Stowage/go
2
0
Fork 0
mirror of https://github.com/golang/go.git synced 2025-11-01 17:20:56 +00:00
Code Issues Projects Releases Packages Wiki Activity
go/src/cmd/compile/internal/ssa/gen/generic.rules

2350 lines
103 KiB
Text
Raw Normal View History

[dev.ssa] cmd/internal/ssa: implement more compiler passes opt: machine-independent optimization fuse: join basic blocks lower: convert to machine-dependent opcodes critical: remove critical edges for register alloc layout: order basic blocks schedule: order values in basic blocks cgen: generate assembly output opt and lower use machine-generated matching rules using the rule generator in rulegen/ cgen will probably change in the real compiler, as we want to generate binary directly instead of ascii assembly. Change-Id: Iedd7ca70f6f55a4cde30e27cfad6a7fa05691b83 Reviewed-on: https://go-review.googlesource.com/7981 Reviewed-by: Alan Donovan <adonovan@google.com> Reviewed-by: Keith Randall <khr@golang.org>
2015-03-23 17:02:11 -07:00
// 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.
cmd/compile: precompute constant square roots If a program wants to evaluate math.Sqrt for any constant value (for example, math.Sqrt(3)), we can replace that expression with its evaluation (1.7320508075688772) at compile time, instead of generating a SQRT assembly command or equivalent. Adds tests that math.Sqrt generates the correct values. I also compiled a short program and verified that the Sqrt expression was replaced by a constant value in the "after opt" step. Adds a short doc to the top of generic.rules explaining what the file does and how other files interact with it. Fixes #15543. Change-Id: I6b6e63ac61cec50763a09ba581024adeee03d4fa Reviewed-on: https://go-review.googlesource.com/27457 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Keith Randall <khr@golang.org>
2016-08-20 22:05:47 -07:00
// 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.
[dev.ssa] cmd/compile/internal/ssa: Implement block rewriting rules Change-Id: I47e5349e34fc18118c4d35bf433f875b958cc3e5 Reviewed-on: https://go-review.googlesource.com/10495 Reviewed-by: Alan Donovan <adonovan@google.com>
2015-05-28 16:45:33 -07:00
// values are specified using the following format:
[dev.ssa] cmd/compiler/internal/ssa: Add auxint field Add an additional int64 auxiliary field to Value. There are two main reasons for doing this: 1) Ints in interfaces require allocation, and we store ints in Aux a lot. 2) I'd like to have both *gc.Sym and int offsets included in lots of operations (e.g. MOVQloadidx8). It will be more efficient to store them as separate fields instead of a pointer to a sym/int pair. It also simplifies a bunch of code. This is just the refactoring. I'll start using this some more in a subsequent changelist. Change-Id: I1ca797ff572553986cf90cab3ac0a0c1d01ad241 Reviewed-on: https://go-review.googlesource.com/10929 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-06-11 21:29:25 -07:00
// (op <type> [auxint] {aux} arg0 arg1 ...)
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
// the type, aux, and auxint fields are optional
[dev.ssa] cmd/compile/internal/ssa: Implement block rewriting rules Change-Id: I47e5349e34fc18118c4d35bf433f875b958cc3e5 Reviewed-on: https://go-review.googlesource.com/10495 Reviewed-by: Alan Donovan <adonovan@google.com>
2015-05-28 16:45:33 -07:00
// on the matching side
[dev.ssa] cmd/compiler/internal/ssa: Add auxint field Add an additional int64 auxiliary field to Value. There are two main reasons for doing this: 1) Ints in interfaces require allocation, and we store ints in Aux a lot. 2) I'd like to have both *gc.Sym and int offsets included in lots of operations (e.g. MOVQloadidx8). It will be more efficient to store them as separate fields instead of a pointer to a sym/int pair. It also simplifies a bunch of code. This is just the refactoring. I'll start using this some more in a subsequent changelist. Change-Id: I1ca797ff572553986cf90cab3ac0a0c1d01ad241 Reviewed-on: https://go-review.googlesource.com/10929 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-06-11 21:29:25 -07:00
// - the type, aux, and auxint fields must match if they are specified.
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
// - 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 "&&".
[dev.ssa] cmd/compile/internal/ssa: Implement block rewriting rules Change-Id: I47e5349e34fc18118c4d35bf433f875b958cc3e5 Reviewed-on: https://go-review.googlesource.com/10495 Reviewed-by: Alan Donovan <adonovan@google.com>
2015-05-28 16:45:33 -07:00
// on the generated side
// - the type of the top-level expression is the same as the one on the left-hand side.
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
// - the type of any subexpressions must be specified explicitly (or
// be specified in the op's type field).
[dev.ssa] cmd/compiler/internal/ssa: Add auxint field Add an additional int64 auxiliary field to Value. There are two main reasons for doing this: 1) Ints in interfaces require allocation, and we store ints in Aux a lot. 2) I'd like to have both *gc.Sym and int offsets included in lots of operations (e.g. MOVQloadidx8). It will be more efficient to store them as separate fields instead of a pointer to a sym/int pair. It also simplifies a bunch of code. This is just the refactoring. I'll start using this some more in a subsequent changelist. Change-Id: I1ca797ff572553986cf90cab3ac0a0c1d01ad241 Reviewed-on: https://go-review.googlesource.com/10929 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-06-11 21:29:25 -07:00
// - auxint will be 0 if not specified.
[dev.ssa] cmd/compile/internal/ssa: Implement block rewriting rules Change-Id: I47e5349e34fc18118c4d35bf433f875b958cc3e5 Reviewed-on: https://go-review.googlesource.com/10495 Reviewed-by: Alan Donovan <adonovan@google.com>
2015-05-28 16:45:33 -07:00
// - 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.
[dev.ssa] cmd/internal/ssa: implement more compiler passes opt: machine-independent optimization fuse: join basic blocks lower: convert to machine-dependent opcodes critical: remove critical edges for register alloc layout: order basic blocks schedule: order values in basic blocks cgen: generate assembly output opt and lower use machine-generated matching rules using the rule generator in rulegen/ cgen will probably change in the real compiler, as we want to generate binary directly instead of ascii assembly. Change-Id: Iedd7ca70f6f55a4cde30e27cfad6a7fa05691b83 Reviewed-on: https://go-review.googlesource.com/7981 Reviewed-by: Alan Donovan <adonovan@google.com> Reviewed-by: Keith Randall <khr@golang.org>
2015-03-23 17:02:11 -07:00
// constant folding
cmd/compile: improve floating point constant propagation Add generic rules to propagate floating point constants through comparisons and integer conversions. These new rules seldom trigger in the standard library so there is no performance change, however I think it is worth adding them anyway for completeness. Change-Id: I9db5222746508a2996f1cafb72f4e0cf2541de07 Reviewed-on: https://go-review.googlesource.com/63795 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-09-14 11:04:37 +01:00
(Trunc16to8 (Const16 [c])) -> (Const8 [int64(int8(c))])
(Trunc32to8 (Const32 [c])) -> (Const8 [int64(int8(c))])
(Trunc32to16 (Const32 [c])) -> (Const16 [int64(int16(c))])
(Trunc64to8 (Const64 [c])) -> (Const8 [int64(int8(c))])
(Trunc64to16 (Const64 [c])) -> (Const16 [int64(int16(c))])
(Trunc64to32 (Const64 [c])) -> (Const32 [int64(int32(c))])
cmd/compile: avoid more float32 <-> float64 conversions in compiler Use the new custom truncate/extension code when storing or extracting float32 values from AuxInts to avoid the value being changed by the host platform's floating point conversion instructions (e.g. sNaN -> qNaN). Updates #27516. Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974 Reviewed-on: https://go-review.googlesource.com/134855 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-09-12 12:16:50 +01:00
(Cvt64Fto32F (Const64F [c])) -> (Const32F [auxFrom32F(float32(auxTo64F(c)))])
cmd/compile: const folding for float32/64 Split the auxFloat type into 32/64 bit versions and perform checking for exactly representable float32 values. Perform const folding on float32/64. Comment out some const negation rules that the frontend already performs. Change-Id: Ib3f8d59fa8b30e50fe0267786cfb3c50a06169d2 Reviewed-on: https://go-review.googlesource.com/20568 Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-03-11 19:36:54 -06:00
(Cvt32Fto64F (Const32F [c])) -> (Const64F [c]) // c is already a 64 bit float
cmd/compile: avoid more float32 <-> float64 conversions in compiler Use the new custom truncate/extension code when storing or extracting float32 values from AuxInts to avoid the value being changed by the host platform's floating point conversion instructions (e.g. sNaN -> qNaN). Updates #27516. Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974 Reviewed-on: https://go-review.googlesource.com/134855 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-09-12 12:16:50 +01:00
(Cvt32to32F (Const32 [c])) -> (Const32F [auxFrom32F(float32(int32(c)))])
(Cvt32to64F (Const32 [c])) -> (Const64F [auxFrom64F(float64(int32(c)))])
(Cvt64to32F (Const64 [c])) -> (Const32F [auxFrom32F(float32(c))])
(Cvt64to64F (Const64 [c])) -> (Const64F [auxFrom64F(float64(c))])
(Cvt32Fto32 (Const32F [c])) -> (Const32 [int64(int32(auxTo32F(c)))])
(Cvt32Fto64 (Const32F [c])) -> (Const64 [int64(auxTo32F(c))])
(Cvt64Fto32 (Const64F [c])) -> (Const32 [int64(int32(auxTo64F(c)))])
(Cvt64Fto64 (Const64F [c])) -> (Const64 [int64(auxTo64F(c))])
cmd/compile: emit fused multiply-{add,subtract} instructions on s390x Explcitly block fused multiply-add pattern matching when a cast is used after the multiplication, for example: - (a * b) + c // can emit fused multiply-add - float64(a * b) + c // cannot emit fused multiply-add float{32,64} and complex{64,128} casts of matching types are now kept as OCONV operations rather than being replaced with OCONVNOP operations because they now imply a rounding operation (and therefore aren't a no-op anymore). Operations (for example, multiplication) on complex types may utilize fused multiply-add and -subtract instructions internally. There is no way to disable this behavior at the moment. Improves the performance of the floating point implementation of poly1305: name old speed new speed delta 64 246MB/s ± 0% 275MB/s ± 0% +11.48% (p=0.000 n=10+8) 1K 312MB/s ± 0% 357MB/s ± 0% +14.41% (p=0.000 n=10+10) 64Unaligned 246MB/s ± 0% 274MB/s ± 0% +11.43% (p=0.000 n=10+10) 1KUnaligned 312MB/s ± 0% 357MB/s ± 0% +14.39% (p=0.000 n=10+8) Updates #17895. Change-Id: Ia771d275bb9150d1a598f8cc773444663de5ce16 Reviewed-on: https://go-review.googlesource.com/36963 Run-TryBot: Michael Munday <munday@ca.ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-02-12 22:12:12 -05:00
(Round32F x:(Const32F)) -> x
(Round64F x:(Const64F)) -> x
[dev.ssa] cmd/compile/internal/ssa: constant fold truncates and bool comparisons Change-Id: I731722eb77f373ff7d6101f93830ab0a50497e2c Reviewed-on: https://go-review.googlesource.com/19542 Run-TryBot: David Chase <drchase@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-02-17 12:17:11 +01:00
cmd/compile: add Trunc-of-Ext simplifications This is a follow-up to the discussion in CL 27853. During make.bash, trigger count: 24 rewrite generic.rules:57 22 rewrite generic.rules:69 10 rewrite generic.rules:54 10 rewrite generic.rules:58 10 rewrite generic.rules:67 7 rewrite generic.rules:66 4 rewrite generic.rules:59 3 rewrite generic.rules:50 3 rewrite generic.rules:51 3 rewrite generic.rules:52 1 rewrite generic.rules:64 Change-Id: Id96cb6a707a4a564831f763c2d4d0e180c94bbef Reviewed-on: https://go-review.googlesource.com/28088 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Martin Möhrmann <martisch@uos.de>
2016-08-30 09:57:48 -07:00
(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
cmd/compile: optimize (ZeroExt (Const [c])) These rules trigger 116 times while running make.bash. And at least for the sample code at https://github.com/golang/go/issues/18906#issuecomment-277174241 they are providing optimizations not already present in amd64. Updates #18906 Change-Id: I410a480f566f5ab176fc573fb5ac74f9cffec225 Reviewed-on: https://go-review.googlesource.com/36217 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-02-02 22:38:04 -08:00
