suffixarray.freq_8_64 317 -> 178 (-43.85%) index/suffixarray.bucketMin_8_32 394 -> 243 (-38.32%) index/suffixarray.bucketMin_8_64 398 -> 234 (-41.21%) index/suffixarray.bucketMax_8_64 398 -> 234 (-41.21%) compress/flate compress/flate.(*huffmanBitWriter).generateCodegen 965 -> 838 (-13.16%) compress/flate.(*compressor).reset 429 -> 409 (-4.66%) cmd/vendor/golang.org/x/sys/unix cmd/vendor/golang.org/x/sys/unix.(*FdSet).Zero 66 -> 60 (-9.09%) cmd/vendor/golang.org/x/sys/unix.(*Ifreq).SetInet4Addr 211 -> 129 (-38.86%) cmd/vendor/golang.org/x/sys/unix.(*Ifreq).SetUint32 98 -> 14 (-85.71%) cmd/vendor/golang.org/x/sys/unix.(*Ifreq).clear 66 -> 11 (-83.33%) cmd/vendor/golang.org/x/sys/unix.(*Ifreq).SetUint16 101 -> 15 (-85.15%) cmd/vendor/golang.org/x/sys/unix.(*CPUSet).Zero 66 -> 60 (-9.09%) internal/coverage/decodemeta internal/coverage/decodemeta.(*CoverageMetaFileReader).rdUint64 325 -> 293 (-9.85%) crypto/tls crypto/tls.(*halfConn).setTrafficSecret 253 -> 247 (-2.37%) crypto/tls.(*Conn).readRecordOrCCS 10315 -> 10283 (-0.31%) crypto/tls.(*halfConn).changeCipherSpec 271 -> 261 (-3.69%) crypto/tls.(*Conn).writeRecordLocked 1765 -> 1748 (-0.96%) file before after Δ % runtime.s 512467 512164 -303 -0.059% crypto/rc4.s 955 946 -9 -0.942% compress/bzip2.s 9586 9554 -32 -0.334% image/jpeg.s 32122 32109 -13 -0.040% index/suffixarray.s 38547 37644 -903 -2.343% compress/flate.s 46668 46521 -147 -0.315% cmd/vendor/golang.org/x/sys/unix.s 118620 118301 -319 -0.269% internal/coverage/decodemeta.s 7224 7192 -32 -0.443% crypto/tls.s 288762 288697 -65 -0.023% cmd/compile/internal/ssa.s 3639799 3640727 +928 +0.025% total 20790248 20789353 -895 -0.004% src/runtime benchmarks (Linux Alder Lake 12600k): name old time/op new time/op delta MakeChan/Byte-16 26.2ns ± 2% 25.6ns ± 3% -2.05% (p=0.003 n=9+10) MakeChan/Int-16 33.9ns ± 2% 33.3ns ± 4% -1.99% (p=0.015 n=10+10) MakeChan/Ptr-16 54.2ns ± 2% 53.7ns ± 1% -0.90% (p=0.016 n=10+9) MakeChan/Struct/0-16 23.8ns ± 3% 23.4ns ± 1% -1.72% (p=0.009 n=10+8) MakeChan/Struct/32-16 55.9ns ± 2% 53.9ns ± 1% -3.48% (p=0.000 n=10+10) MakeChan/Struct/40-16 63.5ns ± 1% 61.1ns ± 2% -3.79% (p=0.000 n=10+9) ChanNonblocking-16 0.22ns ± 0% 0.22ns ± 0% +0.40% (p=0.011 n=9+8) SelectUncontended-16 4.63ns ± 1% 4.62ns ± 0% -0.35% (p=0.001 n=10+8) SelectSyncContended-16 1.58µs ± 2% 1.59µs ± 1% ~ (p=0.540 n=10+10) SelectAsyncContended-16 290ns ± 0% 291ns ± 0% +0.14% (p=0.012 n=8+9) SelectNonblock-16 0.95ns ± 1% 0.95ns ± 1% ~ (p=0.546 n=9+9) ChanUncontended-16 239ns ± 3% 242ns ± 6% ~ (p=0.886 n=9+10) ChanContended-16 17.7µs ± 1% 18.2µs ± 1% +2.87% (p=0.000 n=10+9) ChanSync-16 109ns ± 2% 109ns ± 1% ~ (p=0.342 n=10+10) ChanSyncWork-16 6.55µs ± 1% 6.53µs ± 1% ~ (p=0.101 n=10+10) ChanProdCons0-16 502ns ± 1% 499ns ± 0% -0.55% (p=0.001 n=10+9) ChanProdCons10-16 373ns ± 2% 377ns ± 1% ~ (p=0.095 n=10+9) ChanProdCons100-16 224ns ± 2% 223ns ± 3% ~ (p=0.150 n=9+10) ChanProdConsWork0-16 491ns ± 1% 484ns ± 0% -1.26% (p=0.000 n=10+9) ChanProdConsWork10-16 451ns ± 2% 448ns ± 2% ~ (p=0.210 n=8+10) ChanProdConsWork100-16 406ns ± 0% 407ns ± 1% ~ (p=0.138 n=8+8) SelectProdCons-16 509ns ± 0% 509ns ± 0% ~ (p=0.917 n=9+9) ReceiveDataFromClosedChan-16 12.1ns ± 0% 12.1ns ± 0% ~ (p=0.780 n=10+10) ChanCreation-16 22.6ns ± 1% 22.4ns ± 0% -0.72% (p=0.001 n=10+8) ChanSem-16 165ns ± 1% 166ns ± 1% +0.72% (p=0.002 n=10+10) ChanPopular-16 500µs ± 2% 498µs ± 1% ~ (p=0.218 n=10+10) ChanClosed-16 0.29ns ± 0% 0.29ns ± 0% +0.09% (p=0.019 n=9+8) CallClosure-16 1.28ns ± 0% 1.27ns ± 0% -0.51% (p=0.000 n=9+9) CallClosure1-16 1.50ns ± 0% 1.50ns ± 0% ~ (p=0.123 n=9+9) CallClosure2-16 8.86ns ± 1% 8.86ns ± 3% ~ (p=0.590 n=9+10) CallClosure3-16 8.75ns ± 2% 8.69ns ± 2% ~ (p=0.247 n=10+10) CallClosure4-16 8.65ns ± 2% 8.56ns ± 2% ~ (p=0.105 n=10+10) Complex128DivNormal-16 2.47ns ± 0% 2.47ns ± 0% ~ (p=0.790 n=10+9) Complex128DivNisNaN-16 4.44ns ± 0% 4.43ns ± 0% ~ (p=0.564 n=10+10) Complex128DivDisNaN-16 4.48ns ± 0% 4.48ns ± 0% ~ (p=0.101 n=10+10) Complex128DivNisInf-16 2.58ns ± 0% 2.58ns ± 0% ~ (p=0.808 n=10+10) Complex128DivDisInf-16 6.30ns ± 0% 6.31ns ± 0% ~ (p=0.305 n=10+10) SetTypePtr-16 0.73ns ± 1% 0.73ns ± 3% ~ (p=0.644 n=10+10) SetTypePtr8-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.127 n=10+10) SetTypePtr16-16 4.13ns ± 1% 4.12ns ± 0% ~ (p=0.109 n=10+10) SetTypePtr32-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.203 n=9+10) SetTypePtr64-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.696 n=10+10) SetTypePtr126-16 6.91ns ± 0% 6.91ns ± 0% ~ (p=0.469 n=10+10) SetTypePtr128-16 6.66ns ± 0% 6.67ns ± 0% ~ (p=0.246 n=9+10) SetTypePtrSlice-16 54.1ns ± 1% 54.1ns ± 1% ~ (p=0.509 n=9+10) SetTypeNode1-16 4.13ns ± 1% 4.12ns ± 0% ~ (p=0.342 n=10+10) SetTypeNode1Slice-16 10.1ns ± 1% 10.0ns ± 1% -1.18% (p=0.000 n=10+10) SetTypeNode8-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.137 n=8+8) SetTypeNode8Slice-16 22.6ns ± 0% 22.6ns ± 0% ~ (p=0.423 n=10+10) SetTypeNode64-16 6.90ns ± 0% 6.91ns ± 0% ~ (p=0.275 n=10+10) SetTypeNode64Slice-16 173ns ± 0% 173ns ± 0% ~ (p=0.610 n=9+10) SetTypeNode64Dead-16 5.53ns ± 0% 5.52ns ± 0% ~ (p=0.123 n=10+6) SetTypeNode64DeadSlice-16 150ns ± 0% 150ns ± 0% ~ (p=0.398 n=10+10) SetTypeNode124-16 6.90ns ± 0% 6.90ns ± 0% ~ (p=0.779 n=10+10) SetTypeNode124Slice-16 222ns ± 5% 217ns ± 0% ~ (p=0.302 n=10+10) SetTypeNode126-16 6.66ns ± 0% 6.66ns ± 0% ~ (p=0.324 n=10+9) SetTypeNode126Slice-16 218ns ± 0% 218ns ± 0% ~ (p=0.119 n=9+10) SetTypeNode128-16 9.76ns ± 0% 9.73ns ± 0% -0.31% (p=0.003 n=9+10) SetTypeNode128Slice-16 279ns ± 0% 278ns ± 0% ~ (p=0.112 n=10+9) SetTypeNode130-16 9.77ns ± 0% 9.73ns ± 0% -0.33% (p=0.002 n=10+10) SetTypeNode130Slice-16 284ns ± 0% 284ns ± 0% ~ (p=0.668 n=10+10) SetTypeNode1024-16 51.2ns ± 0% 51.6ns ± 1% ~ (p=0.080 n=9+9) SetTypeNode1024Slice-16 1.83µs ± 0% 1.82µs ± 0% ~ (p=0.115 n=10+10) Allocation-16 4.64µs ± 1% 4.37µs ± 1% -5.69% (p=0.000 n=9+9) ReadMemStats-16 5.62µs ± 2% 5.55µs ± 5% -1.36% (p=0.050 n=10+10) WriteBarrier-16 4.95ns ± 3% 4.99ns ± 3% ~ (p=0.255 n=10+10) BulkWriteBarrier-16 1.69ns ± 2% 1.63ns ± 4% -3.77% (p=0.001 n=10+10) ScanStackNoLocals-16 12.8ms ± 2% 12.9ms ± 1% +0.72% (p=0.019 n=10+10) MSpanCountAlloc/bits=64-16 1.65ns ± 0% 1.65ns ± 0% ~ (p=0.124 n=10+10) MSpanCountAlloc/bits=128-16 2.08ns ± 1% 2.06ns ± 1% -0.87% (p=0.000 n=10+10) MSpanCountAlloc/bits=256-16 2.71ns ± 1% 2.69ns ± 1% -0.74% (p=0.001 n=10+9) MSpanCountAlloc/bits=512-16 4.15ns ± 0% 4.23ns ± 2% +2.15% (p=0.000 n=10+10) MSpanCountAlloc/bits=1024-16 7.89ns ± 1% 7.89ns ± 1% ~ (p=0.867 n=10+10) Hash5-16 1.93ns ± 1% 2.01ns ± 0% +3.99% (p=0.000 n=10+8) Hash16-16 2.04ns ± 1% 2.21ns ± 1% +8.61% (p=0.000 n=10+10) Hash64-16 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.154 n=9+9) Hash1024-16 16.4ns ± 0% 16.4ns ± 0% +0.17% (p=0.020 n=9+10) Hash65536-16 886ns ± 0% 885ns ± 0% ~ (p=0.725 n=10+10) AlignedLoad-16 0.96ns ± 2% 0.95ns ± 3% ~ (p=0.123 n=10+10) UnalignedLoad-16 0.95ns ± 2% 1.01ns ± 2% +6.31% (p=0.000 n=10+10) EqEfaceConcrete-16 0.31ns ± 3% 0.33ns ± 5% +8.10% (p=0.000 n=10+10) EqIfaceConcrete-16 0.31ns ±13% 0.28ns ± 2% -9.23% (p=0.001 n=10+10) NeEfaceConcrete-16 0.29ns ± 1% 0.31ns ± 7% +5.59% (p=0.010 n=8+8) NeIfaceConcrete-16 0.28ns ± 2% 0.29ns ± 1% +4.49% (p=0.000 n=9+8) ConvT2EByteSized/bool-16 0.53ns ± 1% 0.52ns ± 1% -2.18% (p=0.000 n=10+10) ConvT2EByteSized/uint8-16 0.53ns ± 1% 0.53ns ± 0% +1.22% (p=0.000 n=10+10) ConvT2ESmall-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.774 n=9+9) ConvT2EUintptr-16 1.03ns ± 0% 1.04ns ± 0% +0.50% (p=0.000 n=10+8) ConvT2ELarge-16 14.4ns ± 2% 14.4ns ± 1% ~ (p=0.726 n=10+10) ConvT2ISmall-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.693 n=9+10) ConvT2IUintptr-16 1.03ns ± 0% 1.03ns ± 0% +0.44% (p=0.000 n=10+10) ConvT2ILarge-16 14.2ns ± 1% 14.4ns ± 1% +0.85% (p=0.007 n=9+10) ConvI2E-16 0.54ns ± 1% 0.54ns ± 0% -0.39% (p=0.037 n=10+8) ConvI2I-16 2.68ns ± 0% 2.70ns ± 1% +0.73% (p=0.000 n=9+10) AssertE2T-16 0.28ns ± 1% 0.39ns ± 5% +37.38% (p=0.000 n=10+10) AssertE2TLarge-16 0.42ns ± 2% 0.48ns ± 1% +14.92% (p=0.000 n=9+10) AssertE2I-16 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.352 n=9+9) AssertI2T-16 0.37ns ± 3% 0.34ns ± 1% -6.16% (p=0.000 n=10+10) AssertI2I-16 2.67ns ± 0% 2.67ns ± 0% ~ (p=0.286 n=10+10) AssertI2E-16 0.54ns ± 1% 0.54ns ± 0% -0.94% (p=0.000 n=10+10) AssertE2E-16 0.41ns ± 0% 0.41ns ± 1% ~ (p=0.880 n=9+9) AssertE2T2-16 0.41ns ± 1% 0.41ns ± 1% ~ (p=0.725 n=10+10) AssertE2T2Blank-16 0.24ns ± 5% 0.21ns ± 1% -14.79% (p=0.000 n=10+9) AssertI2E2-16 0.69ns ± 0% 0.69ns ± 1% ~ (p=0.541 n=10+10) AssertI2E2Blank-16 0.26ns ± 9% 0.21ns ± 1% -18.86% (p=0.000 n=10+9) AssertE2E2-16 0.53ns ± 1% 0.53ns ± 1% +0.72% (p=0.004 n=10+10) AssertE2E2Blank-16 0.23ns ± 4% 0.21ns ± 1% -8.42% (p=0.000 n=10+10) ConvT2Ezero/zero/16-16 1.13ns ± 0% 1.14ns ± 1% ~ (p=0.583 n=9+10) ConvT2Ezero/zero/32-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.417 n=10+10) ConvT2Ezero/zero/64-16 1.03ns ± 1% 1.03ns ± 0% ~ (p=0.051 n=10+10) ConvT2Ezero/zero/str-16 1.03ns ± 0% 1.03ns ± 0% ~ (p=0.132 n=10+10) ConvT2Ezero/zero/slice-16 1.14ns ± 0% 1.15ns ± 0% +0.49% (p=0.001 n=10+10) ConvT2Ezero/zero/big-16 123ns ± 1% 123ns ± 1% ~ (p=0.171 n=10+10) ConvT2Ezero/nonzero/str-16 19.4ns ± 1% 19.3ns ± 3% ~ (p=0.548 n=9+10) ConvT2Ezero/nonzero/slice-16 22.2ns ± 2% 22.0ns ± 2% ~ (p=0.109 n=10+10) ConvT2Ezero/nonzero/big-16 123ns ± 1% 123ns ± 1% ~ (p=0.446 n=10+8) ConvT2Ezero/smallint/16-16 1.13ns ± 0% 1.14ns ± 1% ~ (p=0.362 n=10+10) ConvT2Ezero/smallint/32-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.907 n=10+9) ConvT2Ezero/smallint/64-16 1.04ns ± 0% 1.03ns ± 0% -0.38% (p=0.002 n=10+10) ConvT2Ezero/largeint/16-16 6.65ns ± 1% 6.65ns ± 2% ~ (p=0.618 n=10+9) ConvT2Ezero/largeint/32-16 6.75ns ± 3% 6.63ns ± 2% -1.77% (p=0.015 n=10+10) ConvT2Ezero/largeint/64-16 9.19ns ± 1% 9.26ns ± 2% ~ (p=0.123 n=10+10) Malloc8-16 8.66ns ± 1% 8.89ns ± 2% +2.74% (p=0.000 n=10+10) Malloc16-16 13.7ns ± 1% 13.8ns ± 1% +0.71% (p=0.022 n=10+8) MallocTypeInfo8-16 11.7ns ± 3% 11.6ns ± 2% ~ (p=0.469 n=10+10) MallocTypeInfo16-16 18.3ns ± 1% 18.2ns ± 2% ~ (p=0.251 n=9+10) MallocLargeStruct-16 195ns ± 1% 198ns ± 1% +1.65% (p=0.000 n=9+10) GoroutineSelect-16 1.10ms ± 1% 1.12ms ± 1% +1.36% (p=0.000 n=10+8) GoroutineBlocking-16 986µs ± 1% 998µs ± 1% +1.23% (p=0.002 n=10+10) GoroutineForRange-16 985µs ± 1% 1001µs ± 1% +1.68% (p=0.000 n=10+10) GoroutineIdle-16 679µs ± 1% 691µs ± 0% +1.74% (p=0.000 n=10+9) HashStringSpeed-16 5.33ns ± 5% 5.19ns ± 4% ~ (p=0.113 n=9+9) HashBytesSpeed-16 8.20ns ± 3% 8.24ns ± 1% ~ (p=0.497 n=10+9) HashInt32Speed-16 4.01ns ± 2% 3.90ns ± 4% -2.63% (p=0.011 n=9+10) HashInt64Speed-16 3.94ns ± 4% 3.79ns ± 1% -3.74% (p=0.003 n=10+9) HashStringArraySpeed-16 12.5ns ± 4% 12.3ns ± 1% ~ (p=0.055 n=10+10) MegMap-16 3.72ns ± 1% 3.73ns ± 1% ~ (p=0.484 n=9+10) MegOneMap-16 2.28ns ± 1% 2.27ns ± 1% ~ (p=0.287 n=10+10) MegEqMap-16 22.0µs ± 3% 22.3µs ± 2% +1.48% (p=0.028 n=10+9) MegEmptyMap-16 0.93ns ± 1% 0.92ns ± 1% -0.52% (p=0.030 n=10+10) SmallStrMap-16 3.77ns ± 0% 3.77ns ± 0% ~ (p=0.324 n=10+10) MapStringKeysEight_16-16 3.91ns ± 0% 3.91ns ± 0% ~ (p=0.088 n=9+9) MapStringKeysEight_32-16 3.58ns ± 1% 3.50ns ± 0% -2.11% (p=0.000 n=10+10) MapStringKeysEight_64-16 3.58ns ± 1% 3.50ns ± 0% -2.23% (p=0.000 n=10+10) MapStringKeysEight_1M-16 3.57ns ± 1% 3.50ns ± 0% -1.92% (p=0.000 n=10+10) IntMap-16 2.89ns ± 1% 2.89ns ± 0% ~ (p=0.381 n=10+10) MapFirst/1-16 1.60ns ± 1% 1.59ns ± 2% -0.49% (p=0.020 n=10+9) MapFirst/2-16 1.61ns ± 0% 1.59ns ± 1% -1.17% (p=0.001 n=10+10) MapFirst/3-16 1.61ns ± 1% 1.59ns ± 1% -1.45% (p=0.000 n=10+10) MapFirst/4-16 1.60ns ± 1% 1.59ns ± 1% -1.16% (p=0.000 n=10+10) MapFirst/5-16 1.60ns ± 1% 1.58ns ± 1% -0.98% (p=0.000 n=10+10) MapFirst/6-16 1.60ns ± 1% 1.59ns ± 1% -0.87% (p=0.001 n=10+10) MapFirst/7-16 1.60ns ± 1% 1.59ns ± 1% -0.79% (p=0.002 n=10+10) MapFirst/8-16 1.60ns ± 1% 1.59ns ± 1% -0.67% (p=0.017 n=9+10) MapFirst/9-16 2.83ns ± 0% 2.83ns ± 0% ~ (p=0.492 n=10+10) MapFirst/10-16 2.83ns ± 0% 2.84ns ± 0% +0.24% (p=0.017 n=10+10) MapFirst/11-16 2.83ns ± 0% 2.83ns ± 0% ~ (p=0.445 n=10+10) MapFirst/12-16 2.83ns ± 0% 2.83ns ± 0% ~ (p=0.564 n=10+10) MapFirst/13-16 2.83ns ± 0% 2.84ns ± 0% ~ (p=0.175 n=9+10) MapFirst/14-16 2.83ns ± 0% 2.83ns ± 0% ~ (p=0.322 n=10+9) MapFirst/15-16 2.83ns ± 0% 2.84ns ± 1% ~ (p=0.209 n=10+10) MapFirst/16-16 2.83ns ± 1% 2.84ns ± 0% ~ (p=0.238 n=10+10) MapMid/1-16 1.64ns ± 0% 1.64ns ± 0% ~ (p=0.453 n=10+9) MapMid/2-16 1.86ns ± 1% 1.86ns ± 0% ~ (p=0.764 n=10+9) MapMid/3-16 1.86ns ± 0% 1.86ns ± 1% ~ (p=1.000 n=10+10) MapMid/4-16 2.06ns ± 0% 2.06ns ± 0% -0.27% (p=0.014 n=10+9) MapMid/5-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.075 n=9+10) MapMid/6-16 2.27ns ± 0% 2.27ns ± 1% ~ (p=0.898 n=10+10) MapMid/7-16 2.27ns ± 1% 2.26ns ± 0% -0.23% (p=0.049 n=10+10) MapMid/8-16 2.47ns ± 0% 2.47ns ± 1% ~ (p=0.840 n=10+10) MapMid/9-16 4.21ns ± 7% 4.13ns ±19% ~ (p=0.315 n=10+10) MapMid/10-16 4.17ns ± 7% 4.31ns ± 5% +3.37% (p=0.021 n=10+9) MapMid/11-16 4.18ns ± 7% 4.32ns ± 6% +3.50% (p=0.015 n=10+10) MapMid/12-16 4.34ns ± 7% 4.30ns ± 5% ~ (p=0.858 n=9+10) MapMid/13-16 4.25ns ± 6% 4.28ns ± 6% ~ (p=0.489 n=9+9) MapMid/14-16 3.75ns ±23% 3.90ns ±16% ~ (p=0.353 n=10+10) MapMid/15-16 3.87ns ±25% 3.95ns ±26% ~ (p=0.315 n=10+10) MapMid/16-16 4.06ns ±19% 3.94ns ±16% ~ (p=0.796 n=10+10) MapLast/1-16 1.65ns ± 0% 1.65ns ± 0% ~ (p=0.607 n=10+10) MapLast/2-16 1.86ns ± 0% 1.86ns ± 0% +0.26% (p=0.029 n=10+10) MapLast/3-16 2.06ns ± 1% 2.06ns ± 0% ~ (p=0.689 n=8+9) MapLast/4-16 2.27ns ± 1% 2.26ns ± 0% ~ (p=0.148 n=10+9) MapLast/5-16 2.47ns ± 0% 2.47ns ± 0% ~ (p=0.385 n=9+10) MapLast/6-16 2.67ns ± 0% 2.68ns ± 0% ~ (p=0.202 n=9+10) MapLast/7-16 2.88ns ± 0% 2.88ns ± 0% ~ (p=0.751 n=10+10) MapLast/8-16 3.08ns ± 0% 3.08ns ± 0% ~ (p=0.826 n=10+9) MapLast/9-16 4.31ns ± 6% 4.54ns ± 5% ~ (p=0.070 n=9+8) MapLast/10-16 4.25ns ± 5% 4.42ns ± 6% ~ (p=0.321 n=9+8) MapLast/11-16 4.59ns ±16% 5.42ns ±44% +17.99% (p=0.019 n=10+10) MapLast/12-16 5.04ns ±19% 6.11ns ±28% +21.35% (p=0.005 n=9+10) MapLast/13-16 6.00ns ±35% 5.76ns ± 3% ~ (p=0.173 n=10+8) MapLast/14-16 4.27ns ± 5% 4.53ns ± 6% +6.14% (p=0.007 n=10+10) MapLast/15-16 4.41ns ± 1% 4.44ns ± 7% ~ (p=0.515 n=8+10) MapLast/16-16 4.18ns ± 6% 4.99ns ±18% +19.48% (p=0.000 n=10+10) MapCycle-16 7.48ns ± 2% 7.46ns ± 1% ~ (p=0.699 n=10+10) RepeatedLookupStrMapKey32-16 6.98ns ± 3% 6.73ns ± 2% -3.63% (p=0.000 n=10+10) RepeatedLookupStrMapKey1M-16 14.7µs ± 5% 14.7µs ± 4% ~ (p=0.604 n=9+10) MakeMap/[Byte]Byte-16 58.5ns ± 1% 58.5ns ± 1% ~ (p=0.780 n=10+9) MakeMap/[Int]Int-16 113ns ± 0% 113ns ± 1% ~ (p=0.100 n=8+10) NewEmptyMap-16 2.47ns ± 0% 2.47ns ± 0% ~ (p=0.638 n=10+10) NewSmallMap-16 11.5ns ± 1% 11.6ns ± 0% +1.18% (p=0.000 n=10+10) MapIter-16 42.2ns ± 0% 42.8ns ± 1% +1.50% (p=0.000 n=10+10) MapIterEmpty-16 1.85ns ± 0% 1.85ns ± 0% ~ (p=0.651 n=10+10) SameLengthMap-16 1.85ns ± 1% 1.85ns ± 0% ~ (p=0.247 n=10+10) BigKeyMap-16 7.18ns ± 1% 7.42ns ± 4% +3.33% (p=0.004 n=10+10) BigValMap-16 7.03ns ± 2% 7.19ns ± 1% +2.33% (p=0.000 n=10+9) SmallKeyMap-16 5.32ns ± 1% 5.24ns ± 1% -1.41% (p=0.000 n=10+10) MapPopulate/1-16 6.30ns ± 0% 6.41ns ± 1% +1.81% (p=0.000 n=8+10) MapPopulate/10-16 239ns ± 2% 234ns ± 2% -2.05% (p=0.001 n=9+10) MapPopulate/100-16 4.19µs ± 2% 4.22µs ± 2% ~ (p=0.171 n=10+10) MapPopulate/1000-16 52.3µs ± 1% 52.5µs ± 1% ~ (p=0.133 n=9+10) MapPopulate/10000-16 459µs ± 1% 466µs ± 2% +1.45% (p=0.005 n=10+10) MapPopulate/100000-16 4.22ms ± 2% 4.25ms ± 2% ~ (p=0.393 n=10+10) ComplexAlgMap-16 12.5ns ± 1% 12.4ns ± 1% -0.95% (p=0.022 n=10+10) GoMapClear/Reflexive/1-16 9.61ns ± 1% 9.58ns ± 0% -0.27% (p=0.027 n=10+10) GoMapClear/Reflexive/10-16 10.0ns ± 1% 10.0ns ± 1% ~ (p=0.648 n=9+9) GoMapClear/Reflexive/100-16 31.4ns ± 0% 31.4ns ± 1% ~ (p=0.305 n=9+10) GoMapClear/Reflexive/1000-16 147ns ± 0% 149ns ± 2% +1.21% (p=0.000 n=10+10) GoMapClear/Reflexive/10000-16 3.99µs ± 0% 4.00µs ± 0% +0.21% (p=0.018 n=9+10) GoMapClear/NonReflexive/1-16 41.4ns ± 2% 41.7ns ± 1% +0.55% (p=0.043 n=9+10) GoMapClear/NonReflexive/10-16 50.3ns ± 1% 50.9ns ± 1% +1.16% (p=0.000 n=10+10) GoMapClear/NonReflexive/100-16 125ns ± 0% 126ns ± 0% +0.96% (p=0.000 n=8+10) GoMapClear/NonReflexive/1000-16 1.08µs ± 0% 1.08µs ± 1% ~ (p=0.097 n=10+10) GoMapClear/NonReflexive/10000-16 8.18µs ± 2% 8.10µs ± 0% -0.91% (p=0.019 n=10+8) MapStringConversion/32/simple-16 4.66ns ± 1% 4.69ns ± 3% ~ (p=0.905 n=9+10) MapStringConversion/32/struct-16 4.65ns ± 3% 4.94ns ± 2% +6.23% (p=0.000 n=10+10) MapStringConversion/32/array-16 4.69ns ± 3% 4.72ns ± 3% ~ (p=0.631 n=10+10) MapStringConversion/64/simple-16 4.14ns ± 0% 4.14ns ± 1% ~ (p=0.342 n=10+10) MapStringConversion/64/struct-16 4.13ns ± 0% 4.13ns ± 0% ~ (p=0.809 n=10+10) MapStringConversion/64/array-16 4.13ns ± 1% 4.13ns ± 1% ~ (p=0.752 n=10+10) MapInterfaceString-16 7.90ns ±23% 8.51ns ±33% ~ (p=0.604 n=9+10) MapInterfacePtr-16 7.68ns ±29% 7.10ns ±36% ~ (p=0.353 n=10+10) NewEmptyMapHintLessThan8-16 3.70ns ± 0% 3.70ns ± 0% ~ (p=0.209 n=10+10) NewEmptyMapHintGreaterThan8-16 270ns ± 1% 272ns ± 1% +0.71% (p=0.005 n=10+9) MapPop100-16 6.45µs ± 0% 6.50µs ± 1% +0.77% (p=0.000 n=10+10) MapPop1000-16 114µs ± 1% 114µs ± 1% ~ (p=0.190 n=10+10) MapPop10000-16 2.28ms ± 2% 2.28ms ± 2% ~ (p=0.912 n=10+10) MapAssign/Int32/256-16 4.75ns ± 2% 4.82ns ± 4% ~ (p=0.101 n=10+10) MapAssign/Int32/65536-16 16.4ns ± 1% 16.7ns ± 0% +1.44% (p=0.000 n=10+9) MapAssign/Int64/256-16 4.79ns ± 5% 4.79ns ± 1% ~ (p=0.616 n=10+8) MapAssign/Int64/65536-16 17.1ns ± 1% 16.8ns ± 0% -1.28% (p=0.000 n=10+8) MapAssign/Str/256-16 6.07ns ± 6% 6.24ns ± 2% +2.84% (p=0.035 n=10+9) MapAssign/Str/65536-16 21.4ns ± 0% 21.4ns ± 3% ~ (p=0.300 n=7+10) MapOperatorAssign/Int32/256-16 4.82ns ± 3% 4.81ns ± 3% ~ (p=0.684 n=10+10) MapOperatorAssign/Int32/65536-16 16.8ns ± 1% 16.5ns ± 1% -1.68% (p=0.000 n=9+10) MapOperatorAssign/Int64/256-16 4.74ns ± 1% 4.77ns ± 3% ~ (p=0.563 n=10+9) MapOperatorAssign/Int64/65536-16 16.9ns ± 1% 17.2ns ± 1% +1.88% (p=0.000 n=10+10) MapOperatorAssign/Str/256-16 1.09µs ± 1% 1.10µs ± 2% ~ (p=0.210 n=10+10) MapOperatorAssign/Str/65536-16 184ns ± 9% 184ns ± 8% ~ (p=0.922 n=10+9) MapAppendAssign/Int32/256-16 13.8ns ±10% 14.4ns ±11% ~ (p=0.190 n=10+10) MapAppendAssign/Int32/65536-16 28.9ns ± 5% 30.7ns ± 6% +6.13% (p=0.001 n=9+10) MapAppendAssign/Int64/256-16 14.5ns ±12% 13.8ns ± 8% -5.02% (p=0.037 n=10+10) MapAppendAssign/Int64/65536-16 30.9ns ± 1% 30.4ns ± 2% -1.56% (p=0.001 n=10+10) MapAppendAssign/Str/256-16 30.2ns ± 6% 30.0ns ±10% ~ (p=0.645 n=10+10) MapAppendAssign/Str/65536-16 44.5ns ± 4% 46.8ns ± 3% +5.17% (p=0.001 n=8+9) MapDelete/Int32/100-16 18.7ns ± 0% 18.7ns ± 0% -0.27% (p=0.017 n=10+10) MapDelete/Int32/1000-16 17.6ns ± 1% 17.5ns ± 1% -0.85% (p=0.000 n=9+10) MapDelete/Int32/10000-16 18.7ns ± 0% 18.3ns ± 1% -1.92% (p=0.000 n=10+10) MapDelete/Int64/100-16 19.1ns ± 0% 19.2ns ± 0% +0.68% (p=0.000 n=10+9) MapDelete/Int64/1000-16 17.7ns ± 2% 18.3ns ± 1% +3.00% (p=0.000 n=10+10) MapDelete/Int64/10000-16 18.8ns ± 1% 19.2ns ± 0% +2.01% (p=0.000 n=10+9) MapDelete/Str/100-16 26.5ns ± 0% 26.4ns ± 1% -0.73% (p=0.000 n=10+10) MapDelete/Str/1000-16 23.5ns ± 2% 23.4ns ± 1% ~ (p=0.425 n=10+10) MapDelete/Str/10000-16 25.1ns ± 0% 25.1ns ± 1% +0.28% (p=0.037 n=10+10) MapDelete/Pointer/100-16 20.6ns ± 1% 20.6ns ± 0% ~ (p=0.117 n=10+10) MapDelete/Pointer/1000-16 19.2ns ± 1% 19.4ns ± 1% +0.97% (p=0.004 n=10+10) MapDelete/Pointer/10000-16 20.0ns ± 0% 20.1ns ± 1% +0.52% (p=0.022 n=10+10) Memmove/0-16 0.21ns ± 2% 0.21ns ± 1% ~ (p=0.671 n=10+10) Memmove/1-16 0.93ns ± 0% 0.93ns ± 0% +0.21% (p=0.034 n=10+10) Memmove/2-16 0.93ns ± 0% 0.93ns ± 0% ~ (p=0.101 n=10+10) Memmove/3-16 0.93ns ± 1% 0.93ns ± 1% +0.49% (p=0.004 n=10+10) Memmove/4-16 1.03ns ± 0% 1.03ns ± 0% ~ (p=0.260 n=10+10) Memmove/5-16 1.13ns ± 0% 1.13ns ± 0% +0.20% (p=0.034 n=10+10) Memmove/6-16 1.13ns ± 0% 1.13ns ± 1% ~ (p=0.126 n=10+10) Memmove/7-16 1.13ns ± 0% 1.13ns ± 1% +0.22% (p=0.028 n=10+10) Memmove/8-16 1.13ns ± 0% 1.13ns ± 0% ~ (p=0.545 n=9+10) Memmove/9-16 1.25ns ± 0% 1.35ns ± 0% +7.98% (p=0.000 n=10+10) Memmove/10-16 1.25ns ± 0% 1.35ns ± 0% +7.96% (p=0.000 n=9+9) Memmove/11-16 1.25ns ± 0% 1.35ns ± 0% +8.53% (p=0.000 n=10+9) Memmove/12-16 1.25ns ± 0% 1.35ns ± 1% +8.24% (p=0.000 n=10+10) Memmove/13-16 1.25ns ± 0% 1.34ns ± 0% +7.75% (p=0.000 n=10+10) Memmove/14-16 1.25ns ± 0% 1.35ns ± 1% +8.28% (p=0.000 n=10+9) Memmove/15-16 1.25ns ± 0% 1.35ns ± 0% +8.07% (p=0.000 n=10+9) Memmove/16-16 1.25ns ± 0% 1.35ns ± 1% +8.35% (p=0.000 n=9+10) Memmove/32-16 1.34ns ± 0% 1.36ns ± 1% +1.22% (p=0.000 n=10+10) Memmove/64-16 1.45ns ± 0% 1.64ns ± 0% +13.07% (p=0.000 n=10+9) Memmove/128-16 1.86ns ± 0% 2.02ns ± 0% +8.64% (p=0.000 n=10+10) Memmove/256-16 2.47ns ± 0% 2.49ns ± 1% +1.14% (p=0.000 n=10+10) Memmove/512-16 3.96ns ± 1% 3.96ns ± 0% ~ (p=0.182 n=10+10) Memmove/1024-16 5.90ns ± 1% 5.87ns ± 1% ~ (p=0.258 n=9+9) Memmove/2048-16 9.62ns ± 1% 9.62ns ± 2% ~ (p=0.963 n=8+9) Memmove/4096-16 16.4ns ± 0% 17.1ns ± 4% +4.19% (p=0.003 n=8+9) MemmoveOverlap/32-16 1.62ns ± 1% 1.68ns ± 1% +3.53% (p=0.000 n=10+10) MemmoveOverlap/64-16 1.64ns ± 0% 1.65ns ± 0% +0.29% (p=0.002 n=9+9) MemmoveOverlap/128-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.070 n=10+10) MemmoveOverlap/256-16 2.67ns ± 0% 2.67ns ± 0% +0.26% (p=0.012 n=10+10) MemmoveOverlap/512-16 6.20ns ±18% 5.74ns ± 0% ~ (p=0.645 n=10+8) MemmoveOverlap/1024-16 7.28ns ± 0% 7.30ns ± 0% +0.28% (p=0.006 n=8+10) MemmoveOverlap/2048-16 11.9ns ± 0% 12.0ns ± 1% +0.37% (p=0.014 n=9+9) MemmoveOverlap/4096-16 23.3ns ± 1% 23.1ns ± 1% -0.84% (p=0.000 n=8+10) MemmoveUnalignedDst/0-16 1.03ns ± 0% 1.03ns ± 0% +0.19% (p=0.007 n=10+10) MemmoveUnalignedDst/1-16 1.24ns ± 0% 1.25ns ± 1% +0.52% (p=0.022 n=10+10) MemmoveUnalignedDst/2-16 1.23ns ± 0% 1.23ns ± 0% ~ (p=0.051 n=10+10) MemmoveUnalignedDst/3-16 1.23ns ± 0% 1.23ns ± 0% +0.14% (p=0.006 n=9+9) MemmoveUnalignedDst/4-16 1.23ns ± 0% 1.24ns ± 1% +0.37% (p=0.004 n=10+10) MemmoveUnalignedDst/5-16 1.35ns ± 0% 1.35ns ± 0% ~ (p=0.075 n=10+10) MemmoveUnalignedDst/6-16 1.34ns ± 0% 1.34ns ± 0% ~ (p=0.779 n=10+10) MemmoveUnalignedDst/7-16 1.34ns ± 0% 1.34ns ± 0% ~ (p=1.000 n=10+10) MemmoveUnalignedDst/8-16 1.34ns ± 0% 1.35ns ± 1% +0.39% (p=0.024 n=10+10) MemmoveUnalignedDst/9-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.849 n=10+10) MemmoveUnalignedDst/10-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.255 n=10+10) MemmoveUnalignedDst/11-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.304 n=10+10) MemmoveUnalignedDst/12-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.672 n=10+10) MemmoveUnalignedDst/13-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.435 n=10+10) MemmoveUnalignedDst/14-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.340 n=10+10) MemmoveUnalignedDst/15-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.911 n=10+9) MemmoveUnalignedDst/16-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.074 n=10+10) MemmoveUnalignedDst/32-16 1.62ns ± 0% 1.63ns ± 0% ~ (p=0.059 n=10+10) MemmoveUnalignedDst/64-16 1.65ns ± 0% 1.65ns ± 0% ~ (p=0.234 n=10+10) MemmoveUnalignedDst/128-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.709 n=10+9) MemmoveUnalignedDst/256-16 3.69ns ± 0% 3.70ns ± 0% ~ (p=0.144 n=10+10) MemmoveUnalignedDst/512-16 4.15ns ± 1% 4.14ns ± 0% ~ (p=0.778 n=10+8) MemmoveUnalignedDst/1024-16 7.52ns ± 0% 7.53ns ± 1% ~ (p=0.650 n=9+9) MemmoveUnalignedDst/2048-16 12.9ns ± 0% 12.9ns ± 1% ~ (p=0.548 n=8+8) MemmoveUnalignedDst/4096-16 25.4ns ± 0% 25.4ns ± 0% ~ (p=0.947 n=9+9) MemmoveUnalignedDstOverlap/32-16 4.08ns ± 0% 4.09ns ± 0% ~ (p=0.360 n=10+10) MemmoveUnalignedDstOverlap/64-16 4.56ns ± 0% 4.56ns ± 0% ~ (p=0.705 n=10+9) MemmoveUnalignedDstOverlap/128-16 4.67ns ± 0% 4.67ns ± 0% ~ (p=0.397 n=10+10) MemmoveUnalignedDstOverlap/256-16 5.08ns ± 0% 5.08ns ± 0% ~ (p=0.159 n=10+9) MemmoveUnalignedDstOverlap/512-16 8.45ns ± 5% 8.19ns ± 0% -3.10% (p=0.021 n=10+9) MemmoveUnalignedDstOverlap/1024-16 9.55ns ± 0% 9.56ns ± 0% ~ (p=0.221 n=8+8) MemmoveUnalignedDstOverlap/2048-16 14.0ns ± 0% 14.0ns ± 1% ~ (p=0.200 n=10+9) MemmoveUnalignedDstOverlap/4096-16 26.5ns ± 0% 26.5ns ± 0% ~ (p=0.458 n=10+9) MemmoveUnalignedSrc/0-16 1.02ns ± 1% 0.99ns ± 1% -2.67% (p=0.000 n=10+9) MemmoveUnalignedSrc/1-16 1.13ns ± 0% 1.13ns ± 1% -0.25% (p=0.027 n=10+9) MemmoveUnalignedSrc/2-16 1.13ns ± 1% 1.13ns ± 0% -0.28% (p=0.012 n=10+9) MemmoveUnalignedSrc/3-16 1.24ns ± 1% 1.23ns ± 0% -0.25% (p=0.022 n=9+10) MemmoveUnalignedSrc/4-16 1.24ns ± 0% 1.23ns ± 1% ~ (p=0.118 n=9+10) MemmoveUnalignedSrc/5-16 1.34ns ± 0% 1.34ns ± 1% ~ (p=0.564 n=8+10) MemmoveUnalignedSrc/6-16 1.34ns ± 0% 1.34ns ± 0% -0.39% (p=0.000 n=10+10) MemmoveUnalignedSrc/7-16 1.34ns ± 0% 1.34ns ± 0% ~ (p=0.235 n=10+10) MemmoveUnalignedSrc/8-16 1.34ns ± 0% 1.34ns ± 0% -0.37% (p=0.002 n=10+9) MemmoveUnalignedSrc/9-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.579 n=10+9) MemmoveUnalignedSrc/10-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.534 n=10+9) MemmoveUnalignedSrc/11-16 1.44ns ± 0% 1.44ns ± 1% ~ (p=0.415 n=10+10) MemmoveUnalignedSrc/12-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.218 n=10+10) MemmoveUnalignedSrc/13-16 1.44ns ± 0% 1.44ns ± 1% ~ (p=0.693 n=10+10) MemmoveUnalignedSrc/14-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.901 n=10+10) MemmoveUnalignedSrc/15-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.379 n=10+10) MemmoveUnalignedSrc/16-16 1.44ns ± 1% 1.44ns ± 0% ~ (p=0.538 n=10+10) MemmoveUnalignedSrc/32-16 1.60ns ± 1% 1.60ns ± 0% ~ (p=0.491 n=10+10) MemmoveUnalignedSrc/64-16 1.65ns ± 0% 1.65ns ± 0% ~ (p=0.564 n=10+10) MemmoveUnalignedSrc/128-16 2.09ns ± 0% 2.09ns ± 0% ~ (p=0.497 n=10+9) MemmoveUnalignedSrc/256-16 2.70ns ± 0% 2.78ns ± 1% +2.81% (p=0.000 n=10+10) MemmoveUnalignedSrc/512-16 4.31ns ± 0% 4.30ns ± 0% -0.26% (p=0.031 n=8+9) MemmoveUnalignedSrc/1024-16 7.28ns ± 0% 7.21ns ± 1% -1.05% (p=0.000 n=8+10) MemmoveUnalignedSrc/2048-16 13.0ns ± 0% 13.0ns ± 0% ~ (p=0.180 n=9+8) MemmoveUnalignedSrc/4096-16 25.4ns ± 0% 25.3ns ± 1% ~ (p=0.054 n=10+10) MemmoveUnalignedSrcOverlap/32-16 4.04ns ± 0% 4.06ns ± 0% +0.62% (p=0.000 n=9+10) MemmoveUnalignedSrcOverlap/64-16 4.12ns ± 0% 4.12ns ± 0% ~ (p=0.421 n=10+10) MemmoveUnalignedSrcOverlap/128-16 4.53ns ± 0% 4.52ns ± 0% ~ (p=0.251 n=10+10) MemmoveUnalignedSrcOverlap/256-16 6.17ns ± 0% 6.15ns ± 0% -0.35% (p=0.000 n=10+9) MemmoveUnalignedSrcOverlap/512-16 7.43ns ± 0% 7.44ns ± 0% ~ (p=0.524 n=9+8) MemmoveUnalignedSrcOverlap/1024-16 8.94ns ± 0% 8.94ns ± 0% ~ (p=0.419 n=8+8) MemmoveUnalignedSrcOverlap/2048-16 13.2ns ± 0% 14.5ns ±21% ~ (p=0.107 n=8+10) MemmoveUnalignedSrcOverlap/4096-16 25.6ns ± 0% 25.6ns ± 1% ~ (p=0.650 n=9+9) Memclr/5-16 0.86ns ± 1% 0.86ns ± 2% ~ (p=0.531 n=9+9) Memclr/16-16 1.04ns ± 0% 1.04ns ± 0% +0.32% (p=0.013 n=9+10) Memclr/64-16 1.23ns ± 0% 1.26ns ± 0% +2.28% (p=0.000 n=10+10) Memclr/256-16 2.27ns ± 0% 2.27ns ± 0% ~ (p=0.127 n=10+10) Memclr/4096-16 17.1ns ± 1% 17.3ns ± 0% +0.88% (p=0.000 n=10+10) Memclr/65536-16 821ns ± 0% 822ns ± 0% ~ (p=0.516 n=10+10) Memclr/1M-16 14.1µs ± 1% 14.0µs ± 1% ~ (p=0.516 n=10+10) Memclr/4M-16 86.1µs ± 1% 85.9µs ± 0% ~ (p=0.123 n=10+10) Memclr/8M-16 174µs ± 2% 173µs ± 0% ~ (p=0.408 n=10+8) Memclr/16M-16 385µs ± 4% 387µs ± 0% ~ (p=0.173 n=10+8) Memclr/64M-16 2.18ms ± 0% 2.19ms ± 0% ~ (p=0.113 n=10+9) GoMemclr/5-16 0.82ns ± 0% 0.82ns ± 0% ~ (p=0.346 n=9+10) GoMemclr/16-16 1.02ns ± 0% 1.02ns ± 0% +0.22% (p=0.003 n=10+8) GoMemclr/64-16 1.14ns ± 0% 1.14ns ± 0% ~ (p=0.948 n=10+9) GoMemclr/256-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.868 n=10+10) MemclrRange/1K_2K-16 457ns ± 0% 428ns ± 1% -6.38% (p=0.000 n=10+10) MemclrRange/2K_8K-16 1.46µs ± 0% 1.46µs ± 0% ~ (p=0.700 n=10+10) MemclrRange/4K_16K-16 1.16µs ± 0% 1.16µs ± 0% ~ (p=0.567 n=9+10) MemclrRange/160K_228K-16 20.7µs ± 0% 20.7µs ± 0% ~ (p=0.160 n=10+10) ClearFat7-16 0.38ns ± 5% 0.21ns ± 1% -45.79% (p=0.000 n=9+10) ClearFat8-16 0.21ns ± 3% 0.12ns ± 2% -44.16% (p=0.000 n=8+9) ClearFat11-16 0.35ns ± 3% 0.21ns ± 1% -40.46% (p=0.000 n=9+9) ClearFat12-16 0.23ns ± 9% 0.21ns ± 1% -10.23% (p=0.000 n=10+9) ClearFat13-16 0.22ns ± 6% 0.21ns ± 2% -6.53% (p=0.000 n=10+10) ClearFat14-16 0.22ns ± 4% 0.21ns ± 1% -5.97% (p=0.000 n=10+10) ClearFat15-16 0.22ns ± 4% 0.21ns ± 1% -6.96% (p=0.000 n=10+9) ClearFat16-16 0.19ns ± 9% 0.12ns ± 6% -34.89% (p=0.000 n=9+10) ClearFat24-16 0.23ns ± 6% 0.21ns ± 1% -10.26% (p=0.000 n=10+9) ClearFat32-16 0.22ns ± 5% 0.21ns ± 2% -5.31% (p=0.000 n=10+10) ClearFat40-16 0.34ns ± 4% 0.62ns ± 1% +83.00% (p=0.000 n=10+10) ClearFat48-16 0.33ns ± 2% 0.41ns ± 0% +26.71% (p=0.000 n=10+10) ClearFat56-16 0.41ns ± 1% 0.41ns ± 0% ~ (p=0.838 n=10+10) ClearFat64-16 0.41ns ± 0% 0.41ns ± 0% ~ (p=0.178 n=10+8) ClearFat72-16 0.82ns ± 0% 0.82ns ± 0% ~ (p=0.669 n=10+10) ClearFat128-16 1.04ns ± 0% 1.04ns ± 0% ~ (p=0.679 n=10+10) ClearFat256-16 1.86ns ± 0% 1.86ns ± 0% ~ (p=0.066 n=9+10) ClearFat512-16 3.50ns ± 0% 3.50ns ± 0% ~ (p=0.626 n=10+10) ClearFat1024-16 6.79ns ± 0% 6.79ns ± 0% ~ (p=0.986 n=10+10) ClearFat1032-16 13.6ns ± 0% 13.6ns ± 0% +0.13% (p=0.044 n=10+10) ClearFat1040-16 10.3ns ± 0% 10.3ns ± 0% ~ (p=0.175 n=10+9) CopyFat7-16 0.37ns ±13% 0.25ns ± 1% -31.74% (p=0.000 n=10+9) CopyFat8-16 0.17ns ± 1% 0.17ns ± 2% +1.35% (p=0.004 n=9+9) CopyFat11-16 0.26ns ± 1% 0.30ns ± 3% +12.58% (p=0.000 n=9+10) CopyFat12-16 0.28ns ± 2% 0.26ns ± 1% -5.66% (p=0.000 n=9+9) CopyFat13-16 0.26ns ± 0% 0.28ns ± 4% +7.35% (p=0.000 n=8+10) CopyFat14-16 0.29ns ± 6% 0.26ns ± 2% -10.46% (p=0.000 n=10+9) CopyFat15-16 0.26ns ± 1% 0.30ns ± 6% +14.12% (p=0.000 n=8+10) CopyFat16-16 0.21ns ± 1% 0.21ns ± 0% ~ (p=0.426 n=8+8) CopyFat24-16 0.29ns ± 3% 0.25ns ± 1% -12.27% (p=0.000 n=9+10) CopyFat32-16 0.26ns ± 4% 0.29ns ± 4% +11.71% (p=0.000 n=10+10) CopyFat64-16 0.46ns ± 8% 0.42ns ± 1% -8.37% (p=0.002 n=10+10) CopyFat72-16 0.82ns ± 0% 0.82ns ± 0% ~ (p=0.563 n=10+10) CopyFat128-16 1.53ns ± 0% 1.54ns ± 0% +0.62% (p=0.000 n=10+10) CopyFat256-16 2.68ns ± 0% 2.65ns ± 1% -1.23% (p=0.000 n=10+10) CopyFat512-16 4.93ns ± 1% 5.19ns ± 3% +5.16% (p=0.000 n=9+9) CopyFat520-16 6.99ns ± 0% 6.99ns ± 0% ~ (p=0.539 n=10+10) CopyFat1024-16 11.5ns ± 1% 9.8ns ± 1% -14.98% (p=0.000 n=9+10) CopyFat1032-16 13.6ns ± 0% 13.6ns ± 0% ~ (p=0.728 n=10+10) CopyFat1040-16 11.0ns ± 0% 11.1ns ± 0% +0.53% (p=0.000 n=10+10) Issue18740/2byte-16 10.1µs ± 0% 10.1µs ± 0% ~ (p=0.342 n=10+10) Issue18740/4byte-16 2.34µs ± 0% 2.35µs ± 0% +0.30% (p=0.002 n=10+8) Issue18740/8byte-16 1.28µs ± 0% 1.28µs ± 0% +0.32% (p=0.000 n=9+10) Finalizer-16 345µs ± 1% 336µs ± 0% -2.55% (p=0.000 n=10+9) FinalizerRun-16 450ns ± 3% 420ns ± 1% -6.65% (p=0.000 n=10+10) PallocBitsSummarize/Unpacked00-16 2.88ns ± 0% 2.88ns ± 0% ~ (p=0.358 n=10+10) PallocBitsSummarize/UnpackedFFFFFFFFFFFFFFFF-16 15.2ns ± 0% 15.2ns ± 0% ~ (p=0.925 n=10+10) PallocBitsSummarize/UnpackedAA-16 16.4ns ± 0% 16.3ns ± 0% ~ (p=0.113 n=9+9) PallocBitsSummarize/UnpackedAAAAAAAAAAAAAAAA-16 16.5ns ± 0% 16.6ns ± 0% ~ (p=0.238 n=10+10) PallocBitsSummarize/Unpacked80000000AAAAAAAA-16 37.8ns ± 1% 36.4ns ± 0% -3.70% (p=0.000 n=10+9) PallocBitsSummarize/UnpackedAAAAAAAA00000001-16 41.8ns ± 1% 39.9ns ± 0% -4.68% (p=0.000 n=9+10) PallocBitsSummarize/UnpackedBBBBBBBBBBBBBBBB-16 18.3ns ± 0% 18.3ns ± 0% ~ (p=0.781 n=10+10) PallocBitsSummarize/Unpacked80000000BBBBBBBB-16 38.8ns ± 1% 38.1ns ± 0% -1.78% (p=0.000 n=9+10) PallocBitsSummarize/UnpackedBBBBBBBB00000001-16 37.5ns ± 0% 36.1ns ± 1% -3.88% (p=0.000 n=8+10) PallocBitsSummarize/UnpackedCCCCCCCCCCCCCCCC-16 21.8ns ± 0% 21.9ns ± 0% +0.20% (p=0.018 n=10+9) PallocBitsSummarize/Unpacked4444444444444444-16 21.8ns ± 0% 21.9ns ± 0% +0.20% (p=0.029 n=10+9) PallocBitsSummarize/Unpacked4040404040404040-16 26.5ns ± 0% 26.5ns ± 0% -0.24% (p=0.001 n=9+10) PallocBitsSummarize/Unpacked4000400040004000-16 