(ZeroExt8to16 (Const8 [c])) -> (Const16 [int64( uint8(c))])
(ZeroExt8to32 (Const8 [c])) -> (Const32 [int64( uint8(c))])
(ZeroExt8to64 (Const8 [c])) -> (Const64 [int64( uint8(c))])
(ZeroExt16to32 (Const16 [c])) -> (Const32 [int64(uint16(c))])
(ZeroExt16to64 (Const16 [c])) -> (Const64 [int64(uint16(c))])
(ZeroExt32to64 (Const32 [c])) -> (Const64 [int64(uint32(c))])
(SignExt8to16 (Const8 [c])) -> (Const16 [int64( int8(c))])
(SignExt8to32 (Const8 [c])) -> (Const32 [int64( int8(c))])
(SignExt8to64 (Const8 [c])) -> (Const64 [int64( int8(c))])
(SignExt16to32 (Const16 [c])) -> (Const32 [int64( int16(c))])
(SignExt16to64 (Const16 [c])) -> (Const64 [int64( int16(c))])
(SignExt32to64 (Const32 [c])) -> (Const64 [int64( int32(c))])
cmd/compile: fold (NegNN (ConstNN ...)) Fix up and enable a few rules. They trigger a handful of times in std, despite the frontend handling. Change-Id: I83378c057cbbc95a4f2b58cd8c36aec0e9dc547f Reviewed-on: https://go-review.googlesource.com/37227 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-02-18 13:58:59 -08:00
(Neg8 (Const8 [c])) -> (Const8 [int64( -int8(c))])
(Neg16 (Const16 [c])) -> (Const16 [int64(-int16(c))])
(Neg32 (Const32 [c])) -> (Const32 [int64(-int32(c))])
(Neg64 (Const64 [c])) -> (Const64 [-c])
cmd/compile: avoid more float32 <-> float64 conversions in compiler Use the new custom truncate/extension code when storing or extracting float32 values from AuxInts to avoid the value being changed by the host platform's floating point conversion instructions (e.g. sNaN -> qNaN). Updates #27516. Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974 Reviewed-on: https://go-review.googlesource.com/134855 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-09-12 12:16:50 +01:00
(Neg32F (Const32F [c])) && auxTo32F(c) != 0 -> (Const32F [auxFrom32F(-auxTo32F(c))])
(Neg64F (Const64F [c])) && auxTo64F(c) != 0 -> (Const64F [auxFrom64F(-auxTo64F(c))])
[dev.ssa] cmd/compile/internal/ssa/gen: constant fold Neg*. Change-Id: Id51e5c97e9653b764b809bf3424f1a6d31b6ffea Reviewed-on: https://go-review.googlesource.com/19338 Run-TryBot: David Chase <drchase@google.com> Reviewed-by: David Chase <drchase@google.com>
2016-02-08 18:55:56 +01:00
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Add8 (Const8 [c]) (Const8 [d])) -> (Const8 [int64(int8(c+d))])
(Add16 (Const16 [c]) (Const16 [d])) -> (Const16 [int64(int16(c+d))])
(Add32 (Const32 [c]) (Const32 [d])) -> (Const32 [int64(int32(c+d))])
(Add64 (Const64 [c]) (Const64 [d])) -> (Const64 [c+d])
cmd/compile: avoid more float32 <-> float64 conversions in compiler Use the new custom truncate/extension code when storing or extracting float32 values from AuxInts to avoid the value being changed by the host platform's floating point conversion instructions (e.g. sNaN -> qNaN). Updates #27516. Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974 Reviewed-on: https://go-review.googlesource.com/134855 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-09-12 12:16:50 +01:00
(Add32F (Const32F [c]) (Const32F [d])) -> (Const32F [auxFrom32F(auxTo32F(c) + auxTo32F(d))])
(Add64F (Const64F [c]) (Const64F [d])) -> (Const64F [auxFrom64F(auxTo64F(c) + auxTo64F(d))])
cmd/compile: add rules to simplify AddPtr Fixes #14849 Change-Id: I86e2dc27ca73bb6b24261a68cbf0094a63167414 Reviewed-on: https://go-review.googlesource.com/20833 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-18 10:50:00 +01:00
(AddPtr <t> x (Const64 [c])) -> (OffPtr <t> x [c])
cmd/compile: add 32 bit (AddPtr (Const)) rule This triggers about 50k times during 32 bit make.bash. Change-Id: Ia0c2b1a8246b92173b4b0d94a4037626f76b6e73 Reviewed-on: https://go-review.googlesource.com/37998 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2017-03-09 14:28:32 -08:00
(AddPtr <t> x (Const32 [c])) -> (OffPtr <t> x [c])
[dev.ssa] cmd/compile: Do pointer arithmetic with int, not uintptr Be more consistent about this. There's no reason to do the pointer arithmetic on a different type, as sizeof(int) >= sizeof(ptr) on all of our platforms. It simplifies our rewrite rules also, except for a few that need duplication. Add some more constant folding to get constant indexing and slicing to fold down to nothing. Change-Id: I3e56cdb14b3dc1a6a0514f0333e883f92c19e3c7 Reviewed-on: https://go-review.googlesource.com/16586 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-02 21:28:13 -08:00
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Sub8 (Const8 [c]) (Const8 [d])) -> (Const8 [int64(int8(c-d))])
(Sub16 (Const16 [c]) (Const16 [d])) -> (Const16 [int64(int16(c-d))])
(Sub32 (Const32 [c]) (Const32 [d])) -> (Const32 [int64(int32(c-d))])
(Sub64 (Const64 [c]) (Const64 [d])) -> (Const64 [c-d])
cmd/compile: avoid more float32 <-> float64 conversions in compiler Use the new custom truncate/extension code when storing or extracting float32 values from AuxInts to avoid the value being changed by the host platform's floating point conversion instructions (e.g. sNaN -> qNaN). Updates #27516. Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974 Reviewed-on: https://go-review.googlesource.com/134855 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-09-12 12:16:50 +01:00
(Sub32F (Const32F [c]) (Const32F [d])) -> (Const32F [auxFrom32F(auxTo32F(c) - auxTo32F(d))])
(Sub64F (Const64F [c]) (Const64F [d])) -> (Const64F [auxFrom64F(auxTo64F(c) - auxTo64F(d))])
[dev.ssa] cmd/compile: Do pointer arithmetic with int, not uintptr Be more consistent about this. There's no reason to do the pointer arithmetic on a different type, as sizeof(int) >= sizeof(ptr) on all of our platforms. It simplifies our rewrite rules also, except for a few that need duplication. Add some more constant folding to get constant indexing and slicing to fold down to nothing. Change-Id: I3e56cdb14b3dc1a6a0514f0333e883f92c19e3c7 Reviewed-on: https://go-review.googlesource.com/16586 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-02 21:28:13 -08:00
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Mul8 (Const8 [c]) (Const8 [d])) -> (Const8 [int64(int8(c*d))])
(Mul16 (Const16 [c]) (Const16 [d])) -> (Const16 [int64(int16(c*d))])
(Mul32 (Const32 [c]) (Const32 [d])) -> (Const32 [int64(int32(c*d))])
(Mul64 (Const64 [c]) (Const64 [d])) -> (Const64 [c*d])
Revert "cmd/compile: don't allow NaNs in floating-point constant ops" This reverts CL 213477. Reason for revert: tests are failing on linux-mips*-rtrk builders. Change-Id: I8168f7450890233f1bd7e53930b73693c26d4dc0 Reviewed-on: https://go-review.googlesource.com/c/go/+/220897 Run-TryBot: Bryan C. Mills <bcmills@google.com> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2020-02-25 14:36:38 +00:00
(Mul32F (Const32F [c]) (Const32F [d])) -> (Const32F [auxFrom32F(auxTo32F(c) * auxTo32F(d))])
(Mul64F (Const64F [c]) (Const64F [d])) -> (Const64F [auxFrom64F(auxTo64F(c) * auxTo64F(d))])
[dev.ssa] cmd/compile: Do pointer arithmetic with int, not uintptr Be more consistent about this. There's no reason to do the pointer arithmetic on a different type, as sizeof(int) >= sizeof(ptr) on all of our platforms. It simplifies our rewrite rules also, except for a few that need duplication. Add some more constant folding to get constant indexing and slicing to fold down to nothing. Change-Id: I3e56cdb14b3dc1a6a0514f0333e883f92c19e3c7 Reviewed-on: https://go-review.googlesource.com/16586 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-02 21:28:13 -08:00
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
(And8 (Const8 [c]) (Const8 [d])) -> (Const8 [int64(int8(c&d))])
(And16 (Const16 [c]) (Const16 [d])) -> (Const16 [int64(int16(c&d))])
(And32 (Const32 [c]) (Const32 [d])) -> (Const32 [int64(int32(c&d))])
(And64 (Const64 [c]) (Const64 [d])) -> (Const64 [c&d])
(Or8 (Const8 [c]) (Const8 [d])) -> (Const8 [int64(int8(c|d))])
(Or16 (Const16 [c]) (Const16 [d])) -> (Const16 [int64(int16(c|d))])
(Or32 (Const32 [c]) (Const32 [d])) -> (Const32 [int64(int32(c|d))])
(Or64 (Const64 [c]) (Const64 [d])) -> (Const64 [c|d])
(Xor8 (Const8 [c]) (Const8 [d])) -> (Const8 [int64(int8(c^d))])
(Xor16 (Const16 [c]) (Const16 [d])) -> (Const16 [int64(int16(c^d))])
(Xor32 (Const32 [c]) (Const32 [d])) -> (Const32 [int64(int32(c^d))])
(Xor64 (Const64 [c]) (Const64 [d])) -> (Const64 [c^d])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Div8 (Const8 [c]) (Const8 [d])) && d != 0 -> (Const8 [int64(int8(c)/int8(d))])
(Div16 (Const16 [c]) (Const16 [d])) && d != 0 -> (Const16 [int64(int16(c)/int16(d))])
(Div32 (Const32 [c]) (Const32 [d])) && d != 0 -> (Const32 [int64(int32(c)/int32(d))])
(Div64 (Const64 [c]) (Const64 [d])) && d != 0 -> (Const64 [c/d])
(Div8u (Const8 [c]) (Const8 [d])) && d != 0 -> (Const8 [int64(int8(uint8(c)/uint8(d)))])
(Div16u (Const16 [c]) (Const16 [d])) && d != 0 -> (Const16 [int64(int16(uint16(c)/uint16(d)))])
(Div32u (Const32 [c]) (Const32 [d])) && d != 0 -> (Const32 [int64(int32(uint32(c)/uint32(d)))])
(Div64u (Const64 [c]) (Const64 [d])) && d != 0 -> (Const64 [int64(uint64(c)/uint64(d))])
Revert "cmd/compile: don't allow NaNs in floating-point constant ops" This reverts CL 213477. Reason for revert: tests are failing on linux-mips*-rtrk builders. Change-Id: I8168f7450890233f1bd7e53930b73693c26d4dc0 Reviewed-on: https://go-review.googlesource.com/c/go/+/220897 Run-TryBot: Bryan C. Mills <bcmills@google.com> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2020-02-25 14:36:38 +00:00
(Div32F (Const32F [c]) (Const32F [d])) -> (Const32F [auxFrom32F(auxTo32F(c) / auxTo32F(d))])
(Div64F (Const64F [c]) (Const64F [d])) -> (Const64F [auxFrom64F(auxTo64F(c) / auxTo64F(d))])
cmd/compile: reduce bits.Div64(0, lo, y) to 64 bit division With this change, these two functions generate identical code: func f(x uint64) (uint64, uint64) { return bits.Div64(0, x, 5) } func g(x uint64) (uint64, uint64) { return x / 5, x % 5 } Updates #31582 Change-Id: Ia96c2e67f8af5dd985823afee5f155608c04a4b6 Reviewed-on: https://go-review.googlesource.com/c/go/+/173197 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2019-04-20 11:09:34 -07:00
(Select0 (Div128u (Const64 [0]) lo y)) -> (Div64u lo y)
(Select1 (Div128u (Const64 [0]) lo y)) -> (Mod64u lo y)
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
cmd/compile: constant fold !true and !false Constant fold Not of boolean constants. Noticed while working on #23504. Change-Id: I965705154ee7348a1a159fad4e029b922d3171b3 Reviewed-on: https://go-review.googlesource.com/88956 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Martin Möhrmann <moehrmann@google.com>
2018-01-22 10:06:10 -08:00
(Not (ConstBool [c])) -> (ConstBool [1-c])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// Convert x * 1 to x.
cmd/compile: use | in the last repetitive generic rules This change or-ifies the last low-hanging rules in generic. Again, this is limited at short and repetitive rules, where the use or ors does not impact readability. Ran rulegen, no change in the actual compiler code. Change-Id: I972b523bc08532f173a3645b47d6936b6e1218c8 Reviewed-on: https://go-review.googlesource.com/96335 Reviewed-by: Giovanni Bajo <rasky@develer.com>
2018-02-22 14:32:57 +01:00
(Mul(8|16|32|64) (Const(8|16|32|64) [1]) x) -> x
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// Convert x * -1 to -x.
cmd/compile: use | in the last repetitive generic rules This change or-ifies the last low-hanging rules in generic. Again, this is limited at short and repetitive rules, where the use or ors does not impact readability. Ran rulegen, no change in the actual compiler code. Change-Id: I972b523bc08532f173a3645b47d6936b6e1218c8 Reviewed-on: https://go-review.googlesource.com/96335 Reviewed-by: Giovanni Bajo <rasky@develer.com>
2018-02-22 14:32:57 +01:00
(Mul(8|16|32|64) (Const(8|16|32|64) [-1]) x) -> (Neg(8|16|32|64) x)
[dev.ssa] cmd/compile: add some constant folding optimizations These were helpful for some autogenerated code I'm working with. Change-Id: I7b89c69552ca99bf560a14bfbcd6bd238595ddf6 Reviewed-on: https://go-review.googlesource.com/24742 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-07-05 10:03:02 -07:00
cmd/compile: remove walkmul Replace with generic rewrite rules. Change-Id: I3ee32076cfd9db5801f1a7bdbb73a994255884a9 Reviewed-on: https://go-review.googlesource.com/36323 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Keith Randall <khr@golang.org>
2017-02-04 21:20:23 -08:00
// Convert multiplication by a power of two to a shift.
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
(Mul8 <t> n (Const8 [c])) && isPowerOfTwo(c) -> (Lsh8x64 <t> n (Const64 <typ.UInt64> [log2(c)]))
(Mul16 <t> n (Const16 [c])) && isPowerOfTwo(c) -> (Lsh16x64 <t> n (Const64 <typ.UInt64> [log2(c)]))
(Mul32 <t> n (Const32 [c])) && isPowerOfTwo(c) -> (Lsh32x64 <t> n (Const64 <typ.UInt64> [log2(c)]))
(Mul64 <t> n (Const64 [c])) && isPowerOfTwo(c) -> (Lsh64x64 <t> n (Const64 <typ.UInt64> [log2(c)]))
(Mul8 <t> n (Const8 [c])) && t.IsSigned() && isPowerOfTwo(-c) -> (Neg8 (Lsh8x64 <t> n (Const64 <typ.UInt64> [log2(-c)])))
(Mul16 <t> n (Const16 [c])) && t.IsSigned() && isPowerOfTwo(-c) -> (Neg16 (Lsh16x64 <t> n (Const64 <typ.UInt64> [log2(-c)])))
(Mul32 <t> n (Const32 [c])) && t.IsSigned() && isPowerOfTwo(-c) -> (Neg32 (Lsh32x64 <t> n (Const64 <typ.UInt64> [log2(-c)])))
(Mul64 <t> n (Const64 [c])) && t.IsSigned() && isPowerOfTwo(-c) -> (Neg64 (Lsh64x64 <t> n (Const64 <typ.UInt64> [log2(-c)])))
cmd/compile: remove walkmul Replace with generic rewrite rules. Change-Id: I3ee32076cfd9db5801f1a7bdbb73a994255884a9 Reviewed-on: https://go-review.googlesource.com/36323 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Keith Randall <khr@golang.org>
2017-02-04 21:20:23 -08:00
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Mod8 (Const8 [c]) (Const8 [d])) && d != 0 -> (Const8 [int64(int8(c % d))])
(Mod16 (Const16 [c]) (Const16 [d])) && d != 0 -> (Const16 [int64(int16(c % d))])
(Mod32 (Const32 [c]) (Const32 [d])) && d != 0 -> (Const32 [int64(int32(c % d))])
(Mod64 (Const64 [c]) (Const64 [d])) && d != 0 -> (Const64 [c % d])
cmd/compile: constant fold modulo Fixes #15079 Change-Id: Ib4dd9eab322da39234008e040100e75cb58761b3 Reviewed-on: https://go-review.googlesource.com/21501 Reviewed-by: David Chase <drchase@google.com> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-04-04 16:13:35 +02:00
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Mod8u (Const8 [c]) (Const8 [d])) && d != 0 -> (Const8 [int64(uint8(c) % uint8(d))])
(Mod16u (Const16 [c]) (Const16 [d])) && d != 0 -> (Const16 [int64(uint16(c) % uint16(d))])
(Mod32u (Const32 [c]) (Const32 [d])) && d != 0 -> (Const32 [int64(uint32(c) % uint32(d))])
(Mod64u (Const64 [c]) (Const64 [d])) && d != 0 -> (Const64 [int64(uint64(c) % uint64(d))])
cmd/compile: constant fold modulo Fixes #15079 Change-Id: Ib4dd9eab322da39234008e040100e75cb58761b3 Reviewed-on: https://go-review.googlesource.com/21501 Reviewed-by: David Chase <drchase@google.com> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-04-04 16:13:35 +02:00
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(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 [int64(int32(c) << uint64(d))])
(Rsh32x64 (Const32 [c]) (Const64 [d])) -> (Const32 [int64(int32(c) >> uint64(d))])
cmd/compile: Fix constant-folding of unsigned shifts Make sure the results of unsigned constant-folded shifts are sign-extended into the AuxInt field. Fixes #15175 Change-Id: I3490d1bc3d9b2e1578ed30964645508577894f58 Reviewed-on: https://go-review.googlesource.com/21586 Reviewed-by: Alexandru Moșoi <alexandru@mosoi.ro> Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-04-07 10:21:35 -07:00
(Rsh32Ux64 (Const32 [c]) (Const64 [d])) -> (Const32 [int64(int32(uint32(c) >> uint64(d)))])
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Lsh16x64 (Const16 [c]) (Const64 [d])) -> (Const16 [int64(int16(c) << uint64(d))])
(Rsh16x64 (Const16 [c]) (Const64 [d])) -> (Const16 [int64(int16(c) >> uint64(d))])
cmd/compile: Fix constant-folding of unsigned shifts Make sure the results of unsigned constant-folded shifts are sign-extended into the AuxInt field. Fixes #15175 Change-Id: I3490d1bc3d9b2e1578ed30964645508577894f58 Reviewed-on: https://go-review.googlesource.com/21586 Reviewed-by: Alexandru Moșoi <alexandru@mosoi.ro> Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-04-07 10:21:35 -07:00
(Rsh16Ux64 (Const16 [c]) (Const64 [d])) -> (Const16 [int64(int16(uint16(c) >> uint64(d)))])
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Lsh8x64 (Const8 [c]) (Const64 [d])) -> (Const8 [int64(int8(c) << uint64(d))])
(Rsh8x64 (Const8 [c]) (Const64 [d])) -> (Const8 [int64(int8(c) >> uint64(d))])
cmd/compile: Fix constant-folding of unsigned shifts Make sure the results of unsigned constant-folded shifts are sign-extended into the AuxInt field. Fixes #15175 Change-Id: I3490d1bc3d9b2e1578ed30964645508577894f58 Reviewed-on: https://go-review.googlesource.com/21586 Reviewed-by: Alexandru Moșoi <alexandru@mosoi.ro> Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-04-07 10:21:35 -07:00
(Rsh8Ux64 (Const8 [c]) (Const64 [d])) -> (Const8 [int64(int8(uint8(c) >> uint64(d)))])
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
cmd/compile: introduce bool operations. Introduce OrB, EqB, NeqB, AndB to handle bool operations. Change-Id: I53e4d5125a8090d5eeb4576db619103f19fff58d Reviewed-on: https://go-review.googlesource.com/22412 Reviewed-by: Keith Randall <khr@golang.org>
2016-04-24 21:21:07 +02:00
// Fold IsInBounds when the range of the index cannot exceed the limit.
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(IsInBounds (ZeroExt8to32 _) (Const32 [c])) && (1 << 8) <= c -> (ConstBool [1])
(IsInBounds (ZeroExt8to64 _) (Const64 [c])) && (1 << 8) <= c -> (ConstBool [1])
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
(IsInBounds (ZeroExt16to32 _) (Const32 [c])) && (1 << 16) <= c -> (ConstBool [1])
cmd/compile: fold IsInBounds with small index For the following example, but there are a few more in the stdlib: func histogram(b []byte, h *[256]int32) { for _, t := range b { h[t]++ } } Change-Id: I56615f341ae52e02ef34025588dc6d1c52122295 Reviewed-on: https://go-review.googlesource.com/20924 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-20 22:15:27 +01:00
(IsInBounds (ZeroExt16to64 _) (Const64 [c])) && (1 << 16) <= c -> (ConstBool [1])
cmd/compile/internal/ssa: generalize prove to all booleans * Refacts a bit saving and restoring parents restrictions * Shaves ~100k from pkg/tools/linux_amd64, but most of the savings come from the rewrite rules. * Improves on the following artificial test case: func f1(a4 bool, a6 bool) bool { return a6 || (a6 || (a6 || a4)) || (a6 || (a4 || a6 || (false || a6))) } Change-Id: I714000f75a37a3a6617c6e6834c75bd23674215f Reviewed-on: https://go-review.googlesource.com/20306 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-07 18:36:16 +01:00
(IsInBounds x x) -> (ConstBool [0])
cmd/compile: improve generic rules for BCE based on AND operations. Match more patterns generated by the compiler where the index for a bound check is bounded through a AND operation, with different register sizes. These rules trigger a dozen of times in a bootstrap. Change-Id: Ic9fff16f21d08580f19a366c3ee1a372e58357d1 Reviewed-on: https://go-review.googlesource.com/37442 TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-02-26 02:54:32 +01:00
(IsInBounds (And8 (Const8 [c]) _) (Const8 [d])) && 0 <= c && c < d -> (ConstBool [1])
(IsInBounds (ZeroExt8to16 (And8 (Const8 [c]) _)) (Const16 [d])) && 0 <= c && c < d -> (ConstBool [1])
(IsInBounds (ZeroExt8to32 (And8 (Const8 [c]) _)) (Const32 [d])) && 0 <= c && c < d -> (ConstBool [1])
(IsInBounds (ZeroExt8to64 (And8 (Const8 [c]) _)) (Const64 [d])) && 0 <= c && c < d -> (ConstBool [1])
(IsInBounds (And16 (Const16 [c]) _) (Const16 [d])) && 0 <= c && c < d -> (ConstBool [1])
(IsInBounds (ZeroExt16to32 (And16 (Const16 [c]) _)) (Const32 [d])) && 0 <= c && c < d -> (ConstBool [1])
(IsInBounds (ZeroExt16to64 (And16 (Const16 [c]) _)) (Const64 [d])) && 0 <= c && c < d -> (ConstBool [1])
(IsInBounds (And32 (Const32 [c]) _) (Const32 [d])) && 0 <= c && c < d -> (ConstBool [1])
(IsInBounds (ZeroExt32to64 (And32 (Const32 [c]) _)) (Const64 [d])) && 0 <= c && c < d -> (ConstBool [1])
(IsInBounds (And64 (Const64 [c]) _) (Const64 [d])) && 0 <= c && c < d -> (ConstBool [1])
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
(IsInBounds (Const32 [c]) (Const32 [d])) -> (ConstBool [b2i(0 <= c && c < d)])
(IsInBounds (Const64 [c]) (Const64 [d])) -> (ConstBool [b2i(0 <= c && c < d)])
cmd/compile: fold IsInBounds of modulo. In b we only need the division by 0 check. func b(i uint, v []byte) byte { return v[i%uint(len(v))] } Updates #15079. Change-Id: Ic7491e677dd57cd6ba577efbce576dbb6e023cbd Reviewed-on: https://go-review.googlesource.com/21502 Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Ahmed Waheed <oneofone@gmail.com>
2016-04-04 16:38:26 +02:00
// (Mod64u x y) is always between 0 (inclusive) and y (exclusive).
(IsInBounds (Mod32u _ y) y) -> (ConstBool [1])
(IsInBounds (Mod64u _ y) y) -> (ConstBool [1])
cmd/compile: fix trivial typos in comments Change-Id: I04880d87e317a1140ec12da6ec5e788991719760 Reviewed-on: https://go-review.googlesource.com/114936 Reviewed-by: Ian Lance Taylor <iant@golang.org>
2018-05-28 13:49:40 -07:00
// Right shifting an unsigned number limits its value.