33.4ns ± 1% 31.3ns ± 0% -6.20% (p=0.000 n=9+10) PallocBitsSummarize/Unpacked1000404044CCAAFF-16 36.4ns ± 1% 35.9ns ± 0% -1.50% (p=0.000 n=10+10) FindBitRange64/Pattern00Size2-16 0.34ns ± 1% 0.35ns ± 1% +3.80% (p=0.000 n=10+9) FindBitRange64/Pattern00Size8-16 0.70ns ± 1% 0.70ns ± 0% -0.68% (p=0.000 n=10+10) FindBitRange64/Pattern00Size32-16 0.70ns ± 1% 0.69ns ± 0% -0.86% (p=0.001 n=10+8) FindBitRange64/PatternFFFFFFFFFFFFFFFFSize2-16 0.34ns ± 1% 0.35ns ± 1% +4.45% (p=0.000 n=9+8) FindBitRange64/PatternFFFFFFFFFFFFFFFFSize8-16 1.54ns ± 0% 1.54ns ± 1% ~ (p=0.914 n=9+9) FindBitRange64/PatternFFFFFFFFFFFFFFFFSize32-16 2.78ns ± 0% 2.78ns ± 0% ~ (p=0.295 n=9+10) FindBitRange64/PatternAASize2-16 0.34ns ± 2% 0.35ns ± 2% +4.61% (p=0.000 n=10+10) FindBitRange64/PatternAASize8-16 0.70ns ± 1% 0.70ns ± 1% -0.82% (p=0.005 n=10+10) FindBitRange64/PatternAASize32-16 0.70ns ± 1% 0.70ns ± 0% -0.73% (p=0.003 n=10+9) FindBitRange64/PatternAAAAAAAAAAAAAAAASize2-16 0.34ns ± 2% 0.35ns ± 2% +3.94% (p=0.000 n=10+10) FindBitRange64/PatternAAAAAAAAAAAAAAAASize8-16 0.70ns ± 1% 0.70ns ± 1% -0.67% (p=0.025 n=10+10) FindBitRange64/PatternAAAAAAAAAAAAAAAASize32-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.118 n=9+10) FindBitRange64/Pattern80000000AAAAAAAASize2-16 0.34ns ± 1% 0.35ns ± 2% +3.72% (p=0.000 n=10+9) FindBitRange64/Pattern80000000AAAAAAAASize8-16 0.70ns ± 1% 0.70ns ± 0% ~ (p=0.102 n=10+10) FindBitRange64/Pattern80000000AAAAAAAASize32-16 0.70ns ± 1% 0.70ns ± 1% -0.55% (p=0.011 n=10+10) FindBitRange64/PatternAAAAAAAA00000001Size2-16 0.34ns ± 2% 0.35ns ± 1% +3.83% (p=0.000 n=10+9) FindBitRange64/PatternAAAAAAAA00000001Size8-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.065 n=10+10) FindBitRange64/PatternAAAAAAAA00000001Size32-16 0.70ns ± 1% 0.70ns ± 1% -0.95% (p=0.002 n=10+10) FindBitRange64/PatternBBBBBBBBBBBBBBBBSize2-16 0.34ns ± 0% 0.35ns ± 1% +4.12% (p=0.000 n=8+10) FindBitRange64/PatternBBBBBBBBBBBBBBBBSize8-16 1.24ns ± 0% 1.23ns ± 0% -0.30% (p=0.002 n=10+9) FindBitRange64/PatternBBBBBBBBBBBBBBBBSize32-16 1.24ns ± 0% 1.24ns ± 0% -0.17% (p=0.023 n=9+10) FindBitRange64/Pattern80000000BBBBBBBBSize2-16 0.34ns ± 1% 0.35ns ± 2% +4.82% (p=0.000 n=9+10) FindBitRange64/Pattern80000000BBBBBBBBSize8-16 1.24ns ± 1% 1.24ns ± 0% ~ (p=0.063 n=10+10) FindBitRange64/Pattern80000000BBBBBBBBSize32-16 1.24ns ± 0% 1.24ns ± 0% ~ (p=0.164 n=9+10) FindBitRange64/PatternBBBBBBBB00000001Size2-16 0.34ns ± 1% 0.35ns ± 1% +4.38% (p=0.000 n=8+10) FindBitRange64/PatternBBBBBBBB00000001Size8-16 1.24ns ± 1% 1.24ns ± 0% ~ (p=0.052 n=10+10) FindBitRange64/PatternBBBBBBBB00000001Size32-16 1.24ns ± 0% 1.23ns ± 0% -0.40% (p=0.000 n=10+10) FindBitRange64/PatternCCCCCCCCCCCCCCCCSize2-16 0.34ns ± 0% 0.35ns ± 2% +3.96% (p=0.000 n=9+10) FindBitRange64/PatternCCCCCCCCCCCCCCCCSize8-16 1.24ns ± 0% 1.23ns ± 0% -0.30% (p=0.000 n=10+9) FindBitRange64/PatternCCCCCCCCCCCCCCCCSize32-16 1.24ns ± 0% 1.24ns ± 1% ~ (p=0.284 n=10+10) FindBitRange64/Pattern4444444444444444Size2-16 0.34ns ± 1% 0.35ns ± 1% +3.91% (p=0.000 n=9+9) FindBitRange64/Pattern4444444444444444Size8-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.617 n=10+10) FindBitRange64/Pattern4444444444444444Size32-16 0.70ns ± 1% 0.70ns ± 1% -0.60% (p=0.006 n=10+10) FindBitRange64/Pattern4040404040404040Size2-16 0.34ns ± 2% 0.35ns ± 2% +3.67% (p=0.000 n=10+10) FindBitRange64/Pattern4040404040404040Size8-16 0.70ns ± 2% 0.70ns ± 1% -0.87% (p=0.014 n=10+10) FindBitRange64/Pattern4040404040404040Size32-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.256 n=10+10) FindBitRange64/Pattern4000400040004000Size2-16 0.34ns ± 2% 0.35ns ± 3% +4.71% (p=0.000 n=10+10) FindBitRange64/Pattern4000400040004000Size8-16 0.70ns ± 1% 0.70ns ± 1% ~ (p=0.393 n=10+10) FindBitRange64/Pattern4000400040004000Size32-16 0.70ns ± 1% 0.70ns ± 1% -0.86% (p=0.014 n=10+10) NetpollBreak-16 1.49µs ± 1% 1.50µs ± 3% ~ (p=0.181 n=8+10) Syscall-16 3.68ns ± 1% 3.66ns ± 2% ~ (p=0.148 n=10+10) SyscallWork-16 5.15ns ± 1% 5.13ns ± 0% ~ (p=0.188 n=10+9) SyscallExcess-16 3.89ns ± 2% 3.83ns ± 1% -1.52% (p=0.001 n=10+10) SyscallExcessWork-16 5.34ns ± 1% 5.31ns ± 0% -0.64% (p=0.000 n=10+9) PingPongHog-16 397ns ± 7% 394ns ±11% ~ (p=0.912 n=10+10) StackGrowth-16 67.9ns ± 0% 68.8ns ± 0% +1.28% (p=0.000 n=10+8) StackGrowthDeep-16 7.70µs ± 1% 8.48µs ± 2% +10.06% (p=0.000 n=9+10) CreateGoroutines-16 124ns ± 1% 124ns ± 1% ~ (p=0.254 n=10+10) CreateGoroutinesParallel-16 25.7ns ± 1% 27.6ns ± 2% +7.51% (p=0.000 n=10+10) CreateGoroutinesCapture-16 823ns ± 1% 821ns ± 2% ~ (p=0.699 n=10+10) CreateGoroutinesSingle-16 175ns ± 3% 172ns ± 3% -1.90% (p=0.011 n=10+10) ClosureCall-16 0.11ns ± 7% 0.12ns ± 3% ~ (p=0.842 n=9+10) WakeupParallelSpinning/0s-16 11.4µs ± 0% 11.4µs ± 0% ~ (p=0.325 n=9+10) WakeupParallelSpinning/1µs-16 15.4µs ± 0% 15.4µs ± 1% ~ (p=0.955 n=10+10) WakeupParallelSpinning/2µs-16 18.7µs ± 2% 18.9µs ± 2% ~ (p=0.052 n=10+10) WakeupParallelSpinning/5µs-16 30.7µs ± 0% 30.7µs ± 0% -0.03% (p=0.003 n=10+10) WakeupParallelSpinning/10µs-16 48.8µs ± 0% 48.8µs ± 0% ~ (p=0.670 n=10+10) WakeupParallelSpinning/20µs-16 90.8µs ± 0% 90.8µs ± 0% -0.02% (p=0.004 n=10+10) WakeupParallelSpinning/50µs-16 211µs ± 0% 211µs ± 0% ~ (p=0.194 n=10+10) WakeupParallelSpinning/100µs-16 323µs ± 0% 323µs ± 0% ~ (p=1.000 n=10+9) WakeupParallelSyscall/0s-16 118µs ± 0% 118µs ± 0% ~ (p=0.447 n=10+9) WakeupParallelSyscall/1µs-16 119µs ± 2% 119µs ± 1% ~ (p=0.604 n=10+9) WakeupParallelSyscall/2µs-16 120µs ± 1% 121µs ± 3% ~ (p=0.263 n=8+10) WakeupParallelSyscall/5µs-16 126µs ± 2% 126µs ± 2% ~ (p=0.510 n=10+9) WakeupParallelSyscall/10µs-16 136µs ± 1% 137µs ± 1% ~ (p=0.095 n=9+10) WakeupParallelSyscall/20µs-16 156µs ± 2% 157µs ± 3% ~ (p=0.604 n=10+9) WakeupParallelSyscall/50µs-16 221µs ± 1% 220µs ± 1% ~ (p=0.063 n=10+10) WakeupParallelSyscall/100µs-16 326µs ± 0% 325µs ± 0% -0.26% (p=0.003 n=9+10) Matmult-16 0.67ns ± 2% 0.66ns ± 2% ~ (p=0.256 n=10+10) Fastrand-16 0.08ns ±11% 0.08ns ±13% ~ (p=0.661 n=9+10) Fastrand64-16 0.08ns ±11% 0.08ns ± 6% ~ (p=0.631 n=10+10) FastrandHashiter-16 1.76ns ± 1% 1.76ns ± 1% ~ (p=0.854 n=8+8) Fastrandn/2-16 0.86ns ± 1% 0.86ns ± 1% +1.09% (p=0.000 n=10+9) Fastrandn/3-16 0.85ns ± 1% 0.86ns ± 1% +1.23% (p=0.001 n=10+10) Fastrandn/4-16 0.85ns ± 1% 0.87ns ± 2% +1.60% (p=0.000 n=10+10) Fastrandn/5-16 0.85ns ± 1% 0.86ns ± 1% +1.05% (p=0.000 n=10+10) IfaceCmp100-16 46.6ns ± 0% 46.1ns ± 0% -1.18% (p=0.000 n=10+10) IfaceCmpNil100-16 26.8ns ± 0% 26.8ns ± 0% ~ (p=0.777 n=10+8) EfaceCmpDiff-16 132ns ± 0% 130ns ± 0% -0.95% (p=0.000 n=10+9) EfaceCmpDiffIndirect-16 209ns ± 0% 211ns ± 0% +1.14% (p=0.000 n=10+9) Defer-16 3.40ns ± 1% 3.04ns ± 0% -10.67% (p=0.000 n=10+10) Defer10-16 29.4ns ± 2% 27.2ns ± 3% -7.26% (p=0.000 n=10+10) DeferMany-16 110ns ± 6% 113ns ± 2% +3.45% (p=0.017 n=9+9) PanicRecover-16 67.6ns ± 0% 67.7ns ± 2% ~ (p=0.436 n=9+9) GoroutineProfile/small-nil/idle-16 3.90µs ± 4% 3.86µs ± 2% ~ (p=0.305 n=10+9) GoroutineProfile/small-nil/loaded-16 4.82µs ± 6% 4.82µs ± 4% ~ (p=0.905 n=10+9) GoroutineProfile/small/idle-16 103µs ± 3% 102µs ± 3% ~ (p=0.113 n=9+9) GoroutineProfile/small/loaded-16 432µs ± 5% 440µs ±13% ~ (p=0.604 n=9+10) GoroutineProfile/large-nil/idle-16 3.86µs ± 3% 3.82µs ± 3% ~ (p=0.210 n=10+10) GoroutineProfile/large-nil/loaded-16 4.90µs ± 2% 4.90µs ± 5% ~ (p=0.780 n=10+9) GoroutineProfile/large/idle-16 2.58ms ± 1% 2.52ms ± 1% -2.38% (p=0.000 n=10+10) GoroutineProfile/large/loaded-16 8.62ms ± 9% 8.90ms ±11% ~ (p=0.400 n=9+10) GoroutineProfile/sparse-nil/idle-16 3.85µs ± 4% 3.81µs ± 3% ~ (p=0.470 n=10+10) GoroutineProfile/sparse-nil/loaded-16 4.82µs ± 4% 4.69µs ± 5% ~ (p=0.052 n=10+10) GoroutineProfile/sparse/idle-16 102µs ± 4% 102µs ± 2% ~ (p=0.497 n=10+9) GoroutineProfile/sparse/loaded-16 438µs ± 7% 437µs ± 6% ~ (p=0.796 n=10+10) RWMutexUncontended-16 6.79ns ± 0% 6.78ns ± 0% ~ (p=0.228 n=10+8) RWMutexWrite100-16 85.4ns ± 0% 87.1ns ± 0% +2.00% (p=0.000 n=10+8) RWMutexWrite10-16 168ns ±25% 152ns ±11% ~ (p=0.063 n=10+10) RWMutexWorkWrite100-16 106ns ± 0% 106ns ± 3% ~ (p=0.136 n=10+10) RWMutexWorkWrite10-16 567ns ± 3% 571ns ± 1% ~ (p=0.326 n=10+9) SemTable/OneAddrCollision/n=1000-16 15.9µs ± 1% 16.0µs ± 1% +0.50% (p=0.031 n=9+9) SemTable/ManyAddrCollision/n=1000-16 56.2µs ± 1% 56.8µs ± 1% +1.06% (p=0.000 n=10+10) SemTable/OneAddrCollision/n=2000-16 32.6µs ± 2% 32.9µs ± 4% ~ (p=0.156 n=10+9) SemTable/ManyAddrCollision/n=2000-16 118µs ± 0% 119µs ± 0% +0.75% (p=0.000 n=9+10) SemTable/OneAddrCollision/n=4000-16 65.3µs ± 1% 65.6µs ± 3% ~ (p=0.497 n=9+10) SemTable/ManyAddrCollision/n=4000-16 245µs ± 0% 248µs ± 2% +1.36% (p=0.000 n=9+10) SemTable/OneAddrCollision/n=8000-16 131µs ± 1% 130µs ± 1% -1.01% (p=0.002 n=9+10) SemTable/ManyAddrCollision/n=8000-16 503µs ± 1% 508µs ± 0% +0.97% (p=0.000 n=10+10) MakeSliceCopy/mallocmove/Byte-16 67.6ns ± 1% 64.1ns ± 2% -5.20% (p=0.000 n=10+10) MakeSliceCopy/mallocmove/Int-16 65.0ns ± 7% 61.7ns ± 4% -5.08% (p=0.009 n=10+10) MakeSliceCopy/mallocmove/Ptr-16 88.1ns ± 1% 79.9ns ± 1% -9.29% (p=0.000 n=10+10) MakeSliceCopy/makecopy/Byte-16 65.2ns ± 6% 63.4ns ± 0% ~ (p=0.500 n=10+8) MakeSliceCopy/makecopy/Int-16 63.2ns ± 1% 64.1ns ± 1% +1.34% (p=0.001 n=9+9) MakeSliceCopy/makecopy/Ptr-16 88.1ns ± 1% 80.1ns ± 1% -9.09% (p=0.000 n=10+10) MakeSliceCopy/nilappend/Byte-16 69.8ns ± 1% 65.7ns ± 3% -5.80% (p=0.000 n=10+10) MakeSliceCopy/nilappend/Int-16 69.6ns ± 2% 67.2ns ± 1% -3.50% (p=0.000 n=10+9) MakeSliceCopy/nilappend/Ptr-16 91.5ns ± 1% 83.8ns ± 1% -8.42% (p=0.000 n=9+10) MakeSlice/Byte-16 6.64ns ± 3% 6.58ns ± 2% ~ (p=0.393 n=10+10) MakeSlice/Int16-16 8.60ns ± 1% 8.38ns ± 3% -2.48% (p=0.001 n=9+10) MakeSlice/Int-16 17.7ns ± 3% 16.9ns ± 1% -4.67% (p=0.000 n=10+9) MakeSlice/Ptr-16 24.0ns ± 3% 23.3ns ± 2% -3.25% (p=0.000 n=10+9) MakeSlice/Struct/24-16 34.1ns ± 1% 32.0ns ± 1% -6.11% (p=0.000 n=10+10) MakeSlice/Struct/32-16 39.1ns ± 4% 38.2ns ± 1% ~ (p=0.829 n=10+8) MakeSlice/Struct/40-16 47.0ns ± 5% 43.0ns ± 2% -8.55% (p=0.000 n=10+9) GrowSlice/Byte-16 15.3ns ± 3% 15.0ns ± 2% -1.75% (p=0.005 n=9+9) GrowSlice/Int16-16 18.9ns ± 2% 18.4ns ± 2% -2.71% (p=0.000 n=10+9) GrowSlice/Int-16 33.9ns ± 1% 32.2ns ± 1% -4.89% (p=0.000 n=10+9) GrowSlice/Ptr-16 45.3ns ± 2% 43.5ns ± 1% -4.12% (p=0.000 n=10+10) GrowSlice/Struct/24-16 61.9ns ± 2% 60.0ns ± 4% -3.10% (p=0.002 n=10+10) GrowSlice/Struct/32-16 79.9ns ± 2% 72.3ns ± 3% -9.58% (p=0.000 n=8+10) GrowSlice/Struct/40-16 97.1ns ± 7% 88.8ns ± 5% -8.49% (p=0.000 n=10+10) ExtendSlice/IntSlice-16 21.1ns ± 2% 20.3ns ± 2% -3.71% (p=0.000 n=10+10) ExtendSlice/PointerSlice-16 26.8ns ± 2% 26.3ns ± 2% -1.86% (p=0.004 n=10+10) ExtendSlice/NoGrow-16 1.23ns ± 0% 1.30ns ± 1% +5.03% (p=0.000 n=10+10) Append-16 4.58ns ± 1% 4.53ns ± 0% -1.11% (p=0.000 n=10+10) AppendGrowByte-16 1.46ms ± 8% 1.42ms ± 7% -3.24% (p=0.035 n=10+10) AppendGrowString-16 27.8ms ± 4% 27.2ms ± 5% ~ (p=0.052 n=10+10) AppendSlice/1Bytes-16 1.03ns ± 1% 1.04ns ± 1% ~ (p=0.303 n=10+10) AppendSlice/4Bytes-16 1.04ns ± 0% 1.05ns ± 0% +0.79% (p=0.000 n=9+10) AppendSlice/7Bytes-16 1.23ns ± 1% 1.24ns ± 0% +0.45% (p=0.001 n=10+10) AppendSlice/8Bytes-16 1.24ns ± 0% 1.24ns ± 0% ~ (p=0.183 n=10+10) AppendSlice/15Bytes-16 1.37ns ± 1% 1.43ns ± 1% +3.88% (p=0.000 n=10+10) AppendSlice/16Bytes-16 1.37ns ± 1% 1.42ns ± 1% +3.63% (p=0.000 n=9+10) AppendSlice/32Bytes-16 1.44ns ± 0% 1.47ns ± 1% +1.83% (p=0.000 n=10+10) AppendSliceLarge/1024Bytes-16 257ns ± 2% 234ns ± 1% -8.96% (p=0.000 n=8+9) AppendSliceLarge/4096Bytes-16 871ns ± 6% 812ns ± 1% -6.80% (p=0.000 n=10+10) AppendSliceLarge/16384Bytes-16 3.15µs ± 6% 3.04µs ± 5% ~ (p=0.052 n=10+10) AppendSliceLarge/65536Bytes-16 10.7µs ± 7% 10.8µs ± 2% ~ (p=0.278 n=10+9) AppendSliceLarge/262144Bytes-16 42.9µs ± 1% 39.6µs ± 5% -7.75% (p=0.000 n=9+10) AppendSliceLarge/1048576Bytes-16 147µs ± 4% 144µs ± 4% -2.21% (p=0.035 n=10+10) AppendStr/1Bytes-16 1.20ns ± 0% 1.20ns ± 0% ~ (p=0.755 n=10+10) AppendStr/4Bytes-16 1.13ns ± 0% 1.14ns ± 1% +1.20% (p=0.000 n=10+10) AppendStr/8Bytes-16 1.24ns ± 0% 1.25ns ± 0% +0.93% (p=0.000 n=10+10) AppendStr/16Bytes-16 1.40ns ± 0% 1.42ns ± 0% +2.10% (p=0.000 n=9+10) AppendStr/32Bytes-16 1.44ns ± 0% 1.45ns ± 0% +0.99% (p=0.000 n=10+10) AppendSpecialCase-16 8.64ns ± 1% 8.89ns ± 2% +2.90% (p=0.000 n=10+10) Copy/1Byte-16 1.24ns ± 1% 1.24ns ± 0% -0.28% (p=0.000 n=10+6) Copy/1String-16 1.24ns ± 0% 1.23ns ± 0% ~ (p=0.160 n=10+10) Copy/2Byte-16 1.24ns ± 0% 1.24ns ± 0% ~ (p=0.115 n=10+10) Copy/2String-16 1.24ns ± 0% 1.24ns ± 1% ~ (p=0.954 n=10+10) Copy/4Byte-16 1.24ns ± 0% 1.24ns ± 0% -0.44% (p=0.001 n=10+10) Copy/4String-16 1.23ns ± 0% 1.23ns ± 0% ~ (p=0.081 n=10+10) Copy/8Byte-16 1.37ns ± 0% 1.34ns ± 0% -1.79% (p=0.000 n=9+9) Copy/8String-16 1.34ns ± 0% 1.34ns ± 0% -0.58% (p=0.000 n=9+10) Copy/12Byte-16 1.44ns ± 0% 1.44ns ± 0% ~ (p=0.149 n=9+9) Copy/12String-16 1.44ns ± 0% 1.45ns ± 0% ~ (p=0.124 n=9+9) Copy/16Byte-16 1.44ns ± 0% 1.44ns ± 0% -0.19% (p=0.004 n=10+9) Copy/16String-16 1.44ns ± 0% 1.45ns ± 0% +0.30% (p=0.008 n=10+10) Copy/32Byte-16 1.63ns ± 1% 1.62ns ± 1% -0.72% (p=0.002 n=10+10) Copy/32String-16 1.60ns ± 1% 1.64ns ± 0% +2.23% (p=0.000 n=10+10) Copy/128Byte-16 2.06ns ± 0% 2.06ns ± 0% ~ (p=0.757 n=9+10) Copy/128String-16 2.07ns ± 0% 2.07ns ± 0% +0.36% (p=0.004 n=10+10) Copy/1024Byte-16 6.07ns ± 2% 6.00ns ± 1% -1.20% (p=0.000 n=9+10) Copy/1024String-16 6.05ns ± 0% 5.95ns ± 1% -1.54% (p=0.000 n=10+9) AppendInPlace/NoGrow/Byte-16 288ns ± 1% 284ns ± 1% -1.58% (p=0.000 n=10+10) AppendInPlace/NoGrow/1Ptr-16 844ns ± 1% 809ns ± 3% -4.13% (p=0.000 n=9+10) AppendInPlace/NoGrow/2Ptr-16 1.47µs ± 1% 1.46µs ± 1% ~ (p=0.388 n=9+10) AppendInPlace/NoGrow/3Ptr-16 1.87µs ± 7% 1.91µs ± 1% ~ (p=0.166 n=10+8) AppendInPlace/NoGrow/4Ptr-16 2.66µs ± 1% 2.67µs ± 3% ~ (p=0.968 n=9+10) AppendInPlace/Grow/Byte-16 126ns ± 2% 121ns ± 2% -4.06% (p=0.000 n=10+10) AppendInPlace/Grow/1Ptr-16 132ns ± 2% 127ns ± 2% -4.28% (p=0.000 n=10+9) AppendInPlace/Grow/2Ptr-16 196ns ± 2% 188ns ± 1% -4.20% (p=0.000 n=10+8) AppendInPlace/Grow/3Ptr-16 264ns ± 1% 260ns ± 1% -1.51% (p=0.000 n=9+10) AppendInPlace/Grow/4Ptr-16 297ns ± 2% 294ns ± 2% ~ (p=0.085 n=10+10) StackCopyPtr-16 36.4ms ± 2% 36.7ms ± 2% ~ (p=0.481 n=10+10) StackCopy-16 33.9ms ± 3% 32.6ms ± 1% -3.87% (p=0.000 n=10+8) StackCopyNoCache-16 1.00ms ± 5% 1.01ms ± 5% ~ (p=0.143 n=10+10) StackCopyWithStkobj-16 11.0ms ± 3% 10.9ms ± 4% ~ (p=0.579 n=10+10) Issue18138-16 49.2µs ± 5% 49.0µs ± 4% ~ (p=1.000 n=10+9) CompareStringEqual-16 1.39ns ± 1% 1.45ns ± 2% +3.80% (p=0.000 n=8+10) CompareStringIdentical-16 0.55ns ± 1% 0.55ns ± 0% +0.42% (p=0.007 n=10+10) CompareStringSameLength-16 1.03ns ± 0% 1.03ns ± 0% ~ (p=0.430 n=9+10) CompareStringDifferentLength-16 0.11ns ± 2% 0.11ns ± 3% ~ (p=0.139 n=9+10) CompareStringBigUnaligned-16 23.9µs ± 1% 24.0µs ± 1% ~ (p=0.370 n=9+8) CompareStringBig-16 22.0µs ± 3% 22.2µs ± 3% ~ (p=0.243 n=9+10) ConcatStringAndBytes-16 10.7ns ± 1% 10.0ns ± 2% -6.33% (p=0.000 n=10+10) SliceByteToString/1-16 1.34ns ± 0% 1.34ns ± 0% ~ (p=0.057 n=10+10) SliceByteToString/2-16 6.67ns ± 2% 6.60ns ± 3% ~ (p=0.101 n=10+10) SliceByteToString/4-16 7.76ns ± 2% 7.56ns ± 3% -2.59% (p=0.001 n=10+10) SliceByteToString/8-16 9.81ns ± 4% 9.57ns ± 2% -2.48% (p=0.005 n=10+10) SliceByteToString/16-16 14.0ns ± 3% 13.7ns ± 2% -2.31% (p=0.009 n=10+10) SliceByteToString/32-16 17.3ns ± 1% 16.7ns ± 2% -3.41% (p=0.000 n=10+10) SliceByteToString/64-16 25.1ns ± 1% 24.1ns ± 2% -3.93% (p=0.000 n=9+10) SliceByteToString/128-16 38.6ns ± 1% 36.5ns ± 1% -5.60% (p=0.000 n=10+10) RuneCount/lenruneslice/ASCII-16 4.12ns ± 0% 4.11ns ± 0% ~ (p=0.382 n=10+10) RuneCount/lenruneslice/Japanese-16 25.4ns ± 2% 25.6ns ± 2% ~ (p=0.138 n=9+10) RuneCount/lenruneslice/MixedLength-16 17.1ns ± 0% 17.2ns ± 0% +0.59% (p=0.000 n=9+9) RuneCount/rangeloop/ASCII-16 3.30ns ± 1% 3.29ns ± 0% ~ (p=0.267 n=10+10) RuneCount/rangeloop/Japanese-16 20.1ns ± 1% 24.9ns ± 1% +24.31% (p=0.000 n=9+9) RuneCount/rangeloop/MixedLength-16 16.5ns ± 1% 16.7ns ± 1% +1.34% (p=0.000 n=10+10) RuneCount/utf8.RuneCountInString/ASCII-16 5.71ns ± 1% 5.73ns ± 2% ~ (p=0.579 n=10+10) RuneCount/utf8.RuneCountInString/Japanese-16 22.0ns ± 6% 18.4ns ± 3% -16.41% (p=0.000 n=9+10) RuneCount/utf8.RuneCountInString/MixedLength-16 15.0ns ± 1% 14.9ns ± 1% -1.01% (p=0.004 n=9+10) RuneIterate/range/ASCII-16 2.69ns ± 1% 2.72ns ± 0% +0.94% (p=0.026 n=10+9) RuneIterate/range/Japanese-16 24.5ns ± 2% 25.3ns ± 2% +3.23% (p=0.000 n=9+10) RuneIterate/range/MixedLength-16 17.0ns ± 1% 17.1ns ± 1% +0.85% (p=0.000 n=10+10) RuneIterate/range1/ASCII-16 2.70ns ± 1% 2.72ns ± 0% ~ (p=0.058 n=9+9) RuneIterate/range1/Japanese-16 24.1ns ± 2% 25.2ns ± 3% +4.30% (p=0.000 n=10+10) RuneIterate/range1/MixedLength-16 16.9ns ± 1% 17.7ns ± 0% +5.04% (p=0.000 n=10+8) RuneIterate/range2/ASCII-16 2.84ns ± 8% 2.72ns ± 1% -4.28% (p=0.003 n=10+9) RuneIterate/range2/Japanese-16 22.7ns ± 4% 25.2ns ± 3% +10.97% (p=0.000 n=10+10) RuneIterate/range2/MixedLength-16 17.0ns ± 1% 17.2ns ± 0% +0.95% (p=0.000 n=10+10) ArrayEqual-16 0.40ns ± 5% 0.35ns ± 2% -11.83% (p=0.000 n=10+10) Func/Name-16 8.05ns ± 1% 8.09ns ± 1% +0.40% (p=0.025 n=8+10) Func/Entry-16 1.73ns ± 1% 1.66ns ± 1% -3.93% (p=0.000 n=10+10) Func/FileLine-16 27.5ns ± 2% 26.0ns ± 0% -5.50% (p=0.000 n=10+10) [Geo mean] 16.7ns 15.7ns -6.08% name old speed new speed delta SetTypePtr-16 11.0GB/s ± 1% 11.0GB/s ± 3% ~ (p=0.684 n=10+10) SetTypePtr8-16 15.5GB/s ± 0% 15.5GB/s ± 0% ~ (p=0.123 n=10+10) SetTypePtr16-16 31.0GB/s ± 1% 31.1GB/s ± 0% ~ (p=0.123 n=10+10) SetTypePtr32-16 62.1GB/s ± 0% 62.2GB/s ± 0% ~ (p=0.123 n=10+10) SetTypePtr64-16 124GB/s ± 0% 124GB/s ± 0% ~ (p=0.684 n=10+10) SetTypePtr126-16 146GB/s ± 0% 146GB/s ± 0% ~ (p=0.481 n=10+10) SetTypePtr128-16 154GB/s ± 0% 154GB/s ± 0% ~ (p=0.243 n=9+10) SetTypePtrSlice-16 151GB/s ± 1% 151GB/s ± 1% ~ (p=0.497 n=9+10) SetTypeNode1-16 5.82GB/s ± 1% 5.82GB/s ± 0% ~ (p=0.353 n=10+10) SetTypeNode1Slice-16 76.1GB/s ± 1% 77.0GB/s ± 1% +1.19% (p=0.000 n=10+10) SetTypeNode8-16 19.4GB/s ± 0% 19.4GB/s ± 0% ~ (p=0.130 n=8+8) SetTypeNode8Slice-16 113GB/s ± 0% 113GB/s ± 0% ~ (p=0.604 n=10+9) SetTypeNode64-16 76.5GB/s ± 0% 76.5GB/s ± 0% ~ (p=0.190 n=10+10) SetTypeNode64Slice-16 97.8GB/s ± 0% 97.7GB/s ± 0% ~ (p=0.549 n=9+10) SetTypeNode64Dead-16 95.5GB/s ± 0% 95.7GB/s ± 0% ~ (p=0.118 n=10+6) SetTypeNode64DeadSlice-16 112GB/s ± 0% 112GB/s ± 0% ~ (p=0.353 n=10+10) SetTypeNode124-16 146GB/s ± 0% 146GB/s ± 0% ~ (p=0.853 n=10+10) SetTypeNode124Slice-16 146GB/s ± 5% 149GB/s ± 0% ~ (p=0.315 n=10+10) SetTypeNode126-16 154GB/s ± 0% 154GB/s ± 0% ~ (p=0.356 n=10+9) SetTypeNode126Slice-16 150GB/s ± 0% 150GB/s ± 0% ~ (p=0.095 n=9+10) SetTypeNode128-16 107GB/s ± 0% 107GB/s ± 0% +0.31% (p=0.003 n=9+10) SetTypeNode128Slice-16 119GB/s ± 0% 120GB/s ± 0% ~ (p=0.156 n=10+9) SetTypeNode130-16 108GB/s ± 0% 108GB/s ± 0% +0.33% (p=0.002 n=10+10) SetTypeNode130Slice-16 119GB/s ± 0% 119GB/s ± 0% ~ (p=0.739 n=10+10) SetTypeNode1024-16 160GB/s ± 0% 159GB/s ± 1% ~ (p=0.113 n=9+9) SetTypeNode1024Slice-16 144GB/s ± 0% 144GB/s ± 0% ~ (p=0.063 n=10+10) Hash5-16 2.59GB/s ± 1% 2.49GB/s ± 0% -3.90% (p=0.000 n=10+9) Hash16-16 7.85GB/s ± 1% 7.23GB/s ± 1% -7.92% (p=0.000 n=10+10) Hash64-16 24.0GB/s ± 0% 23.9GB/s ± 0% ~ (p=0.190 n=9+9) Hash1024-16 62.4GB/s ± 0% 62.3GB/s ± 0% -0.16% (p=0.017 n=9+10) Hash65536-16 74.0GB/s ± 0% 74.0GB/s ± 0% ~ (p=0.796 n=10+10) Memmove/1-16 1.08GB/s ± 0% 1.08GB/s ± 0% -0.21% (p=0.035 n=10+10) Memmove/2-16 2.16GB/s ± 0% 2.15GB/s ± 0% ~ (p=0.105 n=10+10) Memmove/3-16 3.24GB/s ± 1% 3.22GB/s ± 1% -0.49% (p=0.004 n=10+10) Memmove/4-16 3.89GB/s ± 0% 3.89GB/s ± 0% ~ (p=0.218 n=10+10) Memmove/5-16 4.42GB/s ± 0% 4.42GB/s ± 0% ~ (p=0.075 n=10+10) Memmove/6-16 5.31GB/s ± 0% 5.29GB/s ± 1% ~ (p=0.218 n=10+10) Memmove/7-16 6.19GB/s ± 0% 6.18GB/s ± 0% -0.15% (p=0.035 n=10+9) Memmove/8-16 7.07GB/s ± 0% 7.07GB/s ± 0% ~ (p=0.684 n=10+10) Memmove/9-16 7.22GB/s ± 0% 6.68GB/s ± 0% -7.37% (p=0.000 n=10+10) Memmove/10-16 8.02GB/s ± 0% 7.43GB/s ± 0% -7.38% (p=0.000 n=9+9) Memmove/11-16 8.83GB/s ± 0% 8.13GB/s ± 0% -7.87% (p=0.000 n=10+9) Memmove/12-16 9.62GB/s ± 0% 8.89GB/s ± 1% -7.61% (p=0.000 n=10+10) Memmove/13-16 10.4GB/s ± 0% 9.7GB/s ± 0% -7.20% (p=0.000 n=10+10) Memmove/14-16 11.2GB/s ± 0% 10.4GB/s ± 1% -7.64% (p=0.000 n=10+9) Memmove/15-16 12.0GB/s ± 0% 11.1GB/s ± 0% -7.46% (p=0.000 n=10+9) Memmove/16-16 12.8GB/s ± 0% 11.8GB/s ± 1% -7.67% (p=0.000 n=10+10) Memmove/32-16 23.8GB/s ± 0% 23.5GB/s ± 1% -1.20% (p=0.000 n=10+10) Memmove/64-16 44.2GB/s ± 0% 39.1GB/s ± 0% -11.56% (p=0.000 n=10+9) Memmove/128-16 68.7GB/s ± 0% 63.2GB/s ± 0% -7.95% (p=0.000 n=10+10) Memmove/256-16 104GB/s ± 0% 103GB/s ± 0% -1.13% (p=0.000 n=10+10) Memmove/512-16 129GB/s ± 1% 129GB/s ± 0% ~ (p=0.165 n=10+10) Memmove/1024-16 174GB/s ± 1% 174GB/s ± 1% ~ (p=0.258 n=9+9) Memmove/2048-16 213GB/s ± 1% 213GB/s ± 2% ~ (p=0.963 n=8+9) Memmove/4096-16 250GB/s ± 1% 240GB/s ± 4% -3.83% (p=0.006 n=9+9) MemmoveOverlap/32-16 19.8GB/s ± 1% 19.1GB/s ± 1% -3.40% (p=0.000 n=10+10) MemmoveOverlap/64-16 39.0GB/s ± 0% 38.8GB/s ± 0% -0.28% (p=0.001 n=9+9) MemmoveOverlap/128-16 62.2GB/s ± 0% 62.1GB/s ± 0% ~ (p=0.063 n=10+10) MemmoveOverlap/256-16 96.0GB/s ± 0% 95.8GB/s ± 0% -0.26% (p=0.009 n=10+10) MemmoveOverlap/512-16 83.6GB/s ±16% 89.2GB/s ± 0% ~ (p=0.696 n=10+8) MemmoveOverlap/1024-16 141GB/s ± 0% 140GB/s ± 0% -0.28% (p=0.006 n=8+10) MemmoveOverlap/2048-16 172GB/s ± 0% 171GB/s ± 1% -0.38% (p=0.008 n=9+9) MemmoveOverlap/4096-16 176GB/s ± 1% 177GB/s ± 1% +0.84% (p=0.001 n=8+10) MemmoveUnalignedDst/1-16 806MB/s ± 0% 802MB/s ± 1% -0.52% (p=0.023 n=10+10) MemmoveUnalignedDst/2-16 1.62GB/s ± 0% 1.62GB/s ± 0% -0.11% (p=0.041 n=10+10) MemmoveUnalignedDst/3-16 2.43GB/s ± 0% 2.43GB/s ± 0% -0.14% (p=0.006 n=9+9) MemmoveUnalignedDst/4-16 3.24GB/s ± 0% 3.23GB/s ± 1% -0.36% (p=0.007 n=10+10) MemmoveUnalignedDst/5-16 3.71GB/s ± 0% 3.71GB/s ± 0% ~ (p=0.063 n=10+10) MemmoveUnalignedDst/6-16 4.48GB/s ± 0% 4.47GB/s ± 0% ~ (p=0.912 n=10+10) MemmoveUnalignedDst/7-16 5.22GB/s ± 0% 5.22GB/s ± 0% ~ (p=1.000 n=10+10) MemmoveUnalignedDst/8-16 5.95GB/s ± 0% 5.93GB/s ± 1% -0.40% (p=0.023 n=10+10) MemmoveUnalignedDst/9-16 6.24GB/s ± 0% 6.24GB/s ± 0% ~ (p=0.912 n=10+10) MemmoveUnalignedDst/10-16 6.94GB/s ± 0% 6.94GB/s ± 0% ~ (p=0.353 n=10+10) MemmoveUnalignedDst/11-16 7.64GB/s ± 0% 7.63GB/s ± 0% ~ (p=0.393 n=10+10) MemmoveUnalignedDst/12-16 8.33GB/s ± 0% 8.33GB/s ± 0% ~ (p=0.971 n=10+10) MemmoveUnalignedDst/13-16 9.02GB/s ± 0% 9.01GB/s ± 0% ~ (p=0.436 n=10+10) MemmoveUnalignedDst/14-16 9.71GB/s ± 0% 9.71GB/s ± 0% ~ (p=0.280 n=10+10) MemmoveUnalignedDst/15-16 10.4GB/s ± 0% 10.4GB/s ± 1% ~ (p=0.853 n=10+10) MemmoveUnalignedDst/16-16 11.1GB/s ± 0% 11.1GB/s ± 0% ~ (p=0.089 n=10+10) MemmoveUnalignedDst/32-16 19.7GB/s ± 1% 19.6GB/s ± 0% ~ (p=0.075 n=10+10) MemmoveUnalignedDst/64-16 38.9GB/s ± 0% 38.8GB/s ± 0% ~ (p=0.218 n=10+10) MemmoveUnalignedDst/128-16 62.1GB/s ± 0% 62.1GB/s ± 0% ~ (p=0.549 n=10+9) MemmoveUnalignedDst/256-16 69.4GB/s ± 0% 69.3GB/s ± 0% ~ (p=0.105 n=10+10) MemmoveUnalignedDst/512-16 124GB/s ± 1% 124GB/s ± 0% ~ (p=0.762 n=10+8) MemmoveUnalignedDst/1024-16 136GB/s ± 0% 136GB/s ± 1% ~ (p=0.666 n=9+9) MemmoveUnalignedDst/2048-16 159GB/s ± 0% 159GB/s ± 0% ~ (p=0.574 n=8+8) MemmoveUnalignedDst/4096-16 161GB/s ± 0% 161GB/s ± 0% ~ (p=1.000 n=9+9) MemmoveUnalignedDstOverlap/32-16 7.84GB/s ± 0% 7.83GB/s ± 0% ~ (p=0.353 n=10+10) MemmoveUnalignedDstOverlap/64-16 14.0GB/s ± 0% 14.0GB/s ± 0% ~ (p=0.661 n=10+9) MemmoveUnalignedDstOverlap/128-16 27.4GB/s ± 0% 27.4GB/s ± 0% ~ (p=0.353 n=10+10) MemmoveUnalignedDstOverlap/256-16 50.4GB/s ± 0% 50.4GB/s ± 0% ~ (p=0.156 n=10+9) MemmoveUnalignedDstOverlap/512-16 60.7GB/s ± 4% 62.5GB/s ± 0% +3.07% (p=0.022 n=10+9) MemmoveUnalignedDstOverlap/1024-16 107GB/s ± 0% 107GB/s ± 0% ~ (p=0.234 n=8+8) MemmoveUnalignedDstOverlap/2048-16 146GB/s ± 0% 146GB/s ± 1% ~ (p=0.182 n=10+9) MemmoveUnalignedDstOverlap/4096-16 155GB/s ± 0% 155GB/s ± 0% ~ (p=0.400 n=10+9) MemmoveUnalignedSrc/1-16 882MB/s ± 0% 884MB/s ± 1% +0.24% (p=0.033 n=10+9) MemmoveUnalignedSrc/2-16 1.76GB/s ± 1% 1.77GB/s ± 0% +0.27% (p=0.028 n=10+9) MemmoveUnalignedSrc/3-16 2.43GB/s ± 0% 2.43GB/s ± 0% +0.26% (p=0.027 n=9+10) MemmoveUnalignedSrc/4-16 3.24GB/s ± 0% 3.24GB/s ± 1% ~ (p=0.079 n=9+10) MemmoveUnalignedSrc/5-16 3.73GB/s ± 0% 3.73GB/s ± 1% ~ (p=0.829 n=8+10) MemmoveUnalignedSrc/6-16 4.47GB/s ± 0% 4.49GB/s ± 0% +0.39% (p=0.000 n=10+10) MemmoveUnalignedSrc/7-16 5.22GB/s ± 0% 5.23GB/s ± 0% ~ (p=0.280 n=10+10) MemmoveUnalignedSrc/8-16 5.95GB/s ± 0% 5.98GB/s ± 0% +0.39% (p=0.001 n=10+9) MemmoveUnalignedSrc/9-16 6.24GB/s ± 0% 6.25GB/s ± 0% ~ (p=0.549 n=10+9) MemmoveUnalignedSrc/10-16 6.93GB/s ± 0% 6.94GB/s ± 0% ~ (p=0.604 n=10+9) MemmoveUnalignedSrc/11-16 7.63GB/s ± 0% 7.63GB/s ± 1% ~ (p=0.353 n=10+10) MemmoveUnalignedSrc/12-16 8.32GB/s ± 0% 8.32GB/s ± 0% ~ (p=0.218 n=10+10) MemmoveUnalignedSrc/13-16 9.02GB/s ± 0% 9.00GB/s ± 1% ~ (p=0.684 n=10+10) MemmoveUnalignedSrc/14-16 9.71GB/s ± 0% 9.71GB/s ± 0% ~ (p=0.739 n=10+10) MemmoveUnalignedSrc/15-16 10.4GB/s ± 0% 10.4GB/s ± 0% ~ (p=0.353 n=10+10) MemmoveUnalignedSrc/16-16 11.1GB/s ± 1% 11.1GB/s ± 0% ~ (p=0.579 n=10+10) MemmoveUnalignedSrc/32-16 20.0GB/s ± 1% 20.0GB/s ± 0% ~ (p=0.631 n=10+10) MemmoveUnalignedSrc/64-16 38.8GB/s ± 0% 38.8GB/s ± 0% ~ (p=0.579 n=10+10) MemmoveUnalignedSrc/128-16 61.2GB/s ± 0% 61.2GB/s ± 0% ~ (p=0.780 n=10+9) MemmoveUnalignedSrc/256-16 94.8GB/s ± 0% 92.2GB/s ± 1% -2.73% (p=0.000 n=10+10) MemmoveUnalignedSrc/512-16 119GB/s ± 0% 119GB/s ± 0% +0.26% (p=0.027 n=8+9) MemmoveUnalignedSrc/1024-16 141GB/s ± 0% 142GB/s ± 1% +1.07% (p=0.000 n=8+10) MemmoveUnalignedSrc/2048-16 157GB/s ± 0% 157GB/s ± 0% ~ (p=0.167 n=9+8) MemmoveUnalignedSrc/4096-16 161GB/s ± 0% 162GB/s ± 1% ~ (p=0.063 n=10+10) MemmoveUnalignedSrcOverlap/32-16 7.93GB/s ± 0% 7.88GB/s ± 0% -0.63% (p=0.000 n=9+10) MemmoveUnalignedSrcOverlap/64-16 15.5GB/s ± 0% 15.5GB/s ± 0% ~ (p=0.529 n=10+10) MemmoveUnalignedSrcOverlap/128-16 28.3GB/s ± 0% 28.3GB/s ± 0% ~ (p=0.218 n=10+10) MemmoveUnalignedSrcOverlap/256-16 41.5GB/s ± 0% 41.6GB/s ± 0% +0.35% (p=0.000 n=10+9) MemmoveUnalignedSrcOverlap/512-16 68.9GB/s ± 0% 68.8GB/s ± 0% ~ (p=0.541 n=9+8) MemmoveUnalignedSrcOverlap/1024-16 115GB/s ± 0% 115GB/s ± 0% ~ (p=0.382 n=8+8) MemmoveUnalignedSrcOverlap/2048-16 155GB/s ± 0% 144GB/s ±18% ~ (p=0.101 n=8+10) MemmoveUnalignedSrcOverlap/4096-16 160GB/s ± 0% 160GB/s ± 1% ~ (p=0.605 n=9+9) Memclr/5-16 5.81GB/s ± 1% 5.80GB/s ± 2% ~ (p=0.546 n=9+9) Memclr/16-16 15.5GB/s ± 0% 15.4GB/s ± 0% -0.32% (p=0.008 n=9+10) Memclr/64-16 51.8GB/s ± 0% 50.7GB/s ± 0% -2.22% (p=0.000 n=10+10) Memclr/256-16 113GB/s ± 0% 113GB/s ± 0% ~ (p=0.143 n=10+10) Memclr/4096-16 239GB/s ± 1% 237GB/s ± 0% -0.87% (p=0.000 n=10+10) Memclr/65536-16 79.8GB/s ± 0% 79.7GB/s ± 0% ~ (p=0.529 n=10+10) Memclr/1M-16 74.6GB/s ± 1% 74.7GB/s ± 1% ~ (p=0.529 n=10+10) Memclr/4M-16 48.7GB/s ± 1% 48.8GB/s ± 0% ~ (p=0.123 n=10+10) Memclr/8M-16 48.2GB/s ± 2% 48.6GB/s ± 0% ~ (p=0.408 n=10+8) Memclr/16M-16 43.6GB/s ± 4% 43.3GB/s ± 0% ~ (p=0.173 n=10+8) Memclr/64M-16 30.7GB/s ± 0% 30.7GB/s ± 0% ~ (p=0.113 n=10+9) GoMemclr/5-16 6.07GB/s ± 0% 6.08GB/s ± 0% ~ (p=0.367 n=9+10) GoMemclr/16-16 15.6GB/s ± 0% 15.6GB/s ± 0% -0.22% (p=0.004 n=9+9) GoMemclr/64-16 56.1GB/s ± 0% 56.1GB/s ± 0% ~ (p=0.968 n=10+9) GoMemclr/256-16 125GB/s ± 0% 124GB/s ± 0% ~ (p=0.912 n=10+10) MemclrRange/1K_2K-16 210GB/s ± 0% 224GB/s ± 1% +6.81% (p=0.000 n=10+10) MemclrRange/2K_8K-16 228GB/s ± 0% 228GB/s ± 0% ~ (p=0.684 n=10+10) MemclrRange/4K_16K-16 279GB/s ± 0% 279GB/s ± 0% ~ (p=0.780 n=9+10) MemclrRange/160K_228K-16 80.3GB/s ± 0% 80.2GB/s ± 0% ~ (p=0.165 n=10+10) Copy/1Byte-16 808MB/s ± 1% 810MB/s ± 0% +0.28% (p=0.000 n=10+8) Copy/1String-16 810MB/s ± 0% 811MB/s ± 0% ~ (p=0.105 n=10+10) Copy/2Byte-16 1.62GB/s ± 0% 1.62GB/s ± 0% ~ (p=0.182 n=10+9) Copy/2String-16 1.62GB/s ± 1% 1.62GB/s ± 1% ~ (p=1.000 n=10+10) Copy/4Byte-16 3.22GB/s ± 0% 3.24GB/s ± 0% +0.46% (p=0.000 n=10+10) Copy/4String-16 3.24GB/s ± 0% 3.24GB/s ± 0% ~ (p=0.075 n=10+10) Copy/8Byte-16 5.86GB/s ± 0% 5.96GB/s ± 0% +1.82% (p=0.000 n=9+9) Copy/8String-16 5.95GB/s ± 0% 5.99GB/s ± 0% +0.59% (p=0.000 n=9+10) Copy/12Byte-16 8.32GB/s ± 0% 8.32GB/s ± 0% ~ (p=0.190 n=9+9) Copy/12String-16 8.31GB/s ± 0% 8.29GB/s ± 0% ~ (p=0.068 n=9+10) Copy/16Byte-16 11.1GB/s ± 0% 11.1GB/s ± 0% +0.18% (p=0.003 n=10+9) Copy/16String-16 11.1GB/s ± 0% 11.1GB/s ± 0% -0.31% (p=0.009 n=10+10) Copy/32Byte-16 19.6GB/s ± 1% 19.8GB/s ± 1% +0.72% (p=0.002 n=10+10) Copy/32String-16 20.0GB/s ± 0% 19.5GB/s ± 0% -2.19% (p=0.000 n=10+10) Copy/128Byte-16 62.2GB/s ± 0% 62.1GB/s ± 0% ~ (p=0.661 n=9+10) Copy/128String-16 61.9GB/s ± 0% 61.7GB/s ± 0% -0.35% (p=0.005 n=10+10) Copy/1024Byte-16 169GB/s ± 2% 171GB/s ± 1% +1.21% (p=0.000 n=9+10) Copy/1024String-16 169GB/s ± 0% 172GB/s ± 1% +1.57% (p=0.000 n=10+9) CompareStringBigUnaligned-16 43.8GB/s ± 1% 43.7GB/s ± 1% ~ (p=0.370 n=9+8) CompareStringBig-16 47.6GB/s ± 3% 47.3GB/s ± 3% ~ (p=0.243 n=9+10) [Geo mean] 25.3GB/s 28.0GB/s +10.66% name old p50-ns new p50-ns delta ReadMemStatsLatency-16 98.4k ±37% 112.7k ±64% ~ (p=0.436 n=10+10) ReadMetricsLatency-16 1.69k ± 3% 1.70k ± 2% ~ (p=0.646 n=9+10) GoroutineProfile/small-nil/idle-16 3.75k ± 4% 3.72k ± 1% ~ (p=0.447 n=10+9) GoroutineProfile/small-nil/loaded-16 4.33k ± 3% 4.31k ± 5% ~ (p=0.931 n=9+9) GoroutineProfile/small/idle-16 102k ± 3% 101k ± 4% ~ (p=0.113 n=9+9) GoroutineProfile/small/loaded-16 214k ± 3% 215k ± 6% ~ (p=0.842 n=9+10) GoroutineProfile/large-nil/idle-16 3.70k ± 2% 3.65k ± 2% ~ (p=0.075 n=10+10) GoroutineProfile/large-nil/loaded-16 4.36k ± 3% 4.31k ±10% ~ (p=0.631 n=10+10) GoroutineProfile/large/idle-16 2.56M ± 1% 2.51M ± 1% -2.28% (p=0.000 n=10+10) GoroutineProfile/large/loaded-16 6.77M ± 4% 6.85M ±19% ~ (p=0.536 n=7+10) GoroutineProfile/sparse-nil/idle-16 3.66k ± 1% 3.64k ± 2% ~ (p=0.136 n=9+9) GoroutineProfile/sparse-nil/loaded-16 4.25k ± 5% 4.15k ± 4% ~ (p=0.190 n=10+10) GoroutineProfile/sparse/idle-16 102k ± 4% 101k ± 3% ~ (p=0.447 n=10+9) GoroutineProfile/sparse/loaded-16 216k ± 4% 218k ± 4% ~ (p=0.549 n=9+10) [Geo mean] 35.8k 35.9k +0.35% name old p90-ns new p90-ns delta ReadMemStatsLatency-16 983k ±310% 200k ±34% -79.62% (p=0.034 n=10+8) ReadMetricsLatency-16 4.01k ±35% 3.75k ±17% ~ (p=0.315 n=10+10) GoroutineProfile/small-nil/idle-16 4.21k ± 4% 4.27k ± 8% ~ (p=0.968 n=9+10) GoroutineProfile/small-nil/loaded-16 5.58k ± 8% 5.35k ±12% ~ (p=0.190 n=10+10) GoroutineProfile/small/idle-16 108k ± 6% 107k ± 7% ~ (p=0.497 n=9+10) GoroutineProfile/small/loaded-16 450k ± 5% 432k ± 3% -3.92% (p=0.002 n=9+9) GoroutineProfile/large-nil/idle-16 4.13k ± 7% 4.04k ± 2% ~ (p=0.181 n=10+8) GoroutineProfile/large-nil/loaded-16 5.76k ± 4% 5.67k ± 5% ~ (p=0.190 n=10+10) GoroutineProfile/large/idle-16 2.63M ± 2% 2.58M ± 1% -1.97% (p=0.000 n=10+10) GoroutineProfile/large/loaded-16 16.9M ± 4% 17.0M ± 6% ~ (p=0.661 n=9+10) GoroutineProfile/sparse-nil/idle-16 4.21k ±10% 4.07k ± 7% ~ (p=0.128 n=10+10) GoroutineProfile/sparse-nil/loaded-16 5.55k ± 8% 5.38k ± 6% ~ (p=0.089 n=10+10) GoroutineProfile/sparse/idle-16 106k ± 4% 106k ± 3% ~ (p=0.661 n=10+9) GoroutineProfile/sparse/loaded-16 454k ± 6% 441k ± 6% -2.86% (p=0.043 n=10+10) [Geo mean] 58.4k 51.0k -12.61% name old p99-ns new p99-ns delta ReadMemStatsLatency-16 983k ±310% 200k ±34% -79.62% (p=0.034 n=10+8) ReadMetricsLatency-16 26.6k ±22% 26.6k ±17% ~ (p=0.971 n=10+10) GoroutineProfile/small-nil/idle-16 5.27k ±12% 5.19k ±12% ~ (p=0.579 n=10+10) GoroutineProfile/small-nil/loaded-16 7.28k ± 3% 7.04k ± 9% ~ (p=0.113 n=9+10) GoroutineProfile/small/idle-16 114k ± 6% 113k ± 6% ~ (p=0.604 n=9+10) GoroutineProfile/small/loaded-16 4.84M ±70% 6.06M ±131% ~ (p=0.842 n=9+10) GoroutineProfile/large-nil/idle-16 5.38k ±19% 5.26k ±13% ~ (p=0.912 n=10+10) GoroutineProfile/large-nil/loaded-16 7.38k ± 3% 7.23k ± 4% ~ (p=0.143 