cmd/compile: use right shifts to eliminate bounds checks These rules trigger a few times during make.bash. When we eliminate boundedness checks from walk.go we'll rely on them more heavily. Updates #19692 Change-Id: I268c36ae2f1401c68dd685b15f2d30f5d6971176 Reviewed-on: https://go-review.googlesource.com/43775 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-05-19 15:10:10 -07:00
(IsInBounds (ZeroExt8to64 (Rsh8Ux64 _ (Const64 [c]))) (Const64 [d])) && 0 < c && c < 8 && 1<<uint( 8-c)-1 < d -> (ConstBool [1])
(IsInBounds (ZeroExt8to32 (Rsh8Ux64 _ (Const64 [c]))) (Const32 [d])) && 0 < c && c < 8 && 1<<uint( 8-c)-1 < d -> (ConstBool [1])
(IsInBounds (ZeroExt8to16 (Rsh8Ux64 _ (Const64 [c]))) (Const16 [d])) && 0 < c && c < 8 && 1<<uint( 8-c)-1 < d -> (ConstBool [1])
(IsInBounds (Rsh8Ux64 _ (Const64 [c])) (Const64 [d])) && 0 < c && c < 8 && 1<<uint( 8-c)-1 < d -> (ConstBool [1])
(IsInBounds (ZeroExt16to64 (Rsh16Ux64 _ (Const64 [c]))) (Const64 [d])) && 0 < c && c < 16 && 1<<uint(16-c)-1 < d -> (ConstBool [1])
(IsInBounds (ZeroExt16to32 (Rsh16Ux64 _ (Const64 [c]))) (Const64 [d])) && 0 < c && c < 16 && 1<<uint(16-c)-1 < d -> (ConstBool [1])
(IsInBounds (Rsh16Ux64 _ (Const64 [c])) (Const64 [d])) && 0 < c && c < 16 && 1<<uint(16-c)-1 < d -> (ConstBool [1])
(IsInBounds (ZeroExt32to64 (Rsh32Ux64 _ (Const64 [c]))) (Const64 [d])) && 0 < c && c < 32 && 1<<uint(32-c)-1 < d -> (ConstBool [1])
(IsInBounds (Rsh32Ux64 _ (Const64 [c])) (Const64 [d])) && 0 < c && c < 32 && 1<<uint(32-c)-1 < d -> (ConstBool [1])
(IsInBounds (Rsh64Ux64 _ (Const64 [c])) (Const64 [d])) && 0 < c && c < 64 && 1<<uint(64-c)-1 < d -> (ConstBool [1])
cmd/compile: fold IsInBounds of modulo. In b we only need the division by 0 check. func b(i uint, v []byte) byte { return v[i%uint(len(v))] } Updates #15079. Change-Id: Ic7491e677dd57cd6ba577efbce576dbb6e023cbd Reviewed-on: https://go-review.googlesource.com/21502 Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Ahmed Waheed <oneofone@gmail.com>
2016-04-04 16:38:26 +02:00
cmd/compile/internal/ssa: generalize prove to all booleans * Refacts a bit saving and restoring parents restrictions * Shaves ~100k from pkg/tools/linux_amd64, but most of the savings come from the rewrite rules. * Improves on the following artificial test case: func f1(a4 bool, a6 bool) bool { return a6 || (a6 || (a6 || a4)) || (a6 || (a4 || a6 || (false || a6))) } Change-Id: I714000f75a37a3a6617c6e6834c75bd23674215f Reviewed-on: https://go-review.googlesource.com/20306 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-07 18:36:16 +01:00
(IsSliceInBounds x x) -> (ConstBool [1])
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
(IsSliceInBounds (And32 (Const32 [c]) _) (Const32 [d])) && 0 <= c && c <= d -> (ConstBool [1])
(IsSliceInBounds (And64 (Const64 [c]) _) (Const64 [d])) && 0 <= c && c <= d -> (ConstBool [1])
cmd/compile/internal/ssa: generalize prove to all booleans * Refacts a bit saving and restoring parents restrictions * Shaves ~100k from pkg/tools/linux_amd64, but most of the savings come from the rewrite rules. * Improves on the following artificial test case: func f1(a4 bool, a6 bool) bool { return a6 || (a6 || (a6 || a4)) || (a6 || (a4 || a6 || (false || a6))) } Change-Id: I714000f75a37a3a6617c6e6834c75bd23674215f Reviewed-on: https://go-review.googlesource.com/20306 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-07 18:36:16 +01:00
(IsSliceInBounds (Const32 [0]) _) -> (ConstBool [1])
(IsSliceInBounds (Const64 [0]) _) -> (ConstBool [1])
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
(IsSliceInBounds (Const32 [c]) (Const32 [d])) -> (ConstBool [b2i(0 <= c && c <= d)])
(IsSliceInBounds (Const64 [c]) (Const64 [d])) -> (ConstBool [b2i(0 <= c && c <= d)])
cmd/compile/internal/ssa: generalize prove to all booleans * Refacts a bit saving and restoring parents restrictions * Shaves ~100k from pkg/tools/linux_amd64, but most of the savings come from the rewrite rules. * Improves on the following artificial test case: func f1(a4 bool, a6 bool) bool { return a6 || (a6 || (a6 || a4)) || (a6 || (a4 || a6 || (false || a6))) } Change-Id: I714000f75a37a3a6617c6e6834c75bd23674215f Reviewed-on: https://go-review.googlesource.com/20306 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-07 18:36:16 +01:00
(IsSliceInBounds (SliceLen x) (SliceCap x)) -> (ConstBool [1])
[dev.ssa] cmd/compile/internal/ssa: constant fold truncates and bool comparisons Change-Id: I731722eb77f373ff7d6101f93830ab0a50497e2c Reviewed-on: https://go-review.googlesource.com/19542 Run-TryBot: David Chase <drchase@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-02-17 12:17:11 +01:00
cmd/compile: use | in the most repetitive boolean rules For now, limited to a few repetitive boolean rules where the win is substantial (4+ variants). Change-Id: I67bce0d356ca7d71a0f15ff98551fe2caff8abf9 Reviewed-on: https://go-review.googlesource.com/95535 Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2018-02-20 20:18:16 +01:00
(Eq(64|32|16|8) x x) -> (ConstBool [1])
cmd/compile: introduce bool operations. Introduce OrB, EqB, NeqB, AndB to handle bool operations. Change-Id: I53e4d5125a8090d5eeb4576db619103f19fff58d Reviewed-on: https://go-review.googlesource.com/22412 Reviewed-by: Keith Randall <khr@golang.org>
2016-04-24 21:21:07 +02:00
(EqB (ConstBool [c]) (ConstBool [d])) -> (ConstBool [b2i(c == d)])
(EqB (ConstBool [0]) x) -> (Not x)
(EqB (ConstBool [1]) x) -> x
[dev.ssa] cmd/compile/internal/ssa: constant fold truncates and bool comparisons Change-Id: I731722eb77f373ff7d6101f93830ab0a50497e2c Reviewed-on: https://go-review.googlesource.com/19542 Run-TryBot: David Chase <drchase@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-02-17 12:17:11 +01:00
cmd/compile: use | in the most repetitive boolean rules For now, limited to a few repetitive boolean rules where the win is substantial (4+ variants). Change-Id: I67bce0d356ca7d71a0f15ff98551fe2caff8abf9 Reviewed-on: https://go-review.googlesource.com/95535 Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2018-02-20 20:18:16 +01:00
(Neq(64|32|16|8) x x) -> (ConstBool [0])
cmd/compile: introduce bool operations. Introduce OrB, EqB, NeqB, AndB to handle bool operations. Change-Id: I53e4d5125a8090d5eeb4576db619103f19fff58d Reviewed-on: https://go-review.googlesource.com/22412 Reviewed-by: Keith Randall <khr@golang.org>
2016-04-24 21:21:07 +02:00
(NeqB (ConstBool [c]) (ConstBool [d])) -> (ConstBool [b2i(c != d)])
(NeqB (ConstBool [0]) x) -> x
(NeqB (ConstBool [1]) x) -> (Not x)
cmd/compile: add boolean simplification rules These collectively fire a few hundred times during make.bash, mostly rewriting XOR SETNE -> SETEQ. Fixes #17905. Change-Id: Ic5eb241ee93ed67099da3de11f59e4df9fab64a3 Reviewed-on: https://go-review.googlesource.com/42491 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-05-02 16:43:27 -07:00
(NeqB (Not x) (Not y)) -> (NeqB x y)
[dev.ssa] cmd/internal/ssa: implement more compiler passes opt: machine-independent optimization fuse: join basic blocks lower: convert to machine-dependent opcodes critical: remove critical edges for register alloc layout: order basic blocks schedule: order values in basic blocks cgen: generate assembly output opt and lower use machine-generated matching rules using the rule generator in rulegen/ cgen will probably change in the real compiler, as we want to generate binary directly instead of ascii assembly. Change-Id: Iedd7ca70f6f55a4cde30e27cfad6a7fa05691b83 Reviewed-on: https://go-review.googlesource.com/7981 Reviewed-by: Alan Donovan <adonovan@google.com> Reviewed-by: Keith Randall <khr@golang.org>
2015-03-23 17:02:11 -07:00
[dev.ssa] cmd/compile/internal/ssa: simplify comparisons with constants * Simplify comparisons of form a + const1 == const2 or a + const1 != const2. * Canonicalize Eq, Neq, Add, Sub to have a constant as first argument. Needed for the above new rules and helps constant folding. Change-Id: I8078702a5daa706da57106073a3e9f640a67f486 Reviewed-on: https://go-review.googlesource.com/19192 Reviewed-by: Keith Randall <khr@golang.org>
2016-02-03 19:43:46 +01:00
(Eq64 (Const64 <t> [c]) (Add64 (Const64 <t> [d]) x)) -> (Eq64 (Const64 <t> [c-d]) x)
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
(Eq32 (Const32 <t> [c]) (Add32 (Const32 <t> [d]) x)) -> (Eq32 (Const32 <t> [int64(int32(c-d))]) x)
(Eq16 (Const16 <t> [c]) (Add16 (Const16 <t> [d]) x)) -> (Eq16 (Const16 <t> [int64(int16(c-d))]) x)
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Eq8 (Const8 <t> [c]) (Add8 (Const8 <t> [d]) x)) -> (Eq8 (Const8 <t> [int64(int8(c-d))]) x)
[dev.ssa] cmd/compile/internal/ssa: simplify comparisons with constants * Simplify comparisons of form a + const1 == const2 or a + const1 != const2. * Canonicalize Eq, Neq, Add, Sub to have a constant as first argument. Needed for the above new rules and helps constant folding. Change-Id: I8078702a5daa706da57106073a3e9f640a67f486 Reviewed-on: https://go-review.googlesource.com/19192 Reviewed-by: Keith Randall <khr@golang.org>
2016-02-03 19:43:46 +01:00
(Neq64 (Const64 <t> [c]) (Add64 (Const64 <t> [d]) x)) -> (Neq64 (Const64 <t> [c-d]) x)
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
(Neq32 (Const32 <t> [c]) (Add32 (Const32 <t> [d]) x)) -> (Neq32 (Const32 <t> [int64(int32(c-d))]) x)
(Neq16 (Const16 <t> [c]) (Add16 (Const16 <t> [d]) x)) -> (Neq16 (Const16 <t> [int64(int16(c-d))]) x)
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Neq8 (Const8 <t> [c]) (Add8 (Const8 <t> [d]) x)) -> (Neq8 (Const8 <t> [int64(int8(c-d))]) x)
[dev.ssa] cmd/compile/internal/ssa: simplify comparisons with constants * Simplify comparisons of form a + const1 == const2 or a + const1 != const2. * Canonicalize Eq, Neq, Add, Sub to have a constant as first argument. Needed for the above new rules and helps constant folding. Change-Id: I8078702a5daa706da57106073a3e9f640a67f486 Reviewed-on: https://go-review.googlesource.com/19192 Reviewed-by: Keith Randall <khr@golang.org>
2016-02-03 19:43:46 +01:00
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)
[dev.ssa] cmd/compile/internal/ssa: simplify comparisons with constants * Simplify comparisons of form a + const1 == const2 or a + const1 != const2. * Canonicalize Eq, Neq, Add, Sub to have a constant as first argument. Needed for the above new rules and helps constant folding. Change-Id: I8078702a5daa706da57106073a3e9f640a67f486 Reviewed-on: https://go-review.googlesource.com/19192 Reviewed-by: Keith Randall <khr@golang.org>
2016-02-03 19:43:46 +01:00
(Sub64 x (Const64 <t> [c])) && x.Op != OpConst64 -> (Add64 (Const64 <t> [-c]) x)
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
(Sub32 x (Const32 <t> [c])) && x.Op != OpConst32 -> (Add32 (Const32 <t> [int64(int32(-c))]) x)
(Sub16 x (Const16 <t> [c])) && x.Op != OpConst16 -> (Add16 (Const16 <t> [int64(int16(-c))]) x)
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Sub8 x (Const8 <t> [c])) && x.Op != OpConst8 -> (Add8 (Const8 <t> [int64(int8(-c))]) x)
[dev.ssa] cmd/compile/internal/ssa: simplify comparisons with constants * Simplify comparisons of form a + const1 == const2 or a + const1 != const2. * Canonicalize Eq, Neq, Add, Sub to have a constant as first argument. Needed for the above new rules and helps constant folding. Change-Id: I8078702a5daa706da57106073a3e9f640a67f486 Reviewed-on: https://go-review.googlesource.com/19192 Reviewed-by: Keith Randall <khr@golang.org>
2016-02-03 19:43:46 +01:00
cmd/compile: fold negation into comparison operators This allows for example AMD64 ssa to generate (SETNE x) instead of (XORLconst [1] SETE). make.bash trigger count on AMD64: 691 generic.rules:225 1 generic.rules:226 4 generic.rules:228 1 generic.rules:229 8 generic.rules:231 6 generic.rules:238 2 generic.rules:257 Change-Id: I5b9827b2df63c8532675079e5a6026aa47bfd8dc Reviewed-on: https://go-review.googlesource.com/28232 Run-TryBot: Martin Möhrmann <martisch@uos.de> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-08-31 12:37:19 +02:00
// fold negation into comparison operators
cmd/compile: absorb more Not ops into Neq* and Eq* ops We absorbed Not into most integer comparisons but not into pointer and floating point equality checks. The new cases trigger more than 300 times during make.bash. Change-Id: I77c6b31fcacde10da5470b73fc001a19521ce78d Reviewed-on: https://go-review.googlesource.com/c/go/+/200618 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-10-11 03:51:28 -07:00
(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)
cmd/compile: use | in the most repetitive boolean rules For now, limited to a few repetitive boolean rules where the win is substantial (4+ variants). Change-Id: I67bce0d356ca7d71a0f15ff98551fe2caff8abf9 Reviewed-on: https://go-review.googlesource.com/95535 Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2018-02-20 20:18:16 +01:00
cmd/compile: remove Greater* and Geq* generic integer ops The generic Greater and Geq ops can always be replaced with the Less and Leq ops. This CL therefore removes them. This simplifies the compiler since it reduces the number of operations that need handling in both code and in rewrite rules. This will be especially true when adding control flow optimizations such as the integer-in-range optimizations in CL 165998. Change-Id: If0648b2b19998ac1bddccbf251283f3be4ec3040 Reviewed-on: https://go-review.googlesource.com/c/go/+/220417 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2020-02-20 17:46:08 +00:00
(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)
cmd/compile: use | in the most repetitive boolean rules For now, limited to a few repetitive boolean rules where the win is substantial (4+ variants). Change-Id: I67bce0d356ca7d71a0f15ff98551fe2caff8abf9 Reviewed-on: https://go-review.googlesource.com/95535 Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2018-02-20 20:18:16 +01:00
[dev.ssa] cmd/compile/internal/ssa: distribute multiplication into addition * This is a very basic form of straight line strength reduction. * Removes one multiplication from a[b].c++; a[b+1].c++ * It increases pressure on the register allocator because CSE creates more copies of the multiplication sizeof(a[0])*b. Change-Id: I686a18e9c24cc6f8bdfa925713afed034f7d36d0 Reviewed-on: https://go-review.googlesource.com/20091 Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-03-01 13:39:47 +01:00
// Distribute multiplication c * (d+x) -> c*d + c*x. Useful for:
// a[i].b = ...; a[i+1].b = ...