n=10+10) GoroutineProfile/large/idle-16 2.79M ± 5% 2.72M ± 3% ~ (p=0.089 n=10+10) GoroutineProfile/large/loaded-16 24.0M ±24% 24.9M ±29% ~ (p=0.684 n=10+10) GoroutineProfile/sparse-nil/idle-16 5.32k ±17% 5.49k ±17% ~ (p=0.684 n=10+10) GoroutineProfile/sparse-nil/loaded-16 7.25k ± 4% 6.97k ± 5% -3.90% (p=0.005 n=9+10) GoroutineProfile/sparse/idle-16 113k ± 5% 112k ± 6% ~ (p=0.631 n=10+10) GoroutineProfile/sparse/loaded-16 4.00M ±66% 4.26M ±65% ~ (p=0.489 n=9+9) [Geo mean] 107k 97k -9.55% name old alloc/op new alloc/op delta NewEmptyMap-16 0.00B 0.00B ~ (all equal) NewSmallMap-16 0.00B 0.00B ~ (all equal) MapPopulate/1-16 0.00B 0.00B ~ (all equal) MapPopulate/10-16 179B ± 0% 179B ± 0% ~ (all equal) MapPopulate/100-16 3.35kB ± 0% 3.35kB ± 0% ~ (p=0.294 n=10+8) MapPopulate/1000-16 53.3kB ± 0% 53.3kB ± 0% ~ (p=1.000 n=8+10) MapPopulate/10000-16 428kB ± 0% 428kB ± 0% ~ (p=0.469 n=10+10) MapPopulate/100000-16 3.62MB ± 0% 3.62MB ± 0% ~ (p=0.888 n=9+10) MapStringConversion/32/simple-16 0.00B 0.00B ~ (all equal) MapStringConversion/32/struct-16 0.00B 0.00B ~ (all equal) MapStringConversion/32/array-16 0.00B 0.00B ~ (all equal) MapStringConversion/64/simple-16 0.00B 0.00B ~ (all equal) MapStringConversion/64/struct-16 0.00B 0.00B ~ (all equal) MapStringConversion/64/array-16 0.00B 0.00B ~ (all equal) NewEmptyMapHintLessThan8-16 0.00B 0.00B ~ (all equal) NewEmptyMapHintGreaterThan8-16 1.15kB ± 0% 1.15kB ± 0% ~ (all equal) MapAppendAssign/Int32/256-16 41.7B ±15% 44.3B ±12% ~ (p=0.106 n=10+10) MapAppendAssign/Int32/65536-16 22.6B ± 6% 23.5B ± 6% +4.19% (p=0.025 n=9+10) MapAppendAssign/Int64/256-16 43.5B ±10% 42.9B ± 7% ~ (p=0.757 n=10+10) MapAppendAssign/Int64/65536-16 24.7B ± 7% 21.8B ± 6% -11.74% (p=0.000 n=10+10) MapAppendAssign/Str/256-16 87.6B ±10% 89.2B ± 9% ~ (p=0.379 n=10+10) MapAppendAssign/Str/65536-16 45.1B ±14% 47.2B ± 8% ~ (p=0.150 n=10+9) CreateGoroutinesCapture-16 144B ± 0% 144B ± 0% ~ (all equal) [Geo mean] 769B 770B +0.21% name old allocs/op new allocs/op delta NewEmptyMap-16 0.00 0.00 ~ (all equal) NewSmallMap-16 0.00 0.00 ~ (all equal) MapPopulate/1-16 0.00 0.00 ~ (all equal) MapPopulate/10-16 1.00 ± 0% 1.00 ± 0% ~ (all equal) MapPopulate/100-16 17.0 ± 0% 17.0 ± 0% ~ (all equal) MapPopulate/1000-16 73.0 ± 0% 73.0 ± 0% ~ (all equal) MapPopulate/10000-16 320 ± 0% 320 ± 0% ~ (p=1.000 n=10+10) MapPopulate/100000-16 4.00k ± 0% 4.00k ± 0% ~ (p=0.753 n=10+10) MapStringConversion/32/simple-16 0.00 0.00 ~ (all equal) MapStringConversion/32/struct-16 0.00 0.00 ~ (all equal) MapStringConversion/32/array-16 0.00 0.00 ~ (all equal) MapStringConversion/64/simple-16 0.00 0.00 ~ (all equal) MapStringConversion/64/struct-16 0.00 0.00 ~ (all equal) MapStringConversion/64/array-16 0.00 0.00 ~ (all equal) NewEmptyMapHintLessThan8-16 0.00 0.00 ~ (all equal) NewEmptyMapHintGreaterThan8-16 1.00 ± 0% 1.00 ± 0% ~ (all equal) MapAppendAssign/Int32/256-16 0.00 0.00 ~ (all equal) MapAppendAssign/Int32/65536-16 0.00 0.00 ~ (all equal) MapAppendAssign/Int64/256-16 0.00 0.00 ~ (all equal) MapAppendAssign/Int64/65536-16 0.00 0.00 ~ (all equal) MapAppendAssign/Str/256-16 0.00 0.00 ~ (all equal) MapAppendAssign/Str/65536-16 0.00 0.00 ~ (all equal) CreateGoroutinesCapture-16 5.00 ± 0% 5.00 ± 0% ~ (all equal) [Geo mean] 26.0 26.0 +0.00% Change-Id: I5fb03e93df8b380e04795afbdcd1c94aeeecacc6 Reviewed-on: https://go-review.googlesource.com/c/go/+/454255 Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com> Run-TryBot: Jakub Ciolek <jakub@ciolek.dev> TryBot-Result: Gopher Robot <gobot@golang.org>
2022-11-30 15:31:26 +01:00
&& isSameCall(sym, "runtime.memclrNoHeapPointers")
&& call.Uses == 1
&& clobber(call)
=> (Zero {types.Types[types.TUINT8]} [int64(c)] sptr mem)
// Recognise make([]T, 0) and replace it with a pointer to the zerobase
(StaticLECall {callAux} _ (Const(64|32) [0]) (Const(64|32) [0]) mem)
&& isSameCall(callAux, "runtime.makeslice")
=> (MakeResult (Addr <v.Type.FieldType(0)> {ir.Syms.Zerobase} (SB)) mem)
// Evaluate constant address comparisons.
(EqPtr x x) => (ConstBool [true])
(NeqPtr x x) => (ConstBool [false])
cmd/compile: disallow rewrite rules from declaring reserved names If I change a rule in ARM64.rules to use the variable name "b" in a conflicting way, rulegen would previously not complain, and the compiler would later give a confusing error: $ go run *.go && go build cmd/compile/internal/ssa # cmd/compile/internal/ssa ../rewriteARM64.go:24236:10: b.NewValue0 undefined (type int64 has no field or method NewValue0) Make rulegen complain early about those cases. Sometimes they might happen to be harmless, but in general they can easily cause confusion or unintended effect due to shadowing. After the change, with the same conflicting rule: $ go run *.go && go build cmd/compile/internal/ssa 2021/03/22 11:31:49 rule ARM64.rules:495 uses the reserved name b exit status 1 Note that 24 existing rules were using reserved names. It seems like the shadowing was harmless, as it wasn't causing typechecking issues nor did it seem to cause unintended behavior when the rule rewrite code ran. The bool values "b" were renamed "t", since that seems to have a precedent in other rules and in the fmt package. Sequential values like "a b c" were renamed to "x y z", since "b" is reserved. Finally, "typ" was renamed to "_typ", since there doesn't seem to be an obviously better answer. Passes all three of: $ GOARCH=amd64 go build -toolexec 'toolstash -cmp' -a std $ GOARCH=arm64 go build -toolexec 'toolstash -cmp' -a std $ GOARCH=mips64 go build -toolexec 'toolstash -cmp' -a std Fixes #45154. Change-Id: I1cce194dc7b477886a9c218c17973e996bcedccf Reviewed-on: https://go-review.googlesource.com/c/go/+/303549 Trust: Daniel Martí <mvdan@mvdan.cc> Run-TryBot: Daniel Martí <mvdan@mvdan.cc> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2021-03-22 11:35:02 +01:00
(EqPtr (Addr {x} _) (Addr {y} _)) => (ConstBool [x == y])
(EqPtr (Addr {x} _) (OffPtr [o] (Addr {y} _))) => (ConstBool [x == y && o == 0])
(EqPtr (OffPtr [o1] (Addr {x} _)) (OffPtr [o2] (Addr {y} _))) => (ConstBool [x == y && o1 == o2])
(NeqPtr (Addr {x} _) (Addr {y} _)) => (ConstBool [x != y])
(NeqPtr (Addr {x} _) (OffPtr [o] (Addr {y} _))) => (ConstBool [x != y || o != 0])
(NeqPtr (OffPtr [o1] (Addr {x} _)) (OffPtr [o2] (Addr {y} _))) => (ConstBool [x != y || o1 != o2])
(EqPtr (LocalAddr {x} _ _) (LocalAddr {y} _ _)) => (ConstBool [x == y])
(EqPtr (LocalAddr {x} _ _) (OffPtr [o] (LocalAddr {y} _ _))) => (ConstBool [x == y && o == 0])
(EqPtr (OffPtr [o1] (LocalAddr {x} _ _)) (OffPtr [o2] (LocalAddr {y} _ _))) => (ConstBool [x == y && o1 == o2])
(NeqPtr (LocalAddr {x} _ _) (LocalAddr {y} _ _)) => (ConstBool [x != y])
(NeqPtr (LocalAddr {x} _ _) (OffPtr [o] (LocalAddr {y} _ _))) => (ConstBool [x != y || o != 0])
(NeqPtr (OffPtr [o1] (LocalAddr {x} _ _)) (OffPtr [o2] (LocalAddr {y} _ _))) => (ConstBool [x != y || o1 != o2])
(EqPtr (OffPtr [o1] p1) p2) && isSamePtr(p1, p2) => (ConstBool [o1 == 0])
(NeqPtr (OffPtr [o1] p1) p2) && isSamePtr(p1, p2) => (ConstBool [o1 != 0])
(EqPtr (OffPtr [o1] p1) (OffPtr [o2] p2)) && isSamePtr(p1, p2) => (ConstBool [o1 == o2])
(NeqPtr (OffPtr [o1] p1) (OffPtr [o2] p2)) && isSamePtr(p1, p2) => (ConstBool [o1 != o2])
(EqPtr (Const(32|64) [c]) (Const(32|64) [d])) => (ConstBool [c == d])
(NeqPtr (Const(32|64) [c]) (Const(32|64) [d])) => (ConstBool [c != d])
(EqPtr (Convert (Addr {x} _) _) (Addr {y} _)) => (ConstBool [x==y])
(NeqPtr (Convert (Addr {x} _) _) (Addr {y} _)) => (ConstBool [x!=y])
(EqPtr (LocalAddr _ _) (Addr _)) => (ConstBool [false])
(EqPtr (OffPtr (LocalAddr _ _)) (Addr _)) => (ConstBool [false])
(EqPtr (LocalAddr _ _) (OffPtr (Addr _))) => (ConstBool [false])
(EqPtr (OffPtr (LocalAddr _ _)) (OffPtr (Addr _))) => (ConstBool [false])
(NeqPtr (LocalAddr _ _) (Addr _)) => (ConstBool [true])
(NeqPtr (OffPtr (LocalAddr _ _)) (Addr _)) => (ConstBool [true])
(NeqPtr (LocalAddr _ _) (OffPtr (Addr _))) => (ConstBool [true])
(NeqPtr (OffPtr (LocalAddr _ _)) (OffPtr (Addr _))) => (ConstBool [true])
// Simplify address comparisons.
(EqPtr (AddPtr p1 o1) p2) && isSamePtr(p1, p2) => (Not (IsNonNil o1))
(NeqPtr (AddPtr p1 o1) p2) && isSamePtr(p1, p2) => (IsNonNil o1)
(EqPtr (Const(32|64) [0]) p) => (Not (IsNonNil p))
(NeqPtr (Const(32|64) [0]) p) => (IsNonNil p)
(EqPtr (ConstNil) p) => (Not (IsNonNil p))
(NeqPtr (ConstNil) p) => (IsNonNil p)
// Evaluate constant user nil checks.
(IsNonNil (ConstNil)) => (ConstBool [false])
(IsNonNil (Const(32|64) [c])) => (ConstBool [c != 0])
(IsNonNil (Addr _) ) => (ConstBool [true])
(IsNonNil (Convert (Addr _) _)) => (ConstBool [true])
(IsNonNil (LocalAddr _ _)) => (ConstBool [true])
// Inline small or disjoint runtime.memmove calls with constant length.
// See the comment in op Move in genericOps.go for discussion of the type.
//
// Note that we've lost any knowledge of the type and alignment requirements
// of the source and destination. We only know the size, and that the type
// contains no pointers.
// The type of the move is not necessarily v.Args[0].Type().Elem()!
// See issue 55122 for details.
//
// Because expand calls runs after prove, constants useful to this pattern may not appear.
// Both versions need to exist; the memory and register variants.
//
// Match post-expansion calls, memory version.
(SelectN [0] call:(StaticCall {sym} s1:(Store _ (Const(64|32) [sz]) s2:(Store _ src s3:(Store {t} _ dst mem)))))
&& sz >= 0
&& isSameCall(sym, "runtime.memmove")
&& s1.Uses == 1 && s2.Uses == 1 && s3.Uses == 1
&& isInlinableMemmove(dst, src, int64(sz), config)
&& clobber(s1, s2, s3, call)
=> (Move {types.Types[types.TUINT8]} [int64(sz)] dst src mem)
// Match post-expansion calls, register version.
(SelectN [0] call:(StaticCall {sym} dst src (Const(64|32) [sz]) mem))
&& sz >= 0
&& call.Uses == 1 // this will exclude all calls with results
&& isSameCall(sym, "runtime.memmove")
&& isInlinableMemmove(dst, src, int64(sz), config)
&& clobber(call)
=> (Move {types.Types[types.TUINT8]} [int64(sz)] dst src mem)
// Match pre-expansion calls.
(SelectN [0] call:(StaticLECall {sym} dst src (Const(64|32) [sz]) mem))
&& sz >= 0
&& call.Uses == 1 // this will exclude all calls with results
&& isSameCall(sym, "runtime.memmove")
&& isInlinableMemmove(dst, src, int64(sz), config)
&& clobber(call)
=> (Move {types.Types[types.TUINT8]} [int64(sz)] dst src mem)
// De-virtualize late-expanded interface calls into late-expanded static calls.
(InterLECall [argsize] {auxCall} (Addr {fn} (SB)) ___) => devirtLECall(v, fn.(*obj.LSym))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Move and Zero optimizations.
// Move source and destination may overlap.
// Convert Moves into Zeros when the source is known to be zeros.
(Move {t} [n] dst1 src mem:(Zero {t} [n] dst2 _)) && isSamePtr(src, dst2)
=> (Zero {t} [n] dst1 mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Move {t} [n] dst1 src mem:(VarDef (Zero {t} [n] dst0 _))) && isSamePtr(src, dst0)
=> (Zero {t} [n] dst1 mem)
cmd/compile: optimize Move with all-zero ro sym src to Zero We set up static symbols during walk that we later make copies of to initialize local variables. It is difficult to ascertain at that time exactly when copying a symbol is profitable vs locally initializing an autotmp. During SSA, we are much better placed to optimize. This change recognizes when we are copying from a global readonly all-zero symbol and replaces it with direct zeroing. This often allows the all-zero symbol to be deadcode eliminated at link time. This is not ideal--it makes for large object files, and longer link times--but it is the cleanest fix I could find. This makes the final binary for the program in #38554 shrink from >500mb to ~2.2mb. It also shrinks the standard binaries: file before after Δ % addr2line 4412496 4404304 -8192 -0.186% buildid 2893816 2889720 -4096 -0.142% cgo 4841048 4832856 -8192 -0.169% compile 19926480 19922432 -4048 -0.020% cover 5281816 5277720 -4096 -0.078% link 6734648 6730552 -4096 -0.061% nm 4366240 4358048 -8192 -0.188% objdump 4755968 4747776 -8192 -0.172% pprof 14653060 14612100 -40960 -0.280% trace 11805940 11777268 -28672 -0.243% vet 7185560 7181416 -4144 -0.058% total 113588440 113465560 -122880 -0.108% And not just by removing unnecessary symbols; the program text shrinks a bit as well. Fixes #38554 Change-Id: I8381ae6084ae145a5e0cd9410c451e52c0dc51c8 Reviewed-on: https://go-review.googlesource.com/c/go/+/229704 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Keith Randall <khr@golang.org>
2020-04-23 13:28:14 -07:00
(Move {t} [n] dst (Addr {sym} (SB)) mem) && symIsROZero(sym) => (Zero {t} [n] dst mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Don't Store to variables that are about to be overwritten by Move/Zero.
(Zero {t1} [n] p1 store:(Store {t2} (OffPtr [o2] p2) _ mem))
&& isSamePtr(p1, p2) && store.Uses == 1
&& n >= o2 + t2.Size()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& clobber(store)
=> (Zero {t1} [n] p1 mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Move {t1} [n] dst1 src1 store:(Store {t2} op:(OffPtr [o2] dst2) _ mem))
&& isSamePtr(dst1, dst2) && store.Uses == 1
&& n >= o2 + t2.Size()
&& disjoint(src1, n, op, t2.Size())
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& clobber(store)
=> (Move {t1} [n] dst1 src1 mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Don't Move to variables that are immediately completely overwritten.
(Zero {t} [n] dst1 move:(Move {t} [n] dst2 _ mem))
&& move.Uses == 1
&& isSamePtr(dst1, dst2)
&& clobber(move)
=> (Zero {t} [n] dst1 mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Move {t} [n] dst1 src1 move:(Move {t} [n] dst2 _ mem))
&& move.Uses == 1
&& isSamePtr(dst1, dst2) && disjoint(src1, n, dst2, n)
&& clobber(move)
=> (Move {t} [n] dst1 src1 mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Zero {t} [n] dst1 vardef:(VarDef {x} move:(Move {t} [n] dst2 _ mem)))
&& move.Uses == 1 && vardef.Uses == 1
&& isSamePtr(dst1, dst2)
&& clobber(move, vardef)
=> (Zero {t} [n] dst1 (VarDef {x} mem))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Move {t} [n] dst1 src1 vardef:(VarDef {x} move:(Move {t} [n] dst2 _ mem)))
&& move.Uses == 1 && vardef.Uses == 1
&& isSamePtr(dst1, dst2) && disjoint(src1, n, dst2, n)
&& clobber(move, vardef)
=> (Move {t} [n] dst1 src1 (VarDef {x} mem))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} op1:(OffPtr [o1] p1) d1
m2:(Store {t2} op2:(OffPtr [0] p2) d2
m3:(Move [n] p3 _ mem)))
&& m2.Uses == 1 && m3.Uses == 1
&& o1 == t2.Size()
&& n == t2.Size() + t1.Size()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3)
&& clobber(m2, m3)
=> (Store {t1} op1 d1 (Store {t2} op2 d2 mem))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} op1:(OffPtr [o1] p1) d1
m2:(Store {t2} op2:(OffPtr [o2] p2) d2
m3:(Store {t3} op3:(OffPtr [0] p3) d3
m4:(Move [n] p4 _ mem))))
&& m2.Uses == 1 && m3.Uses == 1 && m4.Uses == 1
&& o2 == t3.Size()
&& o1-o2 == t2.Size()
&& n == t3.Size() + t2.Size() + t1.Size()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4)
&& clobber(m2, m3, m4)
=> (Store {t1} op1 d1 (Store {t2} op2 d2 (Store {t3} op3 d3 mem)))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} op1:(OffPtr [o1] p1) d1
m2:(Store {t2} op2:(OffPtr [o2] p2) d2
m3:(Store {t3} op3:(OffPtr [o3] p3) d3
m4:(Store {t4} op4:(OffPtr [0] p4) d4
m5:(Move [n] p5 _ mem)))))
&& m2.Uses == 1 && m3.Uses == 1 && m4.Uses == 1 && m5.Uses == 1
&& o3 == t4.Size()
&& o2-o3 == t3.Size()
&& o1-o2 == t2.Size()
&& n == t4.Size() + t3.Size() + t2.Size() + t1.Size()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5)
&& clobber(m2, m3, m4, m5)
=> (Store {t1} op1 d1 (Store {t2} op2 d2 (Store {t3} op3 d3 (Store {t4} op4 d4 mem))))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Don't Zero variables that are immediately completely overwritten
// before being accessed.
(Move {t} [n] dst1 src1 zero:(Zero {t} [n] dst2 mem))
&& zero.Uses == 1
&& isSamePtr(dst1, dst2) && disjoint(src1, n, dst2, n)
&& clobber(zero)
=> (Move {t} [n] dst1 src1 mem)
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Move {t} [n] dst1 src1 vardef:(VarDef {x} zero:(Zero {t} [n] dst2 mem)))
&& zero.Uses == 1 && vardef.Uses == 1
&& isSamePtr(dst1, dst2) && disjoint(src1, n, dst2, n)
&& clobber(zero, vardef)
=> (Move {t} [n] dst1 src1 (VarDef {x} mem))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} op1:(OffPtr [o1] p1) d1
m2:(Store {t2} op2:(OffPtr [0] p2) d2
m3:(Zero [n] p3 mem)))
&& m2.Uses == 1 && m3.Uses == 1
&& o1 == t2.Size()
&& n == t2.Size() + t1.Size()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3)
&& clobber(m2, m3)
=> (Store {t1} op1 d1 (Store {t2} op2 d2 mem))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} op1:(OffPtr [o1] p1) d1
m2:(Store {t2} op2:(OffPtr [o2] p2) d2
m3:(Store {t3} op3:(OffPtr [0] p3) d3
m4:(Zero [n] p4 mem))))
&& m2.Uses == 1 && m3.Uses == 1 && m4.Uses == 1
&& o2 == t3.Size()
&& o1-o2 == t2.Size()
&& n == t3.Size() + t2.Size() + t1.Size()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4)
&& clobber(m2, m3, m4)
=> (Store {t1} op1 d1 (Store {t2} op2 d2 (Store {t3} op3 d3 mem)))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t1} op1:(OffPtr [o1] p1) d1
m2:(Store {t2} op2:(OffPtr [o2] p2) d2
m3:(Store {t3} op3:(OffPtr [o3] p3) d3
m4:(Store {t4} op4:(OffPtr [0] p4) d4
m5:(Zero [n] p5 mem)))))
&& m2.Uses == 1 && m3.Uses == 1 && m4.Uses == 1 && m5.Uses == 1
&& o3 == t4.Size()
&& o2-o3 == t3.Size()
&& o1-o2 == t2.Size()
&& n == t4.Size() + t3.Size() + t2.Size() + t1.Size()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5)
&& clobber(m2, m3, m4, m5)
=> (Store {t1} op1 d1 (Store {t2} op2 d2 (Store {t3} op3 d3 (Store {t4} op4 d4 mem))))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Don't Move from memory if the values are likely to already be
// in registers.
(Move {t1} [n] dst p1
mem:(Store {t2} op2:(OffPtr <tt2> [o2] p2) d1
(Store {t3} op3:(OffPtr <tt3> [0] p3) d2 _)))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& o2 == t3.Size()
&& n == t2.Size() + t3.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(Store {t3} (OffPtr <tt3> [0] dst) d2 mem))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Move {t1} [n] dst p1
mem:(Store {t2} op2:(OffPtr <tt2> [o2] p2) d1
(Store {t3} op3:(OffPtr <tt3> [o3] p3) d2
(Store {t4} op4:(OffPtr <tt4> [0] p4) d3 _))))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& o3 == t4.Size()
&& o2-o3 == t3.Size()
&& n == t2.Size() + t3.Size() + t4.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Store {t4} (OffPtr <tt4> [0] dst) d3 mem)))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Move {t1} [n] dst p1
mem:(Store {t2} op2:(OffPtr <tt2> [o2] p2) d1
(Store {t3} op3:(OffPtr <tt3> [o3] p3) d2
(Store {t4} op4:(OffPtr <tt4> [o4] p4) d3
(Store {t5} op5:(OffPtr <tt5> [0] p5) d4 _)))))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
&& t5.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& registerizable(b, t5)
&& o4 == t5.Size()
&& o3-o4 == t4.Size()
&& o2-o3 == t3.Size()
&& n == t2.Size() + t3.Size() + t4.Size() + t5.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Store {t4} (OffPtr <tt4> [o4] dst) d3
(Store {t5} (OffPtr <tt5> [0] dst) d4 mem))))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Same thing but with VarDef in the middle.
(Move {t1} [n] dst p1
mem:(VarDef
(Store {t2} op2:(OffPtr <tt2> [o2] p2) d1
(Store {t3} op3:(OffPtr <tt3> [0] p3) d2 _))))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& o2 == t3.Size()
&& n == t2.Size() + t3.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(Store {t3} (OffPtr <tt3> [0] dst) d2 mem))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Move {t1} [n] dst p1
mem:(VarDef
(Store {t2} op2:(OffPtr <tt2> [o2] p2) d1
(Store {t3} op3:(OffPtr <tt3> [o3] p3) d2
(Store {t4} op4:(OffPtr <tt4> [0] p4) d3 _)))))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& o3 == t4.Size()
&& o2-o3 == t3.Size()
&& n == t2.Size() + t3.Size() + t4.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Store {t4} (OffPtr <tt4> [0] dst) d3 mem)))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Move {t1} [n] dst p1
mem:(VarDef
(Store {t2} op2:(OffPtr <tt2> [o2] p2) d1
(Store {t3} op3:(OffPtr <tt3> [o3] p3) d2
(Store {t4} op4:(OffPtr <tt4> [o4] p4) d3
(Store {t5} op5:(OffPtr <tt5> [0] p5) d4 _))))))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
&& t5.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& registerizable(b, t5)
&& o4 == t5.Size()
&& o3-o4 == t4.Size()
&& o2-o3 == t3.Size()
&& n == t2.Size() + t3.Size() + t4.Size() + t5.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Store {t4} (OffPtr <tt4> [o4] dst) d3
(Store {t5} (OffPtr <tt5> [0] dst) d4 mem))))
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
// Prefer to Zero and Store than to Move.
(Move {t1} [n] dst p1
mem:(Store {t2} op2:(OffPtr <tt2> [o2] p2) d1
(Zero {t3} [n] p3 _)))
&& isSamePtr(p1, p2) && isSamePtr(p2, p3)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& n >= o2 + t2.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Zero {t1} [n] dst mem))
(Move {t1} [n] dst p1
mem:(Store {t2} (OffPtr <tt2> [o2] p2) d1
(Store {t3} (OffPtr <tt3> [o3] p3) d2
(Zero {t4} [n] p4 _))))
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& n >= o2 + t2.Size()
&& n >= o3 + t3.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Zero {t1} [n] dst mem)))
(Move {t1} [n] dst p1
mem:(Store {t2} (OffPtr <tt2> [o2] p2) d1
(Store {t3} (OffPtr <tt3> [o3] p3) d2
(Store {t4} (OffPtr <tt4> [o4] p4) d3
(Zero {t5} [n] p5 _)))))
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
&& t5.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& n >= o2 + t2.Size()
&& n >= o3 + t3.Size()
&& n >= o4 + t4.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Store {t4} (OffPtr <tt4> [o4] dst) d3
(Zero {t1} [n] dst mem))))
(Move {t1} [n] dst p1
mem:(Store {t2} (OffPtr <tt2> [o2] p2) d1
(Store {t3} (OffPtr <tt3> [o3] p3) d2
(Store {t4} (OffPtr <tt4> [o4] p4) d3
(Store {t5} (OffPtr <tt5> [o5] p5) d4
(Zero {t6} [n] p6 _))))))
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5) && isSamePtr(p5, p6)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
&& t5.Alignment() <= t1.Alignment()
&& t6.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& registerizable(b, t5)
&& n >= o2 + t2.Size()
&& n >= o3 + t3.Size()
&& n >= o4 + t4.Size()
&& n >= o5 + t5.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Store {t4} (OffPtr <tt4> [o4] dst) d3
(Store {t5} (OffPtr <tt5> [o5] dst) d4
(Zero {t1} [n] dst mem)))))
(Move {t1} [n] dst p1
mem:(VarDef
(Store {t2} op2:(OffPtr <tt2> [o2] p2) d1
(Zero {t3} [n] p3 _))))
&& isSamePtr(p1, p2) && isSamePtr(p2, p3)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& n >= o2 + t2.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Zero {t1} [n] dst mem))
(Move {t1} [n] dst p1
mem:(VarDef
(Store {t2} (OffPtr <tt2> [o2] p2) d1
(Store {t3} (OffPtr <tt3> [o3] p3) d2
(Zero {t4} [n] p4 _)))))
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& n >= o2 + t2.Size()
&& n >= o3 + t3.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Zero {t1} [n] dst mem)))
(Move {t1} [n] dst p1
mem:(VarDef
(Store {t2} (OffPtr <tt2> [o2] p2) d1
(Store {t3} (OffPtr <tt3> [o3] p3) d2
(Store {t4} (OffPtr <tt4> [o4] p4) d3
(Zero {t5} [n] p5 _))))))
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
&& t5.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& n >= o2 + t2.Size()
&& n >= o3 + t3.Size()
&& n >= o4 + t4.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Store {t4} (OffPtr <tt4> [o4] dst) d3
(Zero {t1} [n] dst mem))))
(Move {t1} [n] dst p1
mem:(VarDef
(Store {t2} (OffPtr <tt2> [o2] p2) d1
(Store {t3} (OffPtr <tt3> [o3] p3) d2
(Store {t4} (OffPtr <tt4> [o4] p4) d3
(Store {t5} (OffPtr <tt5> [o5] p5) d4
(Zero {t6} [n] p6 _)))))))
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5) && isSamePtr(p5, p6)
&& t2.Alignment() <= t1.Alignment()
&& t3.Alignment() <= t1.Alignment()
&& t4.Alignment() <= t1.Alignment()
&& t5.Alignment() <= t1.Alignment()
&& t6.Alignment() <= t1.Alignment()
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& registerizable(b, t5)
&& n >= o2 + t2.Size()
&& n >= o3 + t3.Size()
&& n >= o4 + t4.Size()
&& n >= o5 + t5.Size()
=> (Store {t2} (OffPtr <tt2> [o2] dst) d1
cmd/compile: add some generic composite type optimizations Propagate values through some wide Zero/Move operations. Among other things this allows us to optimize some kinds of array initialization. For example, the following code no longer requires a temporary be allocated on the stack. Instead it writes the values directly into the return value. func f(i uint32) [4]uint32 { return [4]uint32{i, i+1, i+2, i+3} } The return value is unnecessarily cleared but removing that is probably a task for dead store analysis (I think it needs to be able to match multiple Store ops to wide Zero ops). In order to reliably remove stack variables that are rendered unnecessary by these new rules I've added a new generic version of the unread autos elimination pass. These rules are triggered more than 5000 times when building and testing the standard library. Updates #15925 (fixes for arrays of up to 4 elements). Updates #24386 (fixes for up to 4 kept elements). Updates #24416. compilebench results: name old time/op new time/op delta Template 353ms ± 5% 359ms ± 3% ~ (p=0.143 n=10+10) Unicode 219ms ± 1% 217ms ± 4% ~ (p=0.740 n=7+10) GoTypes 1.26s ± 1% 1.26s ± 2% ~ (p=0.549 n=9+10) Compiler 6.00s ± 1% 6.08s ± 1% +1.42% (p=0.000 n=9+8) SSA 15.3s ± 2% 15.6s ± 1% +2.43% (p=0.000 n=10+10) Flate 237ms ± 2% 240ms ± 2% +1.31% (p=0.015 n=10+10) GoParser 285ms ± 1% 285ms ± 1% ~ (p=0.878 n=8+8) Reflect 797ms ± 3% 807ms ± 2% ~ (p=0.065 n=9+10) Tar 334ms ± 0% 335ms ± 4% ~ (p=0.460 n=8+10) XML 419ms ± 0% 423ms ± 1% +0.91% (p=0.001 n=7+9) StdCmd 46.0s ± 0% 46.4s ± 0% +0.85% (p=0.000 n=9+9) name old user-time/op new user-time/op delta Template 337ms ± 3% 346ms ± 5% ~ (p=0.053 n=9+10) Unicode 205ms ±10% 205ms ± 8% ~ (p=1.000 n=10+10) GoTypes 1.22s ± 2% 1.21s ± 3% ~ (p=0.436 n=10+10) Compiler 5.85s ± 1% 5.93s ± 0% +1.46% (p=0.000 n=10+8) SSA 14.9s ± 1% 15.3s ± 1% +2.62% (p=0.000 n=10+10) Flate 229ms ± 4% 228ms ± 6% ~ (p=0.796 n=10+10) GoParser 271ms ± 3% 275ms ± 4% ~ (p=0.165 n=10+10) Reflect 779ms ± 5% 775ms ± 2% ~ (p=0.971 n=10+10) Tar 317ms ± 4% 319ms ± 5% ~ (p=0.853 n=10+10) XML 404ms ± 4% 409ms ± 5% ~ (p=0.436 n=10+10) name old alloc/op new alloc/op delta Template 34.9MB ± 0% 35.0MB ± 0% +0.26% (p=0.000 n=10+10) Unicode 29.3MB ± 0% 29.3MB ± 0% +0.02% (p=0.000 n=10+10) GoTypes 115MB ± 0% 115MB ± 0% +0.30% (p=0.000 n=10+10) Compiler 519MB ± 0% 521MB ± 0% +0.30% (p=0.000 n=10+10) SSA 1.55GB ± 0% 1.57GB ± 0% +1.34% (p=0.000 n=10+9) Flate 24.1MB ± 0% 24.2MB ± 0% +0.10% (p=0.000 n=10+10) GoParser 28.1MB ± 0% 28.1MB ± 0% +0.07% (p=0.000 n=10+10) Reflect 78.7MB ± 0% 78.7MB ± 0% +0.03% (p=0.000 n=8+10) Tar 34.4MB ± 0% 34.5MB ± 0% +0.12% (p=0.000 n=10+10) XML 43.2MB ± 0% 43.2MB ± 0% +0.13% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 330k ± 0% 330k ± 0% -0.01% (p=0.017 n=10+10) Unicode 337k ± 0% 337k ± 0% +0.01% (p=0.000 n=9+10) GoTypes 1.15M ± 0% 1.15M ± 0% +0.03% (p=0.000 n=10+10) Compiler 4.77M ± 0% 4.77M ± 0% +0.03% (p=0.000 n=9+10) SSA 12.5M ± 0% 12.6M ± 0% +1.16% (p=0.000 n=10+10) Flate 221k ± 0% 221k ± 0% +0.05% (p=0.000 n=9+10) GoParser 275k ± 0% 275k ± 0% +0.01% (p=0.014 n=10+9) Reflect 944k ± 0% 944k ± 0% -0.02% (p=0.000 n=10+10) Tar 324k ± 0% 323k ± 0% -0.12% (p=0.000 n=10+10) XML 384k ± 0% 384k ± 0% -0.01% (p=0.001 n=10+10) name old object-bytes new object-bytes delta Template 476kB ± 0% 476kB ± 0% -0.04% (p=0.000 n=10+10) Unicode 218kB ± 0% 218kB ± 0% ~ (all equal) GoTypes 1.58MB ± 0% 1.58MB ± 0% -0.04% (p=0.000 n=10+10) Compiler 6.25MB ± 0% 6.24MB ± 0% -0.09% (p=0.000 n=10+10) SSA 15.9MB ± 0% 16.1MB ± 0% +1.22% (p=0.000 n=10+10) Flate 304kB ± 0% 304kB ± 0% -0.13% (p=0.000 n=10+10) GoParser 370kB ± 0% 370kB ± 0% -0.00% (p=0.000 n=10+10) Reflect 1.27MB ± 0% 1.27MB ± 0% -0.12% (p=0.000 n=10+10) Tar 421kB ± 0% 419kB ± 0% -0.64% (p=0.000 n=10+10) XML 518kB ± 0% 517kB ± 0% -0.12% (p=0.000 n=10+10) name old export-bytes new export-bytes delta Template 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) Unicode 6.52kB ± 0% 6.52kB ± 0% ~ (all equal) GoTypes 29.2kB ± 0% 29.2kB ± 0% ~ (all equal) Compiler 88.0kB ± 0% 88.0kB ± 0% ~ (all equal) SSA 109kB ± 0% 109kB ± 0% ~ (all equal) Flate 4.49kB ± 0% 4.49kB ± 0% ~ (all equal) GoParser 8.10kB ± 0% 8.10kB ± 0% ~ (all equal) Reflect 7.71kB ± 0% 7.71kB ± 0% ~ (all equal) Tar 9.15kB ± 0% 9.15kB ± 0% ~ (all equal) XML 12.3kB ± 0% 12.3kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 676kB ± 0% 672kB ± 0% -0.59% (p=0.000 n=10+10) CmdGoSize 7.26MB ± 0% 7.24MB ± 0% -0.18% (p=0.000 n=10+10) name old data-bytes new data-bytes delta HelloSize 10.2kB ± 0% 10.2kB ± 0% ~ (all equal) CmdGoSize 248kB ± 0% 248kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 125kB ± 0% 125kB ± 0% ~ (all equal) CmdGoSize 145kB ± 0% 145kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.46MB ± 0% 1.45MB ± 0% -0.31% (p=0.000 n=10+10) CmdGoSize 14.7MB ± 0% 14.7MB ± 0% -0.17% (p=0.000 n=10+10) Change-Id: Ic72b0c189dd542f391e1c9ab88a76e9148dc4285 Reviewed-on: https://go-review.googlesource.com/106495 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-11 22:47:24 +01:00
(Store {t3} (OffPtr <tt3> [o3] dst) d2
(Store {t4} (OffPtr <tt4> [o4] dst) d3
(Store {t5} (OffPtr <tt5> [o5] dst) d4
(Zero {t1} [n] dst mem)))))
(SelectN [0] call:(StaticLECall {sym} a x)) && needRaceCleanup(sym, call) && clobber(call) => x
(SelectN [0] call:(StaticLECall {sym} x)) && needRaceCleanup(sym, call) && clobber(call) => x
cmd/compile: move argument stack construction to SSA generation The goal of this change is to move work from walk to SSA, and simplify things along the way. This is hard to accomplish cleanly with small incremental changes, so this large commit message aims to provide a roadmap to the diff. High level description: Prior to this change, walk was responsible for constructing (most of) the stack for function calls. ascompatte gathered variadic arguments into a slice. It also rewrote n.List from a list of arguments to a list of assignments to stack slots. ascompatte was called multiple times to handle the receiver in a method call. reorder1 then introduced temporaries into n.List as needed to avoid smashing the stack. adjustargs then made extra stack space for go/defer args as needed. Node to SSA construction evaluated all the statements in n.List, and issued the function call, assuming that the stack was correctly constructed. Intrinsic calls had to dig around inside n.List to extract the arguments, since intrinsics don't use the stack to make function calls. This change moves stack construction to the SSA construction phase. ascompatte, now called walkParams, does all the work that ascompatte and reorder1 did. It handles variadic arguments, inserts the method receiver if needed, and allocates temporaries. It does not, however, make any assignments to stack slots. Instead, it moves the function arguments to n.Rlist, leaving assignments to temporaries in n.List. (It would be better to use Ninit instead of List; future work.) During SSA construction, after doing all the temporary assignments in n.List, the function arguments are assigned to stack slots by constructing the appropriate SSA Value, using (*state).storeArg. SSA construction also now handles adjustments for go/defer args. This change also simplifies intrinsic calls, since we no longer need to undo walk's work. Along the way, we simplify nodarg by pushing the fp==1 case to its callers, where it fits nicely. Generated code differences: There were a few optimizations applied along the way, the old way. f(g()) was rewritten to do a block copy of function results to function arguments. And reorder1 avoided introducing the final "save the stack" temporary in n.List. The f(g()) block copy optimization never actually triggered; the order pass rewrote away g(), so that has been removed. SSA optimizations mostly obviated the need for reorder1's optimization of avoiding the final temporary. The exception was when the temporary's type was not SSA-able; in that case, we got a Move into an autotmp and then an immediate Move onto the stack, with the autotmp never read or used again. This change introduces a new rewrite rule to detect such pointless double Moves and collapse them into a single Move. This is actually more powerful than the original optimization, since the original optimization relied on the imprecise Node.HasCall calculation. The other significant difference in the generated code is that the stack is now constructed completely in SP-offset order. Prior to this change, the stack was constructed somewhat haphazardly: first the final argument that Node.HasCall deemed to require a temporary, then other arguments, then the method receiver, then the defer/go args. SP-offset is probably a good default order. See future work. There are a few minor object file size changes as a result of this change. I investigated some regressions in early versions of this change. One regression (in archive/tar) was the addition of a single CMPQ instruction, which would be eliminated were this TODO from flagalloc to be done: // TODO: Remove original instructions if they are never used. One regression (in text/template) was an ADDQconstmodify that is now a regular MOVQLoad+ADDQconst+MOVQStore, due to an unlucky change in the order in which arguments are written. The argument change order can also now be luckier, so this appears to be a wash. All in all, though there will be minor winners and losers, this change appears to be performance neutral. Future work: Move loading the result of function calls to SSA construction; eliminate OINDREGSP. Consider pushing stack construction deeper into SSA world, perhaps in an arch-specific pass. Among other benefits, this would make it easier to transition to a new calling convention. This would require rethinking the handling of stack conflicts and is non-trivial. Figure out some clean way to indicate that stack construction Stores/Moves do not alias each other, so that subsequent passes may do things like CSE+tighten shared stack setup, do DSE using non-first Stores, etc. This would allow us to eliminate the minor text/template regression. Possibly make assignments to stack slots not treated as statements by DWARF. Compiler benchmarks: name old time/op new time/op delta Template 182ms ± 2% 179ms ± 2% -1.69% (p=0.000 n=47+48) Unicode 86.3ms ± 5% 85.1ms ± 4% -1.36% (p=0.001 n=50+50) GoTypes 646ms ± 1% 642ms ± 1% -0.63% (p=0.000 n=49+48) Compiler 2.89s ± 1% 2.86s ± 2% -1.36% (p=0.000 n=48+50) SSA 8.47s ± 1% 8.37s ± 2% -1.22% (p=0.000 n=47+50) Flate 122ms ± 2% 121ms ± 2% -0.66% (p=0.000 n=47+45) GoParser 147ms ± 2% 146ms ± 2% -0.53% (p=0.006 n=46+49) Reflect 406ms ± 2% 403ms ± 2% -0.76% (p=0.000 n=48+43) Tar 162ms ± 3% 162ms ± 4% ~ (p=0.191 n=46+50) XML 223ms ± 2% 222ms ± 2% -0.37% (p=0.031 n=45+49) [Geo mean] 382ms 378ms -0.89% name old user-time/op new user-time/op delta Template 219ms ± 3% 216ms ± 3% -1.56% (p=0.000 n=50+48) Unicode 109ms ± 6% 109ms ± 5% ~ (p=0.190 n=50+49) GoTypes 836ms ± 2% 828ms ± 2% -0.96% (p=0.000 n=49+48) Compiler 3.87s ± 2% 3.80s ± 1% -1.81% (p=0.000 n=49+46) SSA 12.0s ± 1% 11.8s ± 1% -2.01% (p=0.000 n=48+50) Flate 142ms ± 3% 141ms ± 3% -0.85% (p=0.003 n=50+48) GoParser 178ms ± 4% 175ms ± 4% -1.66% (p=0.000 n=48+46) Reflect 520ms ± 2% 512ms ± 2% -1.44% (p=0.000 n=45+48) Tar 200ms ± 3% 198ms ± 4% -0.61% (p=0.037 n=47+50) XML 277ms ± 3% 275ms ± 3% -0.85% (p=0.000 n=49+48) [Geo mean] 482ms 476ms -1.23% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 35.3MB ± 0% -2.18% (p=0.008 n=5+5) Unicode 29.8MB ± 0% 29.3MB ± 0% -1.58% (p=0.008 n=5+5) GoTypes 125MB ± 0% 123MB ± 0% -2.13% (p=0.008 n=5+5) Compiler 531MB ± 0% 513MB ± 0% -3.40% (p=0.008 n=5+5) SSA 2.00GB ± 0% 1.93GB ± 0% -3.34% (p=0.008 n=5+5) Flate 24.5MB ± 0% 24.3MB ± 0% -1.18% (p=0.008 n=5+5) GoParser 29.4MB ± 0% 28.7MB ± 0% -2.34% (p=0.008 n=5+5) Reflect 87.1MB ± 0% 86.0MB ± 0% -1.33% (p=0.008 n=5+5) Tar 35.3MB ± 0% 34.8MB ± 0% -1.44% (p=0.008 n=5+5) XML 47.9MB ± 0% 47.1MB ± 0% -1.86% (p=0.008 n=5+5) [Geo mean] 82.8MB 81.1MB -2.08% name old allocs/op new allocs/op delta Template 352k ± 0% 347k ± 0% -1.32% (p=0.008 n=5+5) Unicode 342k ± 0% 339k ± 0% -0.66% (p=0.008 n=5+5) GoTypes 1.29M ± 0% 1.27M ± 0% -1.30% (p=0.008 n=5+5) Compiler 4.98M ± 0% 4.87M ± 0% -2.14% (p=0.008 n=5+5) SSA 15.7M ± 0% 15.2M ± 0% -2.86% (p=0.008 n=5+5) Flate 233k ± 0% 231k ± 0% -0.83% (p=0.008 n=5+5) GoParser 296k ± 0% 291k ± 0% -1.54% (p=0.016 n=5+4) Reflect 1.05M ± 0% 1.04M ± 0% -0.65% (p=0.008 n=5+5) Tar 343k ± 0% 339k ± 0% -0.97% (p=0.008 n=5+5) XML 432k ± 0% 426k ± 0% -1.19% (p=0.008 n=5+5) [Geo mean] 815k 804k -1.35% name old object-bytes new object-bytes delta Template 505kB ± 0% 505kB ± 0% -0.01% (p=0.008 n=5+5) Unicode 224kB ± 0% 224kB ± 0% ~ (all equal) GoTypes 1.82MB ± 0% 1.83MB ± 0% +0.06% (p=0.008 n=5+5) Flate 324kB ± 0% 324kB ± 0% +0.00% (p=0.008 n=5+5) GoParser 402kB ± 0% 402kB ± 0% +0.04% (p=0.008 n=5+5) Reflect 1.39MB ± 0% 1.39MB ± 0% -0.01% (p=0.008 n=5+5) Tar 449kB ± 0% 449kB ± 0% -0.02% (p=0.008 n=5+5) XML 598kB ± 0% 597kB ± 0% -0.05% (p=0.008 n=5+5) Change-Id: Ifc9d5c1bd01f90171414b8fb18ffe2290d271143 Reviewed-on: https://go-review.googlesource.com/c/114797 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2018-05-06 12:58:53 -07:00
// When rewriting append to growslice, we use as the new length the result of
// growslice so that we don't have to spill/restore the new length around the growslice call.