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(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> [int64(int32(c*d))]) (Mul32 <t> (Const32 <t> [c]) x))
[dev.ssa] cmd/compile/internal/ssa: distribute multiplication into addition * This is a very basic form of straight line strength reduction. * Removes one multiplication from a[b].c++; a[b+1].c++ * It increases pressure on the register allocator because CSE creates more copies of the multiplication sizeof(a[0])*b. Change-Id: I686a18e9c24cc6f8bdfa925713afed034f7d36d0 Reviewed-on: https://go-review.googlesource.com/20091 Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-03-01 13:39:47 +01:00
cmd/compile: use more ORs in generic.rules No changes in the actual generated compiler code. Change-Id: I206a7bf7b60f70a73640119fc92974f79ed95a6b Reviewed-on: https://go-review.googlesource.com/102416 Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
2018-03-24 12:08:00 +01:00
// 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))
cmd/compile: combine x*n - y*n into (x-y)*n Do the similar thing to CL 55143 to reduce IMUL. Change-Id: I1bd38f618058e3cd74fac181f003610ea13f2294 Reviewed-on: https://go-review.googlesource.com/56252 Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-08-17 15:06:42 +08:00
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
// 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))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Lsh64x8 <t> x (Const8 [c])) -> (Lsh64x64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Rsh64x32 <t> x (Const32 [c])) -> (Rsh64x64 x (Const64 <t> [int64(uint32(c))]))
(Rsh64x16 <t> x (Const16 [c])) -> (Rsh64x64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Rsh64x8 <t> x (Const8 [c])) -> (Rsh64x64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Rsh64Ux32 <t> x (Const32 [c])) -> (Rsh64Ux64 x (Const64 <t> [int64(uint32(c))]))
(Rsh64Ux16 <t> x (Const16 [c])) -> (Rsh64Ux64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Rsh64Ux8 <t> x (Const8 [c])) -> (Rsh64Ux64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Lsh32x32 <t> x (Const32 [c])) -> (Lsh32x64 x (Const64 <t> [int64(uint32(c))]))
(Lsh32x16 <t> x (Const16 [c])) -> (Lsh32x64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Lsh32x8 <t> x (Const8 [c])) -> (Lsh32x64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Rsh32x32 <t> x (Const32 [c])) -> (Rsh32x64 x (Const64 <t> [int64(uint32(c))]))
(Rsh32x16 <t> x (Const16 [c])) -> (Rsh32x64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Rsh32x8 <t> x (Const8 [c])) -> (Rsh32x64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Rsh32Ux32 <t> x (Const32 [c])) -> (Rsh32Ux64 x (Const64 <t> [int64(uint32(c))]))
(Rsh32Ux16 <t> x (Const16 [c])) -> (Rsh32Ux64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Rsh32Ux8 <t> x (Const8 [c])) -> (Rsh32Ux64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Lsh16x32 <t> x (Const32 [c])) -> (Lsh16x64 x (Const64 <t> [int64(uint32(c))]))
(Lsh16x16 <t> x (Const16 [c])) -> (Lsh16x64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Lsh16x8 <t> x (Const8 [c])) -> (Lsh16x64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Rsh16x32 <t> x (Const32 [c])) -> (Rsh16x64 x (Const64 <t> [int64(uint32(c))]))
(Rsh16x16 <t> x (Const16 [c])) -> (Rsh16x64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Rsh16x8 <t> x (Const8 [c])) -> (Rsh16x64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Rsh16Ux32 <t> x (Const32 [c])) -> (Rsh16Ux64 x (Const64 <t> [int64(uint32(c))]))
(Rsh16Ux16 <t> x (Const16 [c])) -> (Rsh16Ux64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Rsh16Ux8 <t> x (Const8 [c])) -> (Rsh16Ux64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Lsh8x32 <t> x (Const32 [c])) -> (Lsh8x64 x (Const64 <t> [int64(uint32(c))]))
(Lsh8x16 <t> x (Const16 [c])) -> (Lsh8x64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Lsh8x8 <t> x (Const8 [c])) -> (Lsh8x64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Rsh8x32 <t> x (Const32 [c])) -> (Rsh8x64 x (Const64 <t> [int64(uint32(c))]))
(Rsh8x16 <t> x (Const16 [c])) -> (Rsh8x64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Rsh8x8 <t> x (Const8 [c])) -> (Rsh8x64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
(Rsh8Ux32 <t> x (Const32 [c])) -> (Rsh8Ux64 x (Const64 <t> [int64(uint32(c))]))
(Rsh8Ux16 <t> x (Const16 [c])) -> (Rsh8Ux64 x (Const64 <t> [int64(uint16(c))]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Rsh8Ux8 <t> x (Const8 [c])) -> (Rsh8Ux64 x (Const64 <t> [int64(uint8(c))]))
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
// shifts by zero
cmd/compile: use more ORs in generic.rules No changes in the actual generated compiler code. Change-Id: I206a7bf7b60f70a73640119fc92974f79ed95a6b Reviewed-on: https://go-review.googlesource.com/102416 Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
2018-03-24 12:08:00 +01:00
(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: use more ORs in generic.rules No changes in the actual generated compiler code. Change-Id: I206a7bf7b60f70a73640119fc92974f79ed95a6b Reviewed-on: https://go-review.googlesource.com/102416 Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
2018-03-24 12:08:00 +01:00
// zero shifted
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(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])
[dev.ssa] cmd/compile/internal/ssa: more simplifications and normalization Found by inspecting random generated code. Change-Id: I57d0fed7c3a8dc91fd13cdccb4819101f9976ec9 Reviewed-on: https://go-review.googlesource.com/19413 Reviewed-by: Keith Randall <khr@golang.org>
2016-02-09 19:13:43 +01:00
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
// large left shifts of all values, and right shifts of unsigned values
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
((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])
[dev.ssa] cmd/compile: rewrites for constant shifts Add rewrite rules to optimize constant shifts. Fixes #10637 Change-Id: I74b724d3e81aeb7098c696d02c050f7fdfd5b523 Reviewed-on: https://go-review.googlesource.com/19106 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-02-03 06:21:24 -05:00
// 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]))
cmd/compile: remove superfluous signed right shift used for signed division by 2 A signed right shift before an unsigned right shift by register width-1 (extracts the sign bit) is superflous. trigger counts during ./make.bash 0 (Rsh8U (Rsh8 x _) 7 ) -> (Rsh8U x 7 ) 0 (Rsh16U (Rsh16 x _) 15 ) -> (Rsh16U x 15) 2 (Rsh32U (Rsh32 x _) 31 ) -> (Rsh32U x 31) 251 (Rsh64U (Rsh64 x _) 63 ) -> (Rsh64U x 63) Changes the instructions generated on AMD64 for x / 2 where x is a signed integer from: MOVQ AX, CX SARQ $63, AX SHRQ $63, AX ADDQ CX, AX SARQ $1, AX to: MOVQ AX, CX SHRQ $63, AX ADDQ CX, AX SARQ $1, AX Change-Id: I86321ae8fc9dc24b8fa9eb80aa5c7299eff8c9dc Reviewed-on: https://go-review.googlesource.com/115956 Run-TryBot: Martin Möhrmann <moehrmann@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-06-02 20:55:20 +02:00
// 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]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
// ((x >> c1) << c2) >> c3
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(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]))
cmd/compile: shift tests, fix triple-shift rules Add a bunch of tests for shifts. Fix triple-shift rules to always take constant shifts as 64 bits. (Earlier rules always promote shift amounts to 64 bits.) Add overflow checks. Increases generic rule coverage to 91% Change-Id: I6b42d368d19d36ac482dbb8e0d4f67e30ad7145d Reviewed-on: https://go-review.googlesource.com/23555 Reviewed-by: Todd Neal <todd@tneal.org>
2016-05-28 21:15:24 -07:00
&& uint64(c1) >= uint64(c2) && uint64(c3) >= uint64(c2) && !uaddOvf(c1-c2, c3)
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
-> (Rsh(64|32|16|8)Ux64 x (Const64 <typ.UInt64> [c1-c2+c3]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
// ((x << c1) >> c2) << c3
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(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]))
cmd/compile: shift tests, fix triple-shift rules Add a bunch of tests for shifts. Fix triple-shift rules to always take constant shifts as 64 bits. (Earlier rules always promote shift amounts to 64 bits.) Add overflow checks. Increases generic rule coverage to 91% Change-Id: I6b42d368d19d36ac482dbb8e0d4f67e30ad7145d Reviewed-on: https://go-review.googlesource.com/23555 Reviewed-by: Todd Neal <todd@tneal.org>
2016-05-28 21:15:24 -07:00
&& uint64(c1) >= uint64(c2) && uint64(c3) >= uint64(c2) && !uaddOvf(c1-c2, c3)
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
-> (Lsh(64|32|16|8)x64 x (Const64 <typ.UInt64> [c1-c2+c3]))
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
cmd/compile: fold bit masking on bits that have been shifted away Spotted while working on #18943, it triggers once during bootstrap. Change-Id: Ia4330ccc6395627c233a8eb4dcc0e3e2a770bea7 Reviewed-on: https://go-review.googlesource.com/94764 Reviewed-by: Keith Randall <khr@golang.org>
2018-02-18 12:58:11 +01:00
// (x >> c) & uppermask = 0
(And64 (Const64 [m]) (Rsh64Ux64 _ (Const64 [c]))) && c >= 64-ntz(m) -> (Const64 [0])
(And32 (Const32 [m]) (Rsh32Ux64 _ (Const64 [c]))) && c >= 64-ntz(m) -> (Const32 [0])
(And16 (Const16 [m]) (Rsh16Ux64 _ (Const64 [c]))) && c >= 64-ntz(m) -> (Const16 [0])
(And8 (Const8 [m]) (Rsh8Ux64 _ (Const64 [c]))) && c >= 64-ntz(m) -> (Const8 [0])
// (x << c) & lowermask = 0
(And64 (Const64 [m]) (Lsh64x64 _ (Const64 [c]))) && c >= 64-nlz(m) -> (Const64 [0])
(And32 (Const32 [m]) (Lsh32x64 _ (Const64 [c]))) && c >= 64-nlz(m) -> (Const32 [0])
(And16 (Const16 [m]) (Lsh16x64 _ (Const64 [c]))) && c >= 64-nlz(m) -> (Const16 [0])
(And8 (Const8 [m]) (Lsh8x64 _ (Const64 [c]))) && c >= 64-nlz(m) -> (Const8 [0])
cmd/compile: rewrite pairs of shifts to extensions Replaces pairs of shifts with sign/zero extension where possible. For example: (uint64(x) << 32) >> 32 -> uint64(uint32(x)) Reduces the execution time of the following code by ~4.5% on s390x: for i := 0; i < N; i++ { x += (uint64(i)<<32)>>32 } Change-Id: Idb2d56f27e80a2e1366bc995922ad3fd958c51a7 Reviewed-on: https://go-review.googlesource.com/37292 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-20 15:54:43 -05:00
// replace shifts with zero extensions
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
(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))
cmd/compile: rewrite pairs of shifts to extensions Replaces pairs of shifts with sign/zero extension where possible. For example: (uint64(x) << 32) >> 32 -> uint64(uint32(x)) Reduces the execution time of the following code by ~4.5% on s390x: for i := 0; i < N; i++ { x += (uint64(i)<<32)>>32 } Change-Id: Idb2d56f27e80a2e1366bc995922ad3fd958c51a7 Reviewed-on: https://go-review.googlesource.com/37292 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-20 15:54:43 -05:00
// replace shifts with sign extensions
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
(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))
cmd/compile: rewrite pairs of shifts to extensions Replaces pairs of shifts with sign/zero extension where possible. For example: (uint64(x) << 32) >> 32 -> uint64(uint32(x)) Reduces the execution time of the following code by ~4.5% on s390x: for i := 0; i < N; i++ { x += (uint64(i)<<32)>>32 } Change-Id: Idb2d56f27e80a2e1366bc995922ad3fd958c51a7 Reviewed-on: https://go-review.googlesource.com/37292 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-20 15:54:43 -05:00
[dev.ssa] cmd/compile/internal/ssa: rewrite rules for const comparisons Shaves ~8k off the go binary on darwin x64. Change-Id: I73396af44ae28cd4cfc675290d6858f304d45b76 Reviewed-on: https://go-review.googlesource.com/16456 Run-TryBot: Todd Neal <todd@tneal.org> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2015-10-27 17:46:53 -05:00
// constant comparisons
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(Eq(64|32|16|8) (Const(64|32|16|8) [c]) (Const(64|32|16|8) [d])) -> (ConstBool [b2i(c == d)])
(Neq(64|32|16|8) (Const(64|32|16|8) [c]) (Const(64|32|16|8) [d])) -> (ConstBool [b2i(c != d)])
(Less(64|32|16|8) (Const(64|32|16|8) [c]) (Const(64|32|16|8) [d])) -> (ConstBool [b2i(c < d)])
(Leq(64|32|16|8) (Const(64|32|16|8) [c]) (Const(64|32|16|8) [d])) -> (ConstBool [b2i(c <= d)])
[dev.ssa] cmd/compile/internal/ssa: rewrite rules for const comparisons Shaves ~8k off the go binary on darwin x64. Change-Id: I73396af44ae28cd4cfc675290d6858f304d45b76 Reviewed-on: https://go-review.googlesource.com/16456 Run-TryBot: Todd Neal <todd@tneal.org> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2015-10-27 17:46:53 -05:00
(Less64U (Const64 [c]) (Const64 [d])) -> (ConstBool [b2i(uint64(c) < uint64(d))])
(Less32U (Const32 [c]) (Const32 [d])) -> (ConstBool [b2i(uint32(c) < uint32(d))])
(Less16U (Const16 [c]) (Const16 [d])) -> (ConstBool [b2i(uint16(c) < uint16(d))])
(Less8U (Const8 [c]) (Const8 [d])) -> (ConstBool [b2i(uint8(c) < uint8(d))])
(Leq64U (Const64 [c]) (Const64 [d])) -> (ConstBool [b2i(uint64(c) <= uint64(d))])
(Leq32U (Const32 [c]) (Const32 [d])) -> (ConstBool [b2i(uint32(c) <= uint32(d))])
(Leq16U (Const16 [c]) (Const16 [d])) -> (ConstBool [b2i(uint16(c) <= uint16(d))])
(Leq8U (Const8 [c]) (Const8 [d])) -> (ConstBool [b2i(uint8(c) <= uint8(d))])
cmd/compile: remove Greater* and Geq* generic integer ops The generic Greater and Geq ops can always be replaced with the Less and Leq ops. This CL therefore removes them. This simplifies the compiler since it reduces the number of operations that need handling in both code and in rewrite rules. This will be especially true when adding control flow optimizations such as the integer-in-range optimizations in CL 165998. Change-Id: If0648b2b19998ac1bddccbf251283f3be4ec3040 Reviewed-on: https://go-review.googlesource.com/c/go/+/220417 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2020-02-20 17:46:08 +00:00
(Leq8 (Const8 [0]) (And8 _ (Const8 [c]))) && int8(c) >= 0 -> (ConstBool [1])
(Leq16 (Const16 [0]) (And16 _ (Const16 [c]))) && int16(c) >= 0 -> (ConstBool [1])
(Leq32 (Const32 [0]) (And32 _ (Const32 [c]))) && int32(c) >= 0 -> (ConstBool [1])
(Leq64 (Const64 [0]) (And64 _ (Const64 [c]))) && int64(c) >= 0 -> (ConstBool [1])
(Leq8 (Const8 [0]) (Rsh8Ux64 _ (Const64 [c]))) && c > 0 -> (ConstBool [1])
(Leq16 (Const16 [0]) (Rsh16Ux64 _ (Const64 [c]))) && c > 0 -> (ConstBool [1])
(Leq32 (Const32 [0]) (Rsh32Ux64 _ (Const64 [c]))) && c > 0 -> (ConstBool [1])
(Leq64 (Const64 [0]) (Rsh64Ux64 _ (Const64 [c]))) && c > 0 -> (ConstBool [1])
cmd/compile: improve floating point constant propagation Add generic rules to propagate floating point constants through comparisons and integer conversions. These new rules seldom trigger in the standard library so there is no performance change, however I think it is worth adding them anyway for completeness. Change-Id: I9db5222746508a2996f1cafb72f4e0cf2541de07 Reviewed-on: https://go-review.googlesource.com/63795 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-09-14 11:04:37 +01:00
// constant floating point comparisons
cmd/compile: avoid more float32 <-> float64 conversions in compiler Use the new custom truncate/extension code when storing or extracting float32 values from AuxInts to avoid the value being changed by the host platform's floating point conversion instructions (e.g. sNaN -> qNaN). Updates #27516. Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974 Reviewed-on: https://go-review.googlesource.com/134855 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-09-12 12:16:50 +01:00
(Eq32F (Const32F [c]) (Const32F [d])) -> (ConstBool [b2i(auxTo32F(c) == auxTo32F(d))])
(Eq64F (Const64F [c]) (Const64F [d])) -> (ConstBool [b2i(auxTo64F(c) == auxTo64F(d))])
(Neq32F (Const32F [c]) (Const32F [d])) -> (ConstBool [b2i(auxTo32F(c) != auxTo32F(d))])
(Neq64F (Const64F [c]) (Const64F [d])) -> (ConstBool [b2i(auxTo64F(c) != auxTo64F(d))])
(Greater32F (Const32F [c]) (Const32F [d])) -> (ConstBool [b2i(auxTo32F(c) > auxTo32F(d))])
(Greater64F (Const64F [c]) (Const64F [d])) -> (ConstBool [b2i(auxTo64F(c) > auxTo64F(d))])
(Geq32F (Const32F [c]) (Const32F [d])) -> (ConstBool [b2i(auxTo32F(c) >= auxTo32F(d))])
(Geq64F (Const64F [c]) (Const64F [d])) -> (ConstBool [b2i(auxTo64F(c) >= auxTo64F(d))])
(Less32F (Const32F [c]) (Const32F [d])) -> (ConstBool [b2i(auxTo32F(c) < auxTo32F(d))])
(Less64F (Const64F [c]) (Const64F [d])) -> (ConstBool [b2i(auxTo64F(c) < auxTo64F(d))])
(Leq32F (Const32F [c]) (Const32F [d])) -> (ConstBool [b2i(auxTo32F(c) <= auxTo32F(d))])
(Leq64F (Const64F [c]) (Const64F [d])) -> (ConstBool [b2i(auxTo64F(c) <= auxTo64F(d))])
cmd/compile: improve floating point constant propagation Add generic rules to propagate floating point constants through comparisons and integer conversions. These new rules seldom trigger in the standard library so there is no performance change, however I think it is worth adding them anyway for completeness. Change-Id: I9db5222746508a2996f1cafb72f4e0cf2541de07 Reviewed-on: https://go-review.googlesource.com/63795 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-09-14 11:04:37 +01:00
[dev.ssa] cmd/compile/internal/ssa/gen: more simplifications and foldings. Change-Id: I74d1267dbfced2663072b4f091732c0fb328690f Reviewed-on: https://go-review.googlesource.com/13641 Reviewed-by: Keith Randall <khr@golang.org>
2015-08-14 12:59:33 +02:00
// simplifications
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(Or(64|32|16|8) x x) -> x
(Or(64|32|16|8) (Const(64|32|16|8) [0]) x) -> x
cmd/compile: use | in the last repetitive generic rules This change or-ifies the last low-hanging rules in generic. Again, this is limited at short and repetitive rules, where the use or ors does not impact readability. Ran rulegen, no change in the actual compiler code. Change-Id: I972b523bc08532f173a3645b47d6936b6e1218c8 Reviewed-on: https://go-review.googlesource.com/96335 Reviewed-by: Giovanni Bajo <rasky@develer.com>
2018-02-22 14:32:57 +01:00
(Or(64|32|16|8) (Const(64|32|16|8) [-1]) _) -> (Const(64|32|16|8) [-1])
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(And(64|32|16|8) x x) -> x
(And(64|32|16|8) (Const(64|32|16|8) [-1]) x) -> x
cmd/compile: use | in the last repetitive generic rules This change or-ifies the last low-hanging rules in generic. Again, this is limited at short and repetitive rules, where the use or ors does not impact readability. Ran rulegen, no change in the actual compiler code. Change-Id: I972b523bc08532f173a3645b47d6936b6e1218c8 Reviewed-on: https://go-review.googlesource.com/96335 Reviewed-by: Giovanni Bajo <rasky@develer.com>
2018-02-22 14:32:57 +01:00
(And(64|32|16|8) (Const(64|32|16|8) [0]) _) -> (Const(64|32|16|8) [0])
(Xor(64|32|16|8) x x) -> (Const(64|32|16|8) [0])
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(Xor(64|32|16|8) (Const(64|32|16|8) [0]) x) -> x
cmd/compile: use | in the last repetitive generic rules This change or-ifies the last low-hanging rules in generic. Again, this is limited at short and repetitive rules, where the use or ors does not impact readability. Ran rulegen, no change in the actual compiler code. Change-Id: I972b523bc08532f173a3645b47d6936b6e1218c8 Reviewed-on: https://go-review.googlesource.com/96335 Reviewed-by: Giovanni Bajo <rasky@develer.com>
2018-02-22 14:32:57 +01:00
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(Add(64|32|16|8) (Const(64|32|16|8) [0]) x) -> x
cmd/compile: use | in the last repetitive generic rules This change or-ifies the last low-hanging rules in generic. Again, this is limited at short and repetitive rules, where the use or ors does not impact readability. Ran rulegen, no change in the actual compiler code. Change-Id: I972b523bc08532f173a3645b47d6936b6e1218c8 Reviewed-on: https://go-review.googlesource.com/96335 Reviewed-by: Giovanni Bajo <rasky@develer.com>
2018-02-22 14:32:57 +01:00
(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])
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(Com(64|32|16|8) (Com(64|32|16|8) x)) -> x
cmd/compile: use | in the last repetitive generic rules This change or-ifies the last low-hanging rules in generic. Again, this is limited at short and repetitive rules, where the use or ors does not impact readability. Ran rulegen, no change in the actual compiler code. Change-Id: I972b523bc08532f173a3645b47d6936b6e1218c8 Reviewed-on: https://go-review.googlesource.com/96335 Reviewed-by: Giovanni Bajo <rasky@develer.com>
2018-02-22 14:32:57 +01:00
(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)
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
cmd/compile: use more ORs in generic.rules No changes in the actual generated compiler code. Change-Id: I206a7bf7b60f70a73640119fc92974f79ed95a6b Reviewed-on: https://go-review.googlesource.com/102416 Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
2018-03-24 12:08:00 +01:00
(Add(64|32|16|8) (Const(64|32|16|8) [1]) (Com(64|32|16|8) 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
(Com(64|32|16|8) (Add(64|32|16|8) (Const(64|32|16|8) [-1]) x)) -> (Neg(64|32|16|8) x)
// -(-x) == x
(Neg(64|32|16|8) (Neg(64|32|16|8) x)) -> x
// -^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)
cmd/compile: use more ORs in generic.rules No changes in the actual generated compiler code. Change-Id: I206a7bf7b60f70a73640119fc92974f79ed95a6b Reviewed-on: https://go-review.googlesource.com/102416 Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
2018-03-24 12:08:00 +01:00
(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)
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(Xor(64|32|16|8) x (Xor(64|32|16|8) x y)) -> y
cmd/compile: logical operation identities Some rewrites to simplify logical operations. Fixes #14363 Change-Id: I45a1e8f227267cbcca0778101125f7bab776a5dd Reviewed-on: https://go-review.googlesource.com/22188 Reviewed-by: Alexandru Moșoi <alexandru@mosoi.ro> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-04-18 09:28:50 -07:00
cmd/compile: add minor bit twiddling optimizations Noticed while adding to the bitset implementation in cmd/compile/internal/gc. The (Com (Const)) optimizations were already present in the AMD64 lowered optimizations. They trigger 118, 44, 262, and 108 times respectively for int sizes 8, 16, 32, and 64 in a run of make.bash. The (Or (And)) optimization is new. It triggers 3 times for int size 8 and once for int size 64 during make.bash, in packages internal/poll, reflect, encoding/asn1, and go/types, so there is a bit of natural test coverage. Change-Id: I44072864ff88831d5ec7dce37c516d29df056e98 Reviewed-on: https://go-review.googlesource.com/41758 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-25 15:55:52 -07:00
// 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)
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Trunc64to8 (And64 (Const64 [y]) x)) && y&0xFF == 0xFF -> (Trunc64to8 x)
cmd/compile: trunc(and(x,K)) rewrite to trunc(x) for some K uint8(s.b & 0xff) ought to produce same code as uint8(s.b) but it did not. RLH found this one looking for moles to whack in the GC code. Change-Id: I883d68ec7a5746d652712be84a274a11256b3b33 Reviewed-on: https://go-review.googlesource.com/20141 Reviewed-by: Keith Randall <khr@golang.org>
2016-03-02 15:49:55 -05:00
(Trunc64to16 (And64 (Const64 [y]) x)) && y&0xFFFF == 0xFFFF -> (Trunc64to16 x)
(Trunc64to32 (And64 (Const64 [y]) x)) && y&0xFFFFFFFF == 0xFFFFFFFF -> (Trunc64to32 x)
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Trunc32to8 (And32 (Const32 [y]) x)) && y&0xFF == 0xFF -> (Trunc32to8 x)
cmd/compile: trunc(and(x,K)) rewrite to trunc(x) for some K uint8(s.b & 0xff) ought to produce same code as uint8(s.b) but it did not. RLH found this one looking for moles to whack in the GC code. Change-Id: I883d68ec7a5746d652712be84a274a11256b3b33 Reviewed-on: https://go-review.googlesource.com/20141 Reviewed-by: Keith Randall <khr@golang.org>
2016-03-02 15:49:55 -05:00
(Trunc32to16 (And32 (Const32 [y]) x)) && y&0xFFFF == 0xFFFF -> (Trunc32to16 x)
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Trunc16to8 (And16 (Const16 [y]) x)) && y&0xFF == 0xFF -> (Trunc16to8 x)
cmd/compile: trunc(and(x,K)) rewrite to trunc(x) for some K uint8(s.b & 0xff) ought to produce same code as uint8(s.b) but it did not. RLH found this one looking for moles to whack in the GC code. Change-Id: I883d68ec7a5746d652712be84a274a11256b3b33 Reviewed-on: https://go-review.googlesource.com/20141 Reviewed-by: Keith Randall <khr@golang.org>
2016-03-02 15:49:55 -05:00
cmd/compile: remove redundant extension after shift var x uint64 uint8(x >> 56) We don't need to generate any code for the uint8(). Update #15090 Change-Id: Ie1ca4e32022dccf7f7bc42d531a285521fb67872 Reviewed-on: https://go-review.googlesource.com/28191 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-08-30 15:25:20 -07:00
(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
cmd/compile: use masks instead of branches for slicing When we do var x []byte = ... y := x[i:] We can't just use y.ptr = x.ptr + i, as the new pointer may point to the next object in memory after the backing array. We used to fix this by doing: y.cap = x.cap - i delta := i if y.cap == 0 { delta = 0 } y.ptr = x.ptr + delta That generates a branch in what is otherwise straight-line code. Better to do: y.cap = x.cap - i mask := (y.cap - 1) >> 63 // -1 if y.cap==0, 0 otherwise y.ptr = x.ptr + i &^ mask It's about the same number of instructions (~4, depending on what parts are constant, and the target architecture), but it is all inline. It plays nicely with CSE, and the mask can be computed in parallel with the index (in cases where a multiply is required). It is a minor win in both speed and space. Change-Id: Ied60465a0b8abb683c02208402e5bb7ac0e8370f Reviewed-on: https://go-review.googlesource.com/32022 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2016-10-25 15:49:52 -07:00