// The exception here is that if the new length is a constant, avoiding spilling it
// is pointless and its constantness is sometimes useful for subsequent optimizations.
// See issue 56440.
// Note there are 2 rules here, one for the pre-decomposed []T result and one for
// the post-decomposed (*T,int,int) result. (The latter is generated after call expansion.)
(SliceLen (SelectN [0] (StaticLECall {sym} _ newLen:(Const(64|32)) _ _ _ _))) && isSameCall(sym, "runtime.growslice") => newLen
(SelectN [1] (StaticCall {sym} _ newLen:(Const(64|32)) _ _ _ _)) && v.Type.IsInteger() && isSameCall(sym, "runtime.growslice") => newLen
cmd/compile: move argument stack construction to SSA generation The goal of this change is to move work from walk to SSA, and simplify things along the way. This is hard to accomplish cleanly with small incremental changes, so this large commit message aims to provide a roadmap to the diff. High level description: Prior to this change, walk was responsible for constructing (most of) the stack for function calls. ascompatte gathered variadic arguments into a slice. It also rewrote n.List from a list of arguments to a list of assignments to stack slots. ascompatte was called multiple times to handle the receiver in a method call. reorder1 then introduced temporaries into n.List as needed to avoid smashing the stack. adjustargs then made extra stack space for go/defer args as needed. Node to SSA construction evaluated all the statements in n.List, and issued the function call, assuming that the stack was correctly constructed. Intrinsic calls had to dig around inside n.List to extract the arguments, since intrinsics don't use the stack to make function calls. This change moves stack construction to the SSA construction phase. ascompatte, now called walkParams, does all the work that ascompatte and reorder1 did. It handles variadic arguments, inserts the method receiver if needed, and allocates temporaries. It does not, however, make any assignments to stack slots. Instead, it moves the function arguments to n.Rlist, leaving assignments to temporaries in n.List. (It would be better to use Ninit instead of List; future work.) During SSA construction, after doing all the temporary assignments in n.List, the function arguments are assigned to stack slots by constructing the appropriate SSA Value, using (*state).storeArg. SSA construction also now handles adjustments for go/defer args. This change also simplifies intrinsic calls, since we no longer need to undo walk's work. Along the way, we simplify nodarg by pushing the fp==1 case to its callers, where it fits nicely. Generated code differences: There were a few optimizations applied along the way, the old way. f(g()) was rewritten to do a block copy of function results to function arguments. And reorder1 avoided introducing the final "save the stack" temporary in n.List. The f(g()) block copy optimization never actually triggered; the order pass rewrote away g(), so that has been removed. SSA optimizations mostly obviated the need for reorder1's optimization of avoiding the final temporary. The exception was when the temporary's type was not SSA-able; in that case, we got a Move into an autotmp and then an immediate Move onto the stack, with the autotmp never read or used again. This change introduces a new rewrite rule to detect such pointless double Moves and collapse them into a single Move. This is actually more powerful than the original optimization, since the original optimization relied on the imprecise Node.HasCall calculation. The other significant difference in the generated code is that the stack is now constructed completely in SP-offset order. Prior to this change, the stack was constructed somewhat haphazardly: first the final argument that Node.HasCall deemed to require a temporary, then other arguments, then the method receiver, then the defer/go args. SP-offset is probably a good default order. See future work. There are a few minor object file size changes as a result of this change. I investigated some regressions in early versions of this change. One regression (in archive/tar) was the addition of a single CMPQ instruction, which would be eliminated were this TODO from flagalloc to be done: // TODO: Remove original instructions if they are never used. One regression (in text/template) was an ADDQconstmodify that is now a regular MOVQLoad+ADDQconst+MOVQStore, due to an unlucky change in the order in which arguments are written. The argument change order can also now be luckier, so this appears to be a wash. All in all, though there will be minor winners and losers, this change appears to be performance neutral. Future work: Move loading the result of function calls to SSA construction; eliminate OINDREGSP. Consider pushing stack construction deeper into SSA world, perhaps in an arch-specific pass. Among other benefits, this would make it easier to transition to a new calling convention. This would require rethinking the handling of stack conflicts and is non-trivial. Figure out some clean way to indicate that stack construction Stores/Moves do not alias each other, so that subsequent passes may do things like CSE+tighten shared stack setup, do DSE using non-first Stores, etc. This would allow us to eliminate the minor text/template regression. Possibly make assignments to stack slots not treated as statements by DWARF. Compiler benchmarks: name old time/op new time/op delta Template 182ms ± 2% 179ms ± 2% -1.69% (p=0.000 n=47+48) Unicode 86.3ms ± 5% 85.1ms ± 4% -1.36% (p=0.001 n=50+50) GoTypes 646ms ± 1% 642ms ± 1% -0.63% (p=0.000 n=49+48) Compiler 2.89s ± 1% 2.86s ± 2% -1.36% (p=0.000 n=48+50) SSA 8.47s ± 1% 8.37s ± 2% -1.22% (p=0.000 n=47+50) Flate 122ms ± 2% 121ms ± 2% -0.66% (p=0.000 n=47+45) GoParser 147ms ± 2% 146ms ± 2% -0.53% (p=0.006 n=46+49) Reflect 406ms ± 2% 403ms ± 2% -0.76% (p=0.000 n=48+43) Tar 162ms ± 3% 162ms ± 4% ~ (p=0.191 n=46+50) XML 223ms ± 2% 222ms ± 2% -0.37% (p=0.031 n=45+49) [Geo mean] 382ms 378ms -0.89% name old user-time/op new user-time/op delta Template 219ms ± 3% 216ms ± 3% -1.56% (p=0.000 n=50+48) Unicode 109ms ± 6% 109ms ± 5% ~ (p=0.190 n=50+49) GoTypes 836ms ± 2% 828ms ± 2% -0.96% (p=0.000 n=49+48) Compiler 3.87s ± 2% 3.80s ± 1% -1.81% (p=0.000 n=49+46) SSA 12.0s ± 1% 11.8s ± 1% -2.01% (p=0.000 n=48+50) Flate 142ms ± 3% 141ms ± 3% -0.85% (p=0.003 n=50+48) GoParser 178ms ± 4% 175ms ± 4% -1.66% (p=0.000 n=48+46) Reflect 520ms ± 2% 512ms ± 2% -1.44% (p=0.000 n=45+48) Tar 200ms ± 3% 198ms ± 4% -0.61% (p=0.037 n=47+50) XML 277ms ± 3% 275ms ± 3% -0.85% (p=0.000 n=49+48) [Geo mean] 482ms 476ms -1.23% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 35.3MB ± 0% -2.18% (p=0.008 n=5+5) Unicode 29.8MB ± 0% 29.3MB ± 0% -1.58% (p=0.008 n=5+5) GoTypes 125MB ± 0% 123MB ± 0% -2.13% (p=0.008 n=5+5) Compiler 531MB ± 0% 513MB ± 0% -3.40% (p=0.008 n=5+5) SSA 2.00GB ± 0% 1.93GB ± 0% -3.34% (p=0.008 n=5+5) Flate 24.5MB ± 0% 24.3MB ± 0% -1.18% (p=0.008 n=5+5) GoParser 29.4MB ± 0% 28.7MB ± 0% -2.34% (p=0.008 n=5+5) Reflect 87.1MB ± 0% 86.0MB ± 0% -1.33% (p=0.008 n=5+5) Tar 35.3MB ± 0% 34.8MB ± 0% -1.44% (p=0.008 n=5+5) XML 47.9MB ± 0% 47.1MB ± 0% -1.86% (p=0.008 n=5+5) [Geo mean] 82.8MB 81.1MB -2.08% name old allocs/op new allocs/op delta Template 352k ± 0% 347k ± 0% -1.32% (p=0.008 n=5+5) Unicode 342k ± 0% 339k ± 0% -0.66% (p=0.008 n=5+5) GoTypes 1.29M ± 0% 1.27M ± 0% -1.30% (p=0.008 n=5+5) Compiler 4.98M ± 0% 4.87M ± 0% -2.14% (p=0.008 n=5+5) SSA 15.7M ± 0% 15.2M ± 0% -2.86% (p=0.008 n=5+5) Flate 233k ± 0% 231k ± 0% -0.83% (p=0.008 n=5+5) GoParser 296k ± 0% 291k ± 0% -1.54% (p=0.016 n=5+4) Reflect 1.05M ± 0% 1.04M ± 0% -0.65% (p=0.008 n=5+5) Tar 343k ± 0% 339k ± 0% -0.97% (p=0.008 n=5+5) XML 432k ± 0% 426k ± 0% -1.19% (p=0.008 n=5+5) [Geo mean] 815k 804k -1.35% name old object-bytes new object-bytes delta Template 505kB ± 0% 505kB ± 0% -0.01% (p=0.008 n=5+5) Unicode 224kB ± 0% 224kB ± 0% ~ (all equal) GoTypes 1.82MB ± 0% 1.83MB ± 0% +0.06% (p=0.008 n=5+5) Flate 324kB ± 0% 324kB ± 0% +0.00% (p=0.008 n=5+5) GoParser 402kB ± 0% 402kB ± 0% +0.04% (p=0.008 n=5+5) Reflect 1.39MB ± 0% 1.39MB ± 0% -0.01% (p=0.008 n=5+5) Tar 449kB ± 0% 449kB ± 0% -0.02% (p=0.008 n=5+5) XML 598kB ± 0% 597kB ± 0% -0.05% (p=0.008 n=5+5) Change-Id: Ifc9d5c1bd01f90171414b8fb18ffe2290d271143 Reviewed-on: https://go-review.googlesource.com/c/114797 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2018-05-06 12:58:53 -07:00
// Collapse moving A -> B -> C into just A -> C.
// Later passes (deadstore, elim unread auto) will remove the A -> B move, if possible.
// This happens most commonly when B is an autotmp inserted earlier
// during compilation to ensure correctness.
// Take care that overlapping moves are preserved.
// Restrict this optimization to the stack, to avoid duplicating loads from the heap;
// see CL 145208 for discussion.
(Move {t1} [s] dst tmp1 midmem:(Move {t2} [s] tmp2 src _))
&& t1.Compare(t2) == types.CMPeq
cmd/compile: move argument stack construction to SSA generation The goal of this change is to move work from walk to SSA, and simplify things along the way. This is hard to accomplish cleanly with small incremental changes, so this large commit message aims to provide a roadmap to the diff. High level description: Prior to this change, walk was responsible for constructing (most of) the stack for function calls. ascompatte gathered variadic arguments into a slice. It also rewrote n.List from a list of arguments to a list of assignments to stack slots. ascompatte was called multiple times to handle the receiver in a method call. reorder1 then introduced temporaries into n.List as needed to avoid smashing the stack. adjustargs then made extra stack space for go/defer args as needed. Node to SSA construction evaluated all the statements in n.List, and issued the function call, assuming that the stack was correctly constructed. Intrinsic calls had to dig around inside n.List to extract the arguments, since intrinsics don't use the stack to make function calls. This change moves stack construction to the SSA construction phase. ascompatte, now called walkParams, does all the work that ascompatte and reorder1 did. It handles variadic arguments, inserts the method receiver if needed, and allocates temporaries. It does not, however, make any assignments to stack slots. Instead, it moves the function arguments to n.Rlist, leaving assignments to temporaries in n.List. (It would be better to use Ninit instead of List; future work.) During SSA construction, after doing all the temporary assignments in n.List, the function arguments are assigned to stack slots by constructing the appropriate SSA Value, using (*state).storeArg. SSA construction also now handles adjustments for go/defer args. This change also simplifies intrinsic calls, since we no longer need to undo walk's work. Along the way, we simplify nodarg by pushing the fp==1 case to its callers, where it fits nicely. Generated code differences: There were a few optimizations applied along the way, the old way. f(g()) was rewritten to do a block copy of function results to function arguments. And reorder1 avoided introducing the final "save the stack" temporary in n.List. The f(g()) block copy optimization never actually triggered; the order pass rewrote away g(), so that has been removed. SSA optimizations mostly obviated the need for reorder1's optimization of avoiding the final temporary. The exception was when the temporary's type was not SSA-able; in that case, we got a Move into an autotmp and then an immediate Move onto the stack, with the autotmp never read or used again. This change introduces a new rewrite rule to detect such pointless double Moves and collapse them into a single Move. This is actually more powerful than the original optimization, since the original optimization relied on the imprecise Node.HasCall calculation. The other significant difference in the generated code is that the stack is now constructed completely in SP-offset order. Prior to this change, the stack was constructed somewhat haphazardly: first the final argument that Node.HasCall deemed to require a temporary, then other arguments, then the method receiver, then the defer/go args. SP-offset is probably a good default order. See future work. There are a few minor object file size changes as a result of this change. I investigated some regressions in early versions of this change. One regression (in archive/tar) was the addition of a single CMPQ instruction, which would be eliminated were this TODO from flagalloc to be done: // TODO: Remove original instructions if they are never used. One regression (in text/template) was an ADDQconstmodify that is now a regular MOVQLoad+ADDQconst+MOVQStore, due to an unlucky change in the order in which arguments are written. The argument change order can also now be luckier, so this appears to be a wash. All in all, though there will be minor winners and losers, this change appears to be performance neutral. Future work: Move loading the result of function calls to SSA construction; eliminate OINDREGSP. Consider pushing stack construction deeper into SSA world, perhaps in an arch-specific pass. Among other benefits, this would make it easier to transition to a new calling convention. This would require rethinking the handling of stack conflicts and is non-trivial. Figure out some clean way to indicate that stack construction Stores/Moves do not alias each other, so that subsequent passes may do things like CSE+tighten shared stack setup, do DSE using non-first Stores, etc. This would allow us to eliminate the minor text/template regression. Possibly make assignments to stack slots not treated as statements by DWARF. Compiler benchmarks: name old time/op new time/op delta Template 182ms ± 2% 179ms ± 2% -1.69% (p=0.000 n=47+48) Unicode 86.3ms ± 5% 85.1ms ± 4% -1.36% (p=0.001 n=50+50) GoTypes 646ms ± 1% 642ms ± 1% -0.63% (p=0.000 n=49+48) Compiler 2.89s ± 1% 2.86s ± 2% -1.36% (p=0.000 n=48+50) SSA 8.47s ± 1% 8.37s ± 2% -1.22% (p=0.000 n=47+50) Flate 122ms ± 2% 121ms ± 2% -0.66% (p=0.000 n=47+45) GoParser 147ms ± 2% 146ms ± 2% -0.53% (p=0.006 n=46+49) Reflect 406ms ± 2% 403ms ± 2% -0.76% (p=0.000 n=48+43) Tar 162ms ± 3% 162ms ± 4% ~ (p=0.191 n=46+50) XML 223ms ± 2% 222ms ± 2% -0.37% (p=0.031 n=45+49) [Geo mean] 382ms 378ms -0.89% name old user-time/op new user-time/op delta Template 219ms ± 3% 216ms ± 3% -1.56% (p=0.000 n=50+48) Unicode 109ms ± 6% 109ms ± 5% ~ (p=0.190 n=50+49) GoTypes 836ms ± 2% 828ms ± 2% -0.96% (p=0.000 n=49+48) Compiler 3.87s ± 2% 3.80s ± 1% -1.81% (p=0.000 n=49+46) SSA 12.0s ± 1% 11.8s ± 1% -2.01% (p=0.000 n=48+50) Flate 142ms ± 3% 141ms ± 3% -0.85% (p=0.003 n=50+48) GoParser 178ms ± 4% 175ms ± 4% -1.66% (p=0.000 n=48+46) Reflect 520ms ± 2% 512ms ± 2% -1.44% (p=0.000 n=45+48) Tar 200ms ± 3% 198ms ± 4% -0.61% (p=0.037 n=47+50) XML 277ms ± 3% 275ms ± 3% -0.85% (p=0.000 n=49+48) [Geo mean] 482ms 476ms -1.23% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 35.3MB ± 0% -2.18% (p=0.008 n=5+5) Unicode 29.8MB ± 0% 29.3MB ± 0% -1.58% (p=0.008 n=5+5) GoTypes 125MB ± 0% 123MB ± 0% -2.13% (p=0.008 n=5+5) Compiler 531MB ± 0% 513MB ± 0% -3.40% (p=0.008 n=5+5) SSA 2.00GB ± 0% 1.93GB ± 0% -3.34% (p=0.008 n=5+5) Flate 24.5MB ± 0% 24.3MB ± 0% -1.18% (p=0.008 n=5+5) GoParser 29.4MB ± 0% 28.7MB ± 0% -2.34% (p=0.008 n=5+5) Reflect 87.1MB ± 0% 86.0MB ± 0% -1.33% (p=0.008 n=5+5) Tar 35.3MB ± 0% 34.8MB ± 0% -1.44% (p=0.008 n=5+5) XML 47.9MB ± 0% 47.1MB ± 0% -1.86% (p=0.008 n=5+5) [Geo mean] 82.8MB 81.1MB -2.08% name old allocs/op new allocs/op delta Template 352k ± 0% 347k ± 0% -1.32% (p=0.008 n=5+5) Unicode 342k ± 0% 339k ± 0% -0.66% (p=0.008 n=5+5) GoTypes 1.29M ± 0% 1.27M ± 0% -1.30% (p=0.008 n=5+5) Compiler 4.98M ± 0% 4.87M ± 0% -2.14% (p=0.008 n=5+5) SSA 15.7M ± 0% 15.2M ± 0% -2.86% (p=0.008 n=5+5) Flate 233k ± 0% 231k ± 0% -0.83% (p=0.008 n=5+5) GoParser 296k ± 0% 291k ± 0% -1.54% (p=0.016 n=5+4) Reflect 1.05M ± 0% 1.04M ± 0% -0.65% (p=0.008 n=5+5) Tar 343k ± 0% 339k ± 0% -0.97% (p=0.008 n=5+5) XML 432k ± 0% 426k ± 0% -1.19% (p=0.008 n=5+5) [Geo mean] 815k 804k -1.35% name old object-bytes new object-bytes delta Template 505kB ± 0% 505kB ± 0% -0.01% (p=0.008 n=5+5) Unicode 224kB ± 0% 224kB ± 0% ~ (all equal) GoTypes 1.82MB ± 0% 1.83MB ± 0% +0.06% (p=0.008 n=5+5) Flate 324kB ± 0% 324kB ± 0% +0.00% (p=0.008 n=5+5) GoParser 402kB ± 0% 402kB ± 0% +0.04% (p=0.008 n=5+5) Reflect 1.39MB ± 0% 1.39MB ± 0% -0.01% (p=0.008 n=5+5) Tar 449kB ± 0% 449kB ± 0% -0.02% (p=0.008 n=5+5) XML 598kB ± 0% 597kB ± 0% -0.05% (p=0.008 n=5+5) Change-Id: Ifc9d5c1bd01f90171414b8fb18ffe2290d271143 Reviewed-on: https://go-review.googlesource.com/c/114797 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2018-05-06 12:58:53 -07:00
&& isSamePtr(tmp1, tmp2)
&& isStackPtr(src) && !isVolatile(src)
&& disjoint(src, s, tmp2, s)
&& (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))
=> (Move {t1} [s] dst src midmem)
cmd/compile: move argument stack construction to SSA generation The goal of this change is to move work from walk to SSA, and simplify things along the way. This is hard to accomplish cleanly with small incremental changes, so this large commit message aims to provide a roadmap to the diff. High level description: Prior to this change, walk was responsible for constructing (most of) the stack for function calls. ascompatte gathered variadic arguments into a slice. It also rewrote n.List from a list of arguments to a list of assignments to stack slots. ascompatte was called multiple times to handle the receiver in a method call. reorder1 then introduced temporaries into n.List as needed to avoid smashing the stack. adjustargs then made extra stack space for go/defer args as needed. Node to SSA construction evaluated all the statements in n.List, and issued the function call, assuming that the stack was correctly constructed. Intrinsic calls had to dig around inside n.List to extract the arguments, since intrinsics don't use the stack to make function calls. This change moves stack construction to the SSA construction phase. ascompatte, now called walkParams, does all the work that ascompatte and reorder1 did. It handles variadic arguments, inserts the method receiver if needed, and allocates temporaries. It does not, however, make any assignments to stack slots. Instead, it moves the function arguments to n.Rlist, leaving assignments to temporaries in n.List. (It would be better to use Ninit instead of List; future work.) During SSA construction, after doing all the temporary assignments in n.List, the function arguments are assigned to stack slots by constructing the appropriate SSA Value, using (*state).storeArg. SSA construction also now handles adjustments for go/defer args. This change also simplifies intrinsic calls, since we no longer need to undo walk's work. Along the way, we simplify nodarg by pushing the fp==1 case to its callers, where it fits nicely. Generated code differences: There were a few optimizations applied along the way, the old way. f(g()) was rewritten to do a block copy of function results to function arguments. And reorder1 avoided introducing the final "save the stack" temporary in n.List. The f(g()) block copy optimization never actually triggered; the order pass rewrote away g(), so that has been removed. SSA optimizations mostly obviated the need for reorder1's optimization of avoiding the final temporary. The exception was when the temporary's type was not SSA-able; in that case, we got a Move into an autotmp and then an immediate Move onto the stack, with the autotmp never read or used again. This change introduces a new rewrite rule to detect such pointless double Moves and collapse them into a single Move. This is actually more powerful than the original optimization, since the original optimization relied on the imprecise Node.HasCall calculation. The other significant difference in the generated code is that the stack is now constructed completely in SP-offset order. Prior to this change, the stack was constructed somewhat haphazardly: first the final argument that Node.HasCall deemed to require a temporary, then other arguments, then the method receiver, then the defer/go args. SP-offset is probably a good default order. See future work. There are a few minor object file size changes as a result of this change. I investigated some regressions in early versions of this change. One regression (in archive/tar) was the addition of a single CMPQ instruction, which would be eliminated were this TODO from flagalloc to be done: // TODO: Remove original instructions if they are never used. One regression (in text/template) was an ADDQconstmodify that is now a regular MOVQLoad+ADDQconst+MOVQStore, due to an unlucky change in the order in which arguments are written. The argument change order can also now be luckier, so this appears to be a wash. All in all, though there will be minor winners and losers, this change appears to be performance neutral. Future work: Move loading the result of function calls to SSA construction; eliminate OINDREGSP. Consider pushing stack construction deeper into SSA world, perhaps in an arch-specific pass. Among other benefits, this would make it easier to transition to a new calling convention. This would require rethinking the handling of stack conflicts and is non-trivial. Figure out some clean way to indicate that stack construction Stores/Moves do not alias each other, so that subsequent passes may do things like CSE+tighten shared stack setup, do DSE using non-first Stores, etc. This would allow us to eliminate the minor text/template regression. Possibly make assignments to stack slots not treated as statements by DWARF. Compiler benchmarks: name old time/op new time/op delta Template 182ms ± 2% 179ms ± 2% -1.69% (p=0.000 n=47+48) Unicode 86.3ms ± 5% 85.1ms ± 4% -1.36% (p=0.001 n=50+50) GoTypes 646ms ± 1% 642ms ± 1% -0.63% (p=0.000 n=49+48) Compiler 2.89s ± 1% 2.86s ± 2% -1.36% (p=0.000 n=48+50) SSA 8.47s ± 1% 8.37s ± 2% -1.22% (p=0.000 n=47+50) Flate 122ms ± 2% 121ms ± 2% -0.66% (p=0.000 n=47+45) GoParser 147ms ± 2% 146ms ± 2% -0.53% (p=0.006 n=46+49) Reflect 406ms ± 2% 403ms ± 2% -0.76% (p=0.000 n=48+43) Tar 162ms ± 3% 162ms ± 4% ~ (p=0.191 n=46+50) XML 223ms ± 2% 222ms ± 2% -0.37% (p=0.031 n=45+49) [Geo mean] 382ms 378ms -0.89% name old user-time/op new user-time/op delta Template 219ms ± 3% 216ms ± 3% -1.56% (p=0.000 n=50+48) Unicode 109ms ± 6% 109ms ± 5% ~ (p=0.190 n=50+49) GoTypes 836ms ± 2% 828ms ± 2% -0.96% (p=0.000 n=49+48) Compiler 3.87s ± 2% 3.80s ± 1% -1.81% (p=0.000 n=49+46) SSA 12.0s ± 1% 11.8s ± 1% -2.01% (p=0.000 n=48+50) Flate 142ms ± 3% 141ms ± 3% -0.85% (p=0.003 n=50+48) GoParser 178ms ± 4% 175ms ± 4% -1.66% (p=0.000 n=48+46) Reflect 520ms ± 2% 512ms ± 2% -1.44% (p=0.000 n=45+48) Tar 200ms ± 3% 198ms ± 4% -0.61% (p=0.037 n=47+50) XML 277ms ± 3% 275ms ± 3% -0.85% (p=0.000 n=49+48) [Geo mean] 482ms 476ms -1.23% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 35.3MB ± 0% -2.18% (p=0.008 n=5+5) Unicode 29.8MB ± 0% 29.3MB ± 0% -1.58% (p=0.008 n=5+5) GoTypes 125MB ± 0% 123MB ± 0% -2.13% (p=0.008 n=5+5) Compiler 531MB ± 0% 513MB ± 0% -3.40% (p=0.008 n=5+5) SSA 2.00GB ± 0% 1.93GB ± 0% -3.34% (p=0.008 n=5+5) Flate 24.5MB ± 0% 24.3MB ± 0% -1.18% (p=0.008 n=5+5) GoParser 29.4MB ± 0% 28.7MB ± 0% -2.34% (p=0.008 n=5+5) Reflect 87.1MB ± 0% 86.0MB ± 0% -1.33% (p=0.008 n=5+5) Tar 35.3MB ± 0% 34.8MB ± 0% -1.44% (p=0.008 n=5+5) XML 47.9MB ± 0% 47.1MB ± 0% -1.86% (p=0.008 n=5+5) [Geo mean] 82.8MB 81.1MB -2.08% name old allocs/op new allocs/op delta Template 352k ± 0% 347k ± 0% -1.32% (p=0.008 n=5+5) Unicode 342k ± 0% 339k ± 0% -0.66% (p=0.008 n=5+5) GoTypes 1.29M ± 0% 1.27M ± 0% -1.30% (p=0.008 n=5+5) Compiler 4.98M ± 0% 4.87M ± 0% -2.14% (p=0.008 n=5+5) SSA 15.7M ± 0% 15.2M ± 0% -2.86% (p=0.008 n=5+5) Flate 233k ± 0% 231k ± 0% -0.83% (p=0.008 n=5+5) GoParser 296k ± 0% 291k ± 0% -1.54% (p=0.016 n=5+4) Reflect 1.05M ± 0% 1.04M ± 0% -0.65% (p=0.008 n=5+5) Tar 343k ± 0% 339k ± 0% -0.97% (p=0.008 n=5+5) XML 432k ± 0% 426k ± 0% -1.19% (p=0.008 n=5+5) [Geo mean] 815k 804k -1.35% name old object-bytes new object-bytes delta Template 505kB ± 0% 505kB ± 0% -0.01% (p=0.008 n=5+5) Unicode 224kB ± 0% 224kB ± 0% ~ (all equal) GoTypes 1.82MB ± 0% 1.83MB ± 0% +0.06% (p=0.008 n=5+5) Flate 324kB ± 0% 324kB ± 0% +0.00% (p=0.008 n=5+5) GoParser 402kB ± 0% 402kB ± 0% +0.04% (p=0.008 n=5+5) Reflect 1.39MB ± 0% 1.39MB ± 0% -0.01% (p=0.008 n=5+5) Tar 449kB ± 0% 449kB ± 0% -0.02% (p=0.008 n=5+5) XML 598kB ± 0% 597kB ± 0% -0.05% (p=0.008 n=5+5) Change-Id: Ifc9d5c1bd01f90171414b8fb18ffe2290d271143 Reviewed-on: https://go-review.googlesource.com/c/114797 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2018-05-06 12:58:53 -07:00
// Same, but for large types that require VarDefs.
(Move {t1} [s] dst tmp1 midmem:(VarDef (Move {t2} [s] tmp2 src _)))
&& t1.Compare(t2) == types.CMPeq
&& isSamePtr(tmp1, tmp2)
&& isStackPtr(src) && !isVolatile(src)
&& disjoint(src, s, tmp2, s)
&& (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))
=> (Move {t1} [s] dst src midmem)
// Don't zero the same bits twice.
(Zero {t} [s] dst1 zero:(Zero {t} [s] dst2 _)) && isSamePtr(dst1, dst2) => zero
(Zero {t} [s] dst1 vardef:(VarDef (Zero {t} [s] dst2 _))) && isSamePtr(dst1, dst2) => vardef
cmd/compile: move argument stack construction to SSA generation The goal of this change is to move work from walk to SSA, and simplify things along the way. This is hard to accomplish cleanly with small incremental changes, so this large commit message aims to provide a roadmap to the diff. High level description: Prior to this change, walk was responsible for constructing (most of) the stack for function calls. ascompatte gathered variadic arguments into a slice. It also rewrote n.List from a list of arguments to a list of assignments to stack slots. ascompatte was called multiple times to handle the receiver in a method call. reorder1 then introduced temporaries into n.List as needed to avoid smashing the stack. adjustargs then made extra stack space for go/defer args as needed. Node to SSA construction evaluated all the statements in n.List, and issued the function call, assuming that the stack was correctly constructed. Intrinsic calls had to dig around inside n.List to extract the arguments, since intrinsics don't use the stack to make function calls. This change moves stack construction to the SSA construction phase. ascompatte, now called walkParams, does all the work that ascompatte and reorder1 did. It handles variadic arguments, inserts the method receiver if needed, and allocates temporaries. It does not, however, make any assignments to stack slots. Instead, it moves the function arguments to n.Rlist, leaving assignments to temporaries in n.List. (It would be better to use Ninit instead of List; future work.) During SSA construction, after doing all the temporary assignments in n.List, the function arguments are assigned to stack slots by constructing the appropriate SSA Value, using (*state).storeArg. SSA construction also now handles adjustments for go/defer args. This change also simplifies intrinsic calls, since we no longer need to undo walk's work. Along the way, we simplify nodarg by pushing the fp==1 case to its callers, where it fits nicely. Generated code differences: There were a few optimizations applied along the way, the old way. f(g()) was rewritten to do a block copy of function results to function arguments. And reorder1 avoided introducing the final "save the stack" temporary in n.List. The f(g()) block copy optimization never actually triggered; the order pass rewrote away g(), so that has been removed. SSA optimizations mostly obviated the need for reorder1's optimization of avoiding the final temporary. The exception was when the temporary's type was not SSA-able; in that case, we got a Move into an autotmp and then an immediate Move onto the stack, with the autotmp never read or used again. This change introduces a new rewrite rule to detect such pointless double Moves and collapse them into a single Move. This is actually more powerful than the original optimization, since the original optimization relied on the imprecise Node.HasCall calculation. The other significant difference in the generated code is that the stack is now constructed completely in SP-offset order. Prior to this change, the stack was constructed somewhat haphazardly: first the final argument that Node.HasCall deemed to require a temporary, then other arguments, then the method receiver, then the defer/go args. SP-offset is probably a good default order. See future work. There are a few minor object file size changes as a result of this change. I investigated some regressions in early versions of this change. One regression (in archive/tar) was the addition of a single CMPQ instruction, which would be eliminated were this TODO from flagalloc to be done: // TODO: Remove original instructions if they are never used. One regression (in text/template) was an ADDQconstmodify that is now a regular MOVQLoad+ADDQconst+MOVQStore, due to an unlucky change in the order in which arguments are written. The argument change order can also now be luckier, so this appears to be a wash. All in all, though there will be minor winners and losers, this change appears to be performance neutral. Future work: Move loading the result of function calls to SSA construction; eliminate OINDREGSP. Consider pushing stack construction deeper into SSA world, perhaps in an arch-specific pass. Among other benefits, this would make it easier to transition to a new calling convention. This would require rethinking the handling of stack conflicts and is non-trivial. Figure out some clean way to indicate that stack construction Stores/Moves do not alias each other, so that subsequent passes may do things like CSE+tighten shared stack setup, do DSE using non-first Stores, etc. This would allow us to eliminate the minor text/template regression. Possibly make assignments to stack slots not treated as statements by DWARF. Compiler benchmarks: name old time/op new time/op delta Template 182ms ± 2% 179ms ± 2% -1.69% (p=0.000 n=47+48) Unicode 86.3ms ± 5% 85.1ms ± 4% -1.36% (p=0.001 n=50+50) GoTypes 646ms ± 1% 642ms ± 1% -0.63% (p=0.000 n=49+48) Compiler 2.89s ± 1% 2.86s ± 2% -1.36% (p=0.000 n=48+50) SSA 8.47s ± 1% 8.37s ± 2% -1.22% (p=0.000 n=47+50) Flate 122ms ± 2% 121ms ± 2% -0.66% (p=0.000 n=47+45) GoParser 147ms ± 2% 146ms ± 2% -0.53% (p=0.006 n=46+49) Reflect 406ms ± 2% 403ms ± 2% -0.76% (p=0.000 n=48+43) Tar 162ms ± 3% 162ms ± 4% ~ (p=0.191 n=46+50) XML 223ms ± 2% 222ms ± 2% -0.37% (p=0.031 n=45+49) [Geo mean] 382ms 378ms -0.89% name old user-time/op new user-time/op delta Template 219ms ± 3% 216ms ± 3% -1.56% (p=0.000 n=50+48) Unicode 109ms ± 6% 109ms ± 5% ~ (p=0.190 n=50+49) GoTypes 836ms ± 2% 828ms ± 2% -0.96% (p=0.000 n=49+48) Compiler 3.87s ± 2% 3.80s ± 1% -1.81% (p=0.000 n=49+46) SSA 12.0s ± 1% 11.8s ± 1% -2.01% (p=0.000 n=48+50) Flate 142ms ± 3% 141ms ± 3% -0.85% (p=0.003 n=50+48) GoParser 178ms ± 4% 175ms ± 4% -1.66% (p=0.000 n=48+46) Reflect 520ms ± 2% 512ms ± 2% -1.44% (p=0.000 n=45+48) Tar 200ms ± 3% 198ms ± 4% -0.61% (p=0.037 n=47+50) XML 277ms ± 3% 275ms ± 3% -0.85% (p=0.000 n=49+48) [Geo mean] 482ms 476ms -1.23% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 35.3MB ± 0% -2.18% (p=0.008 n=5+5) Unicode 29.8MB ± 0% 29.3MB ± 0% -1.58% (p=0.008 n=5+5) GoTypes 125MB ± 0% 123MB ± 0% -2.13% (p=0.008 n=5+5) Compiler 531MB ± 0% 513MB ± 0% -3.40% (p=0.008 n=5+5) SSA 2.00GB ± 0% 1.93GB ± 0% -3.34% (p=0.008 n=5+5) Flate 24.5MB ± 0% 24.3MB ± 0% -1.18% (p=0.008 n=5+5) GoParser 29.4MB ± 0% 28.7MB ± 0% -2.34% (p=0.008 n=5+5) Reflect 87.1MB ± 0% 86.0MB ± 0% -1.33% (p=0.008 n=5+5) Tar 35.3MB ± 0% 34.8MB ± 0% -1.44% (p=0.008 n=5+5) XML 47.9MB ± 0% 47.1MB ± 0% -1.86% (p=0.008 n=5+5) [Geo mean] 82.8MB 81.1MB -2.08% name old allocs/op new allocs/op delta Template 352k ± 0% 347k ± 0% -1.32% (p=0.008 n=5+5) Unicode 342k ± 0% 339k ± 0% -0.66% (p=0.008 n=5+5) GoTypes 1.29M ± 0% 1.27M ± 0% -1.30% (p=0.008 n=5+5) Compiler 4.98M ± 0% 4.87M ± 0% -2.14% (p=0.008 n=5+5) SSA 15.7M ± 0% 15.2M ± 0% -2.86% (p=0.008 n=5+5) Flate 233k ± 0% 231k ± 0% -0.83% (p=0.008 n=5+5) GoParser 296k ± 0% 291k ± 0% -1.54% (p=0.016 n=5+4) Reflect 1.05M ± 0% 1.04M ± 0% -0.65% (p=0.008 n=5+5) Tar 343k ± 0% 339k ± 0% -0.97% (p=0.008 n=5+5) XML 432k ± 0% 426k ± 0% -1.19% (p=0.008 n=5+5) [Geo mean] 815k 804k -1.35% name old object-bytes new object-bytes delta Template 505kB ± 0% 505kB ± 0% -0.01% (p=0.008 n=5+5) Unicode 224kB ± 0% 224kB ± 0% ~ (all equal) GoTypes 1.82MB ± 0% 1.83MB ± 0% +0.06% (p=0.008 n=5+5) Flate 324kB ± 0% 324kB ± 0% +0.00% (p=0.008 n=5+5) GoParser 402kB ± 0% 402kB ± 0% +0.04% (p=0.008 n=5+5) Reflect 1.39MB ± 0% 1.39MB ± 0% -0.01% (p=0.008 n=5+5) Tar 449kB ± 0% 449kB ± 0% -0.02% (p=0.008 n=5+5) XML 598kB ± 0% 597kB ± 0% -0.05% (p=0.008 n=5+5) Change-Id: Ifc9d5c1bd01f90171414b8fb18ffe2290d271143 Reviewed-on: https://go-review.googlesource.com/c/114797 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2018-05-06 12:58:53 -07:00
// Elide self-moves. This only happens rarely (e.g test/fixedbugs/bug277.go).
// However, this rule is needed to prevent the previous rule from looping forever in such cases.
(Move dst src mem) && isSamePtr(dst, src) => mem
// Constant rotate detection.
((Add64|Or64|Xor64) (Lsh64x64 x z:(Const64 <t> [c])) (Rsh64Ux64 x (Const64 [d]))) && c < 64 && d == 64-c && canRotate(config, 64) => (RotateLeft64 x z)
((Add32|Or32|Xor32) (Lsh32x64 x z:(Const64 <t> [c])) (Rsh32Ux64 x (Const64 [d]))) && c < 32 && d == 32-c && canRotate(config, 32) => (RotateLeft32 x z)
((Add16|Or16|Xor16) (Lsh16x64 x z:(Const64 <t> [c])) (Rsh16Ux64 x (Const64 [d]))) && c < 16 && d == 16-c && canRotate(config, 16) => (RotateLeft16 x z)
((Add8|Or8|Xor8) (Lsh8x64 x z:(Const64 <t> [c])) (Rsh8Ux64 x (Const64 [d]))) && c < 8 && d == 8-c && canRotate(config, 8) => (RotateLeft8 x z)
// Non-constant rotate detection.
// We use shiftIsBounded to make sure that neither of the shifts are >64.
// Note: these rules are subtle when the shift amounts are 0/64, as Go shifts
// are different from most native shifts. But it works out.
((Add64|Or64|Xor64) left:(Lsh64x64 x y) right:(Rsh64Ux64 x (Sub64 (Const64 [64]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 64) => (RotateLeft64 x y)
((Add64|Or64|Xor64) left:(Lsh64x32 x y) right:(Rsh64Ux32 x (Sub32 (Const32 [64]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 64) => (RotateLeft64 x y)
((Add64|Or64|Xor64) left:(Lsh64x16 x y) right:(Rsh64Ux16 x (Sub16 (Const16 [64]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 64) => (RotateLeft64 x y)
((Add64|Or64|Xor64) left:(Lsh64x8 x y) right:(Rsh64Ux8 x (Sub8 (Const8 [64]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 64) => (RotateLeft64 x y)
((Add64|Or64|Xor64) right:(Rsh64Ux64 x y) left:(Lsh64x64 x z:(Sub64 (Const64 [64]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 64) => (RotateLeft64 x z)
((Add64|Or64|Xor64) right:(Rsh64Ux32 x y) left:(Lsh64x32 x z:(Sub32 (Const32 [64]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 64) => (RotateLeft64 x z)
((Add64|Or64|Xor64) right:(Rsh64Ux16 x y) left:(Lsh64x16 x z:(Sub16 (Const16 [64]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 64) => (RotateLeft64 x z)
((Add64|Or64|Xor64) right:(Rsh64Ux8 x y) left:(Lsh64x8 x z:(Sub8 (Const8 [64]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 64) => (RotateLeft64 x z)
((Add32|Or32|Xor32) left:(Lsh32x64 x y) right:(Rsh32Ux64 x (Sub64 (Const64 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x y)
((Add32|Or32|Xor32) left:(Lsh32x32 x y) right:(Rsh32Ux32 x (Sub32 (Const32 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x y)
((Add32|Or32|Xor32) left:(Lsh32x16 x y) right:(Rsh32Ux16 x (Sub16 (Const16 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x y)
((Add32|Or32|Xor32) left:(Lsh32x8 x y) right:(Rsh32Ux8 x (Sub8 (Const8 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x y)
((Add32|Or32|Xor32) right:(Rsh32Ux64 x y) left:(Lsh32x64 x z:(Sub64 (Const64 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x z)
((Add32|Or32|Xor32) right:(Rsh32Ux32 x y) left:(Lsh32x32 x z:(Sub32 (Const32 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x z)
((Add32|Or32|Xor32) right:(Rsh32Ux16 x y) left:(Lsh32x16 x z:(Sub16 (Const16 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x z)
((Add32|Or32|Xor32) right:(Rsh32Ux8 x y) left:(Lsh32x8 x z:(Sub8 (Const8 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x z)
((Add16|Or16|Xor16) left:(Lsh16x64 x y) right:(Rsh16Ux64 x (Sub64 (Const64 [16]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 16) => (RotateLeft16 x y)
((Add16|Or16|Xor16) left:(Lsh16x32 x y) right:(Rsh16Ux32 x (Sub32 (Const32 [16]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 16) => (RotateLeft16 x y)
((Add16|Or16|Xor16) left:(Lsh16x16 x y) right:(Rsh16Ux16 x (Sub16 (Const16 [16]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 16) => (RotateLeft16 x y)
((Add16|Or16|Xor16) left:(Lsh16x8 x y) right:(Rsh16Ux8 x (Sub8 (Const8 [16]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 16) => (RotateLeft16 x y)
((Add16|Or16|Xor16) right:(Rsh16Ux64 x y) left:(Lsh16x64 x z:(Sub64 (Const64 [16]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 16) => (RotateLeft16 x z)
((Add16|Or16|Xor16) right:(Rsh16Ux32 x y) left:(Lsh16x32 x z:(Sub32 (Const32 [16]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 16) => (RotateLeft16 x z)
((Add16|Or16|Xor16) right:(Rsh16Ux16 x y) left:(Lsh16x16 x z:(Sub16 (Const16 [16]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 16) => (RotateLeft16 x z)
((Add16|Or16|Xor16) right:(Rsh16Ux8 x y) left:(Lsh16x8 x z:(Sub8 (Const8 [16]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 16) => (RotateLeft16 x z)
((Add8|Or8|Xor8) left:(Lsh8x64 x y) right:(Rsh8Ux64 x (Sub64 (Const64 [8]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 8) => (RotateLeft8 x y)
((Add8|Or8|Xor8) left:(Lsh8x32 x y) right:(Rsh8Ux32 x (Sub32 (Const32 [8]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 8) => (RotateLeft8 x y)
((Add8|Or8|Xor8) left:(Lsh8x16 x y) right:(Rsh8Ux16 x (Sub16 (Const16 [8]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 8) => (RotateLeft8 x y)
((Add8|Or8|Xor8) left:(Lsh8x8 x y) right:(Rsh8Ux8 x (Sub8 (Const8 [8]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 8) => (RotateLeft8 x y)
((Add8|Or8|Xor8) right:(Rsh8Ux64 x y) left:(Lsh8x64 x z:(Sub64 (Const64 [8]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 8) => (RotateLeft8 x z)
((Add8|Or8|Xor8) right:(Rsh8Ux32 x y) left:(Lsh8x32 x z:(Sub32 (Const32 [8]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 8) => (RotateLeft8 x z)
((Add8|Or8|Xor8) right:(Rsh8Ux16 x y) left:(Lsh8x16 x z:(Sub16 (Const16 [8]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 8) => (RotateLeft8 x z)
((Add8|Or8|Xor8) right:(Rsh8Ux8 x y) left:(Lsh8x8 x z:(Sub8 (Const8 [8]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 8) => (RotateLeft8 x z)
// Rotating by y&c, with c a mask that doesn't change the bottom bits, is the same as rotating by y.
(RotateLeft64 x (And(64|32|16|8) y (Const(64|32|16|8) [c]))) && c&63 == 63 => (RotateLeft64 x y)
(RotateLeft32 x (And(64|32|16|8) y (Const(64|32|16|8) [c]))) && c&31 == 31 => (RotateLeft32 x y)
(RotateLeft16 x (And(64|32|16|8) y (Const(64|32|16|8) [c]))) && c&15 == 15 => (RotateLeft16 x y)
(RotateLeft8 x (And(64|32|16|8) y (Const(64|32|16|8) [c]))) && c&7 == 7 => (RotateLeft8 x y)
// Rotating by -(y&c), with c a mask that doesn't change the bottom bits, is the same as rotating by -y.
(RotateLeft64 x (Neg(64|32|16|8) (And(64|32|16|8) y (Const(64|32|16|8) [c])))) && c&63 == 63 => (RotateLeft64 x (Neg(64|32|16|8) <y.Type> y))
(RotateLeft32 x (Neg(64|32|16|8) (And(64|32|16|8) y (Const(64|32|16|8) [c])))) && c&31 == 31 => (RotateLeft32 x (Neg(64|32|16|8) <y.Type> y))
(RotateLeft16 x (Neg(64|32|16|8) (And(64|32|16|8) y (Const(64|32|16|8) [c])))) && c&15 == 15 => (RotateLeft16 x (Neg(64|32|16|8) <y.Type> y))
(RotateLeft8 x (Neg(64|32|16|8) (And(64|32|16|8) y (Const(64|32|16|8) [c])))) && c&7 == 7 => (RotateLeft8 x (Neg(64|32|16|8) <y.Type> y))
// Rotating by y+c, with c a multiple of the value width, is the same as rotating by y.
(RotateLeft64 x (Add(64|32|16|8) y (Const(64|32|16|8) [c]))) && c&63 == 0 => (RotateLeft64 x y)
(RotateLeft32 x (Add(64|32|16|8) y (Const(64|32|16|8) [c]))) && c&31 == 0 => (RotateLeft32 x y)
(RotateLeft16 x (Add(64|32|16|8) y (Const(64|32|16|8) [c]))) && c&15 == 0 => (RotateLeft16 x y)
(RotateLeft8 x (Add(64|32|16|8) y (Const(64|32|16|8) [c]))) && c&7 == 0 => (RotateLeft8 x y)
// Rotating by c-y, with c a multiple of the value width, is the same as rotating by -y.
(RotateLeft64 x (Sub(64|32|16|8) (Const(64|32|16|8) [c]) y)) && c&63 == 0 => (RotateLeft64 x (Neg(64|32|16|8) <y.Type> y))
(RotateLeft32 x (Sub(64|32|16|8) (Const(64|32|16|8) [c]) y)) && c&31 == 0 => (RotateLeft32 x (Neg(64|32|16|8) <y.Type> y))
(RotateLeft16 x (Sub(64|32|16|8) (Const(64|32|16|8) [c]) y)) && c&15 == 0 => (RotateLeft16 x (Neg(64|32|16|8) <y.Type> y))
(RotateLeft8 x (Sub(64|32|16|8) (Const(64|32|16|8) [c]) y)) && c&7 == 0 => (RotateLeft8 x (Neg(64|32|16|8) <y.Type> y))
// Ensure we don't do Const64 rotates in a 32-bit system.
(RotateLeft64 x (Const64 <t> [c])) && config.PtrSize == 4 => (RotateLeft64 x (Const32 <t> [int32(c)]))
(RotateLeft32 x (Const64 <t> [c])) && config.PtrSize == 4 => (RotateLeft32 x (Const32 <t> [int32(c)]))
(RotateLeft16 x (Const64 <t> [c])) && config.PtrSize == 4 => (RotateLeft16 x (Const32 <t> [int32(c)]))
(RotateLeft8 x (Const64 <t> [c])) && config.PtrSize == 4 => (RotateLeft8 x (Const32 <t> [int32(c)]))
// Rotating by c, then by d, is the same as rotating by c+d.
// We're trading a rotate for an add, which seems generally a good choice. It is especially good when c and d are constants.
// This rule is a bit tricky as c and d might be different widths. We handle only cases where they are the same width.
(RotateLeft(64|32|16|8) (RotateLeft(64|32|16|8) x c) d) && c.Type.Size() == 8 && d.Type.Size() == 8 => (RotateLeft(64|32|16|8) x (Add64 <c.Type> c d))
(RotateLeft(64|32|16|8) (RotateLeft(64|32|16|8) x c) d) && c.Type.Size() == 4 && d.Type.Size() == 4 => (RotateLeft(64|32|16|8) x (Add32 <c.Type> c d))
(RotateLeft(64|32|16|8) (RotateLeft(64|32|16|8) x c) d) && c.Type.Size() == 2 && d.Type.Size() == 2 => (RotateLeft(64|32|16|8) x (Add16 <c.Type> c d))
(RotateLeft(64|32|16|8) (RotateLeft(64|32|16|8) x c) d) && c.Type.Size() == 1 && d.Type.Size() == 1 => (RotateLeft(64|32|16|8) x (Add8 <c.Type> c d))
// Loading constant values from dictionaries and itabs.
(Load <t> (OffPtr [off] (Addr {s} sb) ) _) && t.IsUintptr() && isFixedSym(s, off) => (Addr {fixedSym(b.Func, s, off)} sb)
(Load <t> (OffPtr [off] (Convert (Addr {s} sb) _) ) _) && t.IsUintptr() && isFixedSym(s, off) => (Addr {fixedSym(b.Func, s, off)} sb)
(Load <t> (OffPtr [off] (ITab (IMake (Addr {s} sb) _))) _) && t.IsUintptr() && isFixedSym(s, off) => (Addr {fixedSym(b.Func, s, off)} sb)
(Load <t> (OffPtr [off] (ITab (IMake (Convert (Addr {s} sb) _) _))) _) && t.IsUintptr() && isFixedSym(s, off) => (Addr {fixedSym(b.Func, s, off)} sb)
// Loading constant values from runtime._type.hash.
(Load <t> (OffPtr [off] (Addr {sym} _) ) _) && t.IsInteger() && t.Size() == 4 && isFixed32(config, sym, off) => (Const32 [fixed32(config, sym, off)])
(Load <t> (OffPtr [off] (Convert (Addr {sym} _) _) ) _) && t.IsInteger() && t.Size() == 4 && isFixed32(config, sym, off) => (Const32 [fixed32(config, sym, off)])
(Load <t> (OffPtr [off] (ITab (IMake (Addr {sym} _) _))) _) && t.IsInteger() && t.Size() == 4 && isFixed32(config, sym, off) => (Const32 [fixed32(config, sym, off)])
(Load <t> (OffPtr [off] (ITab (IMake (Convert (Addr {sym} _) _) _))) _) && t.IsInteger() && t.Size() == 4 && isFixed32(config, sym, off) => (Const32 [fixed32(config, sym, off)])
// Calling cmpstring a second time with the same arguments in the
// same memory state can reuse the results of the first call.
// See issue 61725.
// Note that this could pretty easily generalize to any pure function.
(SelectN [0] (StaticLECall {f} x y (SelectN [1] c:(StaticLECall {g} x y mem))))
&& isSameCall(f, "runtime.cmpstring")
&& isSameCall(g, "runtime.cmpstring")
=> @c.Block (SelectN [0] <typ.Int> c)
// If we don't use the result of cmpstring, might as well not call it.
// Note that this could pretty easily generalize to any pure function.
(SelectN [1] c:(StaticLECall {f} _ _ mem)) && c.Uses == 1 && isSameCall(f, "runtime.cmpstring") && clobber(c) => mem