(Slicemask (Const32 [x])) && x > 0 -> (Const32 [-1])
(Slicemask (Const32 [0])) -> (Const32 [0])
(Slicemask (Const64 [x])) && x > 0 -> (Const64 [-1])
(Slicemask (Const64 [0])) -> (Const64 [0])
[dev.ssa] cmd/compile: some SSA optimizations Some optimizations of things I've seen looking at generated code. (x+y)-x == y x-0 == x The ptr portion of the constant string "" can be nil. Also update TODO with recent changes. Change-Id: I02c41ca2f9e9e178bf889058d3e083b446672dbe Reviewed-on: https://go-review.googlesource.com/16771 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-09 20:54:34 -08:00
// simplifications often used for lengths. e.g. len(s[i:i+5])==5
cmd/compile: add | operator to make rewrite rules more succinct Instead of (And64 x x) -> x (And32 x x) -> x (And16 x x) -> x (And8 x x) -> x we can now do: (And(64|32|16|8) x x) -> x Any part of an opcode can have a parenthesized, |-separated list of possibilites. The rule is then expanded using each piece of the | combo. If there are multiple | clauses, they get expanded in tandem. (All the first positions, then all the second positions, etc.) All places | opcodes appear must have the same count. A more complicated example: (MOV(L|SS)load [off1] {sym1} (LEAQ4 [off2] {sym2} ptr idx) mem) && is32Bit(off1+off2) && canMergeSym(sym1, sym2) -> (MOV(L|SS)loadidx4 [off1+off2] {mergeSym(sym1,sym2)} ptr idx mem) This meta-rule generates 2 rules, a MOVL and a MOVSS rule. This CL is carefully orchestrated to not change the generated rules file at all. In some cases, this means we can't align the rules nicely because it changes the whitespace in the generated code. I'll clean that up as a separate step. There are many more opportunites to compactify rules using this new mechanism. I've just done some examples, there's more to do. Change-Id: I8a5e748cd0761ccbb12d09b01925b2f1f4b2f608 Reviewed-on: https://go-review.googlesource.com/86595 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-01-07 13:23:59 -08:00
(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
[dev.ssa] cmd/compile: some SSA optimizations Some optimizations of things I've seen looking at generated code. (x+y)-x == y x-0 == x The ptr portion of the constant string "" can be nil. Also update TODO with recent changes. Change-Id: I02c41ca2f9e9e178bf889058d3e083b446672dbe Reviewed-on: https://go-review.googlesource.com/16771 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-09 20:54:34 -08:00
[dev.ssa] cmd/compile/internal/ssa: handle rewrite of Phis. * Phis can have variable number of arguments, but rulegen assumed that each operation has fixed number of arguments. * Rewriting Phis is necessary to handle the following case: func f1_ssa(a bool, x int) int { v := 0 if a { v = -1 } else { v = -1 } return x|v } Change-Id: Iff6bd411b854f3d1d6d3ce21934bf566757094f2 Reviewed-on: https://go-review.googlesource.com/19412 Reviewed-by: Keith Randall <khr@golang.org>
2016-02-09 19:46:26 +01:00
// basic phi simplifications
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
(Phi (Const8 [c]) (Const8 [c])) -> (Const8 [c])
cmd/compile: define high bits of AuxInt Previously if we were only using the low bits of AuxInt, the high bits were ignored and could be junk. This CL changes that behavior to define the high bits to be the sign-extended version of the low bits for all cases. There are 2 main benefits: - Deterministic representation. This helps with CSE. (Const8 [0x1]) and (Const8 [0x101]) used to be the same "value" but CSE couldn't see them as such. - Testability. We can check that all ops leave AuxInt in a state consistent with the new rule. In the old scheme, it was hard to check whether a rule correctly used only the low-order bits. Side benefits: - ==0 and !=0 tests are easier. Drawbacks: - This differs from the runtime representation in registers, where it is important that we allow upper bits to be undefined (so we're not sign/zero-extending all the time). - Ops that treat AuxInt as unsigned (shifts, mostly) need to be a bit more careful. Change-Id: I9a685ff27e36dc03287c9ab1cecd6c0b4045c819 Reviewed-on: https://go-review.googlesource.com/21256 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2016-03-29 16:39:53 -07:00
(Phi (Const16 [c]) (Const16 [c])) -> (Const16 [c])
(Phi (Const32 [c]) (Const32 [c])) -> (Const32 [c])
[dev.ssa] cmd/compile/internal/ssa: handle rewrite of Phis. * Phis can have variable number of arguments, but rulegen assumed that each operation has fixed number of arguments. * Rewriting Phis is necessary to handle the following case: func f1_ssa(a bool, x int) int { v := 0 if a { v = -1 } else { v = -1 } return x|v } Change-Id: Iff6bd411b854f3d1d6d3ce21934bf566757094f2 Reviewed-on: https://go-review.googlesource.com/19412 Reviewed-by: Keith Randall <khr@golang.org>
2016-02-09 19:46:26 +01:00
(Phi (Const64 [c]) (Const64 [c])) -> (Const64 [c])
[dev.ssa] cmd/compile: implement non-numeric comparisons The only slice/interface comparisons that reach the backend are comparisons to nil. Funcs, maps, and channels are references types, so pointer equality is enough. Change-Id: I60a71da46a36202e9bd62ed370ab7d7f2e2800e7 Reviewed-on: https://go-review.googlesource.com/12715 Reviewed-by: Keith Randall <khr@golang.org>
2015-07-27 13:17:45 -07:00
// slice and interface comparisons
[dev.ssa] cmd/compile/internal/ssa: fix iface and slice comparisons A simpler way to do iface/slice comparisons. Fixes some cases of failed lowerings. Change-Id: Ia252bc8648293a2d460f63c41f1591785543a1e9 Reviewed-on: https://go-review.googlesource.com/14493 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-09-10 13:53:27 -07:00
// 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))
[dev.ssa] cmd/compile: implement non-numeric comparisons The only slice/interface comparisons that reach the backend are comparisons to nil. Funcs, maps, and channels are references types, so pointer equality is enough. Change-Id: I60a71da46a36202e9bd62ed370ab7d7f2e2800e7 Reviewed-on: https://go-review.googlesource.com/12715 Reviewed-by: Keith Randall <khr@golang.org>
2015-07-27 13:17:45 -07:00
[dev.ssa] cmd/compile : replace load of store with a copy Loads of stores from the same pointer with compatible types can be replaced with a copy. Change-Id: I514b3ed8e5b6a9c432946880eac67a51b1607932 Reviewed-on: https://go-review.googlesource.com/19743 Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-02-13 17:37:19 -06:00
// 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() == sizeof(t2)
-> x
(Load <t1> p1 (Store {t2} p2 _ (Store {t3} p3 x _)))
&& isSamePtr(p1, p3)
&& t1.Compare(x.Type) == types.CMPeq
&& t1.Size() == sizeof(t2)
&& disjoint(p3, sizeof(t3), p2, sizeof(t2))
-> x
(Load <t1> p1 (Store {t2} p2 _ (Store {t3} p3 _ (Store {t4} p4 x _))))
&& isSamePtr(p1, p4)
&& t1.Compare(x.Type) == types.CMPeq
&& t1.Size() == sizeof(t2)
&& disjoint(p4, sizeof(t4), p2, sizeof(t2))
&& disjoint(p4, sizeof(t4), p3, sizeof(t3))
-> x
(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() == sizeof(t2)
&& disjoint(p5, sizeof(t5), p2, sizeof(t2))
&& disjoint(p5, sizeof(t5), p3, sizeof(t3))
&& disjoint(p5, sizeof(t5), p4, sizeof(t4))
-> x
[dev.ssa] cmd/compile : replace load of store with a copy Loads of stores from the same pointer with compatible types can be replaced with a copy. Change-Id: I514b3ed8e5b6a9c432946880eac67a51b1607932 Reviewed-on: https://go-review.googlesource.com/19743 Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-02-13 17:37:19 -06:00
cmd/compile: propagate constants through math.Float{32,64}{,from}bits This CL adds generic SSA rules to propagate constants through raw bits conversions between floats and integers. This allows constants to propagate through some math functions. For example, math.Copysign(0, -1) is now constant folded to a load of -0.0. Requires a fix to the ARM assembler which loaded -0.0 as +0.0. Change-Id: I52649a4691077c7414f19d17bb599a6743c23ac2 Reviewed-on: https://go-review.googlesource.com/62250 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2017-09-08 01:31:13 +01:00
// Pass constants through math.Float{32,64}bits and math.Float{32,64}frombits
Revert "cmd/compile: don't allow NaNs in floating-point constant ops" This reverts CL 213477. Reason for revert: tests are failing on linux-mips*-rtrk builders. Change-Id: I8168f7450890233f1bd7e53930b73693c26d4dc0 Reviewed-on: https://go-review.googlesource.com/c/go/+/220897 Run-TryBot: Bryan C. Mills <bcmills@google.com> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2020-02-25 14:36:38 +00:00
(Load <t1> p1 (Store {t2} p2 (Const64 [x]) _)) && isSamePtr(p1,p2) && sizeof(t2) == 8 && is64BitFloat(t1) -> (Const64F [x])
(Load <t1> p1 (Store {t2} p2 (Const32 [x]) _)) && isSamePtr(p1,p2) && sizeof(t2) == 4 && is32BitFloat(t1) -> (Const32F [auxFrom32F(math.Float32frombits(uint32(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 (Const64F [x]) _)) && isSamePtr(p1,p2) && sizeof(t2) == 8 && is64BitInt(t1) -> (Const64 [x])
cmd/compile: avoid more float32 <-> float64 conversions in compiler Use the new custom truncate/extension code when storing or extracting float32 values from AuxInts to avoid the value being changed by the host platform's floating point conversion instructions (e.g. sNaN -> qNaN). Updates #27516. Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974 Reviewed-on: https://go-review.googlesource.com/134855 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-09-12 12:16:50 +01:00
(Load <t1> p1 (Store {t2} p2 (Const32F [x]) _)) && isSamePtr(p1,p2) && sizeof(t2) == 4 && is32BitInt(t1) -> (Const32 [int64(int32(math.Float32bits(auxTo32F(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)
cmd/compile: fix Zero-to-Load forwarding rules Avoid using values that do not dominate the block the Zero op is in. Should fix the SSA check builder. The additional OffPtr ops inserted by these rules should always be optimized away when the Load is replaced with a const zero. Fixes #25288. Change-Id: I4163b58e60364f77c8a206ba084073a58ca6320a Reviewed-on: https://go-review.googlesource.com/112136 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-05-08 16:30:48 +01:00
&& fe.CanSSA(t1)
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
&& disjoint(op, t1.Size(), p2, sizeof(t2))
cmd/compile: fix Zero-to-Load forwarding rules Avoid using values that do not dominate the block the Zero op is in. Should fix the SSA check builder. The additional OffPtr ops inserted by these rules should always be optimized away when the Load is replaced with a const zero. Fixes #25288. Change-Id: I4163b58e60364f77c8a206ba084073a58ca6320a Reviewed-on: https://go-review.googlesource.com/112136 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-05-08 16:30:48 +01:00
-> @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)
cmd/compile: fix Zero-to-Load forwarding rules Avoid using values that do not dominate the block the Zero op is in. Should fix the SSA check builder. The additional OffPtr ops inserted by these rules should always be optimized away when the Load is replaced with a const zero. Fixes #25288. Change-Id: I4163b58e60364f77c8a206ba084073a58ca6320a Reviewed-on: https://go-review.googlesource.com/112136 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-05-08 16:30:48 +01:00
&& fe.CanSSA(t1)
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
&& disjoint(op, t1.Size(), p2, sizeof(t2))
&& disjoint(op, t1.Size(), p3, sizeof(t3))
cmd/compile: fix Zero-to-Load forwarding rules Avoid using values that do not dominate the block the Zero op is in. Should fix the SSA check builder. The additional OffPtr ops inserted by these rules should always be optimized away when the Load is replaced with a const zero. Fixes #25288. Change-Id: I4163b58e60364f77c8a206ba084073a58ca6320a Reviewed-on: https://go-review.googlesource.com/112136 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-05-08 16:30:48 +01:00
-> @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)
cmd/compile: fix Zero-to-Load forwarding rules Avoid using values that do not dominate the block the Zero op is in. Should fix the SSA check builder. The additional OffPtr ops inserted by these rules should always be optimized away when the Load is replaced with a const zero. Fixes #25288. Change-Id: I4163b58e60364f77c8a206ba084073a58ca6320a Reviewed-on: https://go-review.googlesource.com/112136 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-05-08 16:30:48 +01:00
&& fe.CanSSA(t1)
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
&& disjoint(op, t1.Size(), p2, sizeof(t2))
&& disjoint(op, t1.Size(), p3, sizeof(t3))
&& disjoint(op, t1.Size(), p4, sizeof(t4))
cmd/compile: fix Zero-to-Load forwarding rules Avoid using values that do not dominate the block the Zero op is in. Should fix the SSA check builder. The additional OffPtr ops inserted by these rules should always be optimized away when the Load is replaced with a const zero. Fixes #25288. Change-Id: I4163b58e60364f77c8a206ba084073a58ca6320a Reviewed-on: https://go-review.googlesource.com/112136 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-05-08 16:30:48 +01:00
-> @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)
cmd/compile: fix Zero-to-Load forwarding rules Avoid using values that do not dominate the block the Zero op is in. Should fix the SSA check builder. The additional OffPtr ops inserted by these rules should always be optimized away when the Load is replaced with a const zero. Fixes #25288. Change-Id: I4163b58e60364f77c8a206ba084073a58ca6320a Reviewed-on: https://go-review.googlesource.com/112136 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-05-08 16:30:48 +01:00
&& fe.CanSSA(t1)
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
&& disjoint(op, t1.Size(), p2, sizeof(t2))
&& disjoint(op, t1.Size(), p3, sizeof(t3))
&& disjoint(op, t1.Size(), p4, sizeof(t4))
&& disjoint(op, t1.Size(), p5, sizeof(t5))
cmd/compile: fix Zero-to-Load forwarding rules Avoid using values that do not dominate the block the Zero op is in. Should fix the SSA check builder. The additional OffPtr ops inserted by these rules should always be optimized away when the Load is replaced with a const zero. Fixes #25288. Change-Id: I4163b58e60364f77c8a206ba084073a58ca6320a Reviewed-on: https://go-review.googlesource.com/112136 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-05-08 16:30:48 +01:00
-> @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 [0])
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is8BitInt(t1)
&& isSamePtr(p1, p2)
&& n >= o + 1
-> (Const8 [0])
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is16BitInt(t1)
&& isSamePtr(p1, p2)
&& n >= o + 2
-> (Const16 [0])
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is32BitInt(t1)
&& isSamePtr(p1, p2)
&& n >= o + 4
-> (Const32 [0])
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is64BitInt(t1)
&& isSamePtr(p1, p2)
&& n >= o + 8
-> (Const64 [0])
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is32BitFloat(t1)
&& isSamePtr(p1, p2)
&& n >= o + 4
-> (Const32F [0])
(Load <t1> (OffPtr [o] p1) (Zero [n] p2 _))
&& is64BitFloat(t1)
&& isSamePtr(p1, p2)
&& n >= o + 8
-> (Const64F [0])
cmd/compile: propagate constants through math.Float{32,64}{,from}bits This CL adds generic SSA rules to propagate constants through raw bits conversions between floats and integers. This allows constants to propagate through some math functions. For example, math.Copysign(0, -1) is now constant folded to a load of -0.0. Requires a fix to the ARM assembler which loaded -0.0 as +0.0. Change-Id: I52649a4691077c7414f19d17bb599a6743c23ac2 Reviewed-on: https://go-review.googlesource.com/62250 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2017-09-08 01:31:13 +01:00
cmd/compile: add generic rules to eliminate some unnecessary stores Eliminates stores of values that have just been loaded from the same location. Handles the common case where there are up to 3 intermediate stores to non-overlapping struct fields. For example the loads and stores of x.a, x.b and x.d in the following function are now removed: type T struct { a, b, c, d int } func f(x *T) { y := *x y.c += 8 *x = y } Before this CL (s390x): TEXT "".f(SB) MOVD "".x(R15), R5 MOVD (R5), R1 MOVD 8(R5), R2 MOVD 16(R5), R0 MOVD 24(R5), R4 ADD $8, R0, R3 STMG R1, R4, (R5) RET After this CL (s390x): TEXT "".f(SB) MOVD "".x(R15), R1 MOVD 16(R1), R0 ADD $8, R0, R0 MOVD R0, 16(R1) RET In total these rules are triggered ~5091 times during all.bash, which is broken down as: Intermediate stores | Triggered --------------------+---------- 0 | 1434 1 | 2508 2 | 888 3 | 261 --------------------+---------- Change-Id: Ia4721ae40146aceec1fdd3e65b0e9283770bfba5 Reviewed-on: https://go-review.googlesource.com/38793 Run-TryBot: Michael Munday <munday@ca.ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-29 16:37:12 -04:00
// 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() == sizeof(t1)
-> mem
(Store {t1} p1 (Load <t2> p2 oldmem) mem:(Store {t3} p3 _ oldmem))
&& isSamePtr(p1, p2)
&& t2.Size() == sizeof(t1)
&& disjoint(p1, sizeof(t1), p3, sizeof(t3))
-> mem
(Store {t1} p1 (Load <t2> p2 oldmem) mem:(Store {t3} p3 _ (Store {t4} p4 _ oldmem)))
&& isSamePtr(p1, p2)
&& t2.Size() == sizeof(t1)
&& disjoint(p1, sizeof(t1), p3, sizeof(t3))
&& disjoint(p1, sizeof(t1), p4, sizeof(t4))
-> mem
(Store {t1} p1 (Load <t2> p2 oldmem) mem:(Store {t3} p3 _ (Store {t4} p4 _ (Store {t5} p5 _ oldmem))))
&& isSamePtr(p1, p2)
&& t2.Size() == sizeof(t1)
&& disjoint(p1, sizeof(t1), p3, sizeof(t3))
&& disjoint(p1, sizeof(t1), p4, sizeof(t4))
&& disjoint(p1, sizeof(t1), p5, sizeof(t5))
-> mem
// Don't Store zeros to cleared variables.
(Store {t} (OffPtr [o] p1) x mem:(Zero [n] p2 _))
&& isConstZero(x)
&& o >= 0 && sizeof(t) + o <= n && isSamePtr(p1, p2)
-> mem
(Store {t1} op:(OffPtr [o1] p1) x mem:(Store {t2} p2 _ (Zero [n] p3 _)))
&& isConstZero(x)
&& o1 >= 0 && sizeof(t1) + o1 <= n && isSamePtr(p1, p3)
&& disjoint(op, sizeof(t1), p2, sizeof(t2))
-> mem
(Store {t1} op:(OffPtr [o1] p1) x mem:(Store {t2} p2 _ (Store {t3} p3 _ (Zero [n] p4 _))))
&& isConstZero(x)
&& o1 >= 0 && sizeof(t1) + o1 <= n && isSamePtr(p1, p4)
&& disjoint(op, sizeof(t1), p2, sizeof(t2))
&& disjoint(op, sizeof(t1), p3, sizeof(t3))
-> mem
(Store {t1} op:(OffPtr [o1] p1) x mem:(Store {t2} p2 _ (Store {t3} p3 _ (Store {t4} p4 _ (Zero [n] p5 _)))))
&& isConstZero(x)
&& o1 >= 0 && sizeof(t1) + o1 <= n && isSamePtr(p1, p5)
&& disjoint(op, sizeof(t1), p2, sizeof(t2))
&& disjoint(op, sizeof(t1), p3, sizeof(t3))
&& disjoint(op, sizeof(t1), p4, sizeof(t4))
-> mem
cmd/compile: add generic rules to eliminate some unnecessary stores Eliminates stores of values that have just been loaded from the same location. Handles the common case where there are up to 3 intermediate stores to non-overlapping struct fields. For example the loads and stores of x.a, x.b and x.d in the following function are now removed: type T struct { a, b, c, d int } func f(x *T) { y := *x y.c += 8 *x = y } Before this CL (s390x): TEXT "".f(SB) MOVD "".x(R15), R5 MOVD (R5), R1 MOVD 8(R5), R2 MOVD 16(R5), R0 MOVD 24(R5), R4 ADD $8, R0, R3 STMG R1, R4, (R5) RET After this CL (s390x): TEXT "".f(SB) MOVD "".x(R15), R1 MOVD 16(R1), R0 ADD $8, R0, R0 MOVD R0, 16(R1) RET In total these rules are triggered ~5091 times during all.bash, which is broken down as: Intermediate stores | Triggered --------------------+---------- 0 | 1434 1 | 2508 2 | 888 3 | 261 --------------------+---------- Change-Id: Ia4721ae40146aceec1fdd3e65b0e9283770bfba5 Reviewed-on: https://go-review.googlesource.com/38793 Run-TryBot: Michael Munday <munday@ca.ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-29 16:37:12 -04:00
cmd/compile: collapse OffPtr sequences This triggers an astonishing 160k times during make.bash. The second biggest generic rewrite triggers 100k times. However, this is really just moving rewrites that were happening at the architecture level to the generic level. Change-Id: Ife06fe5234f31433328460cb2e0741c071deda41 Reviewed-on: https://go-review.googlesource.com/20235 Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-03-04 14:12:44 -08:00
// Collapse OffPtr
(OffPtr (OffPtr p [b]) [a]) -> (OffPtr p [a+b])
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
(OffPtr p [0]) && v.Type.Compare(p.Type) == types.CMPeq -> p
cmd/compile: collapse OffPtr sequences This triggers an astonishing 160k times during make.bash. The second biggest generic rewrite triggers 100k times. However, this is really just moving rewrites that were happening at the architecture level to the generic level. Change-Id: Ife06fe5234f31433328460cb2e0741c071deda41 Reviewed-on: https://go-review.googlesource.com/20235 Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-03-04 14:12:44 -08:00
[dev.ssa] cmd/internal/ssa: Handle more instructions + some cleanup Add & as an input op. Add several output ops (loads & stores, TESTB, LEAQglobal, branches, memcopy) Some other small things: - Add exprAddr to builder to generate addresses of expressions. Use it in various places that had ad-hoc code. - Separate out nil & bounds check generation to separate functions. - Add explicit FP and SP ops so we dont need specialized *FP and *SP opcodes. - Fix fallthrough at end of functions with no return values. - rematerialization of more opcodes. Change-Id: I781decfcef9770fb15f0cd6b061547f7824a2d5e Reviewed-on: https://go-review.googlesource.com/10213 Reviewed-by: Alan Donovan <adonovan@google.com>
2015-05-18 16:44:20 -07:00
// indexing operations
[dev.ssa] cmd/internal/gc: convert standard IR into SSA. Hook into the current compiler to convert the existing IR (after walk) into SSA. Any function ending in "_ssa" will take this path. The resulting assembly is printed and then discarded. Use gc.Type directly in ssa instead of a wrapper for go types. It makes the IR->SSA rewrite a lot simpler. Only a few opcodes are implemented in this change. It is enough to compile simple examples like func f(p *int) int { return *p } func g(a []int, i int) int { return a[i] } Change-Id: I5e18841b752a83ca0519aa1b2d36ef02ce1de6f9 Reviewed-on: https://go-review.googlesource.com/8971 Reviewed-by: Alan Donovan <adonovan@google.com>
2015-04-15 15:51:25 -07:00
// Note: bounds check has already been done
cmd/compile/internal/types: remove ElemType wrapper This was an artifact from when we had a separate ssa.Type interface to break circular dependency between packages ssa and gc. It's no longer needed now that package ssa directly uses package types. Change-Id: I6a93e5d79082815f7f0eb89507381969cc6cb403 Reviewed-on: https://go-review.googlesource.com/109137 Run-TryBot: Matthew Dempsky <mdempsky@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-04-24 13:39:51 -07:00
(PtrIndex <t> ptr idx) && config.PtrSize == 4 -> (AddPtr ptr (Mul32 <typ.Int> idx (Const32 <typ.Int> [t.Elem().Size()])))
(PtrIndex <t> ptr idx) && config.PtrSize == 8 -> (AddPtr ptr (Mul64 <typ.Int> idx (Const64 <typ.Int> [t.Elem().Size()])))
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
// struct operations
(StructSelect (StructMake1 x)) -> x
(StructSelect [0] (StructMake2 x _)) -> x
(StructSelect [1] (StructMake2 _ x)) -> x
(StructSelect [0] (StructMake3 x _ _)) -> x
(StructSelect [1] (StructMake3 _ x _)) -> x
(StructSelect [2] (StructMake3 _ _ x)) -> x
(StructSelect [0] (StructMake4 x _ _ _)) -> x
(StructSelect [1] (StructMake4 _ x _ _)) -> x
(StructSelect [2] (StructMake4 _ _ x _)) -> x
(StructSelect [3] (StructMake4 _ _ _ x)) -> x
cmd/compile: evaluate config as needed in rewrite rules Prior to this CL, config was an explicit argument to the SSA rewrite rules, and rules that needed a Frontend got at it via config. An upcoming CL moves Frontend from Config to Func, so rules can no longer reach Frontend via Config. Passing a Frontend as an argument to the rewrite rules causes a 2-3% regression in compile times. This CL takes a different approach: It treats the variable names "config" and "fe" as special and calculates them as needed. The "as needed part" is also important to performance: If they are calculated eagerly, the nilchecks themselves cause a regression. This introduces a little bit of magic into the rewrite generator. However, from the perspective of the rules, the config variable was already more or less magic. And it makes the upcoming changes much clearer. Passes toolstash -cmp. Change-Id: I173f2bcc124cba43d53138bfa3775e21316a9107 Reviewed-on: https://go-review.googlesource.com/38326 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2017-03-17 10:50:20 -07:00
(Load <t> _ _) && t.IsStruct() && t.NumFields() == 0 && fe.CanSSA(t) ->
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(StructMake0)
cmd/compile: evaluate config as needed in rewrite rules Prior to this CL, config was an explicit argument to the SSA rewrite rules, and rules that needed a Frontend got at it via config. An upcoming CL moves Frontend from Config to Func, so rules can no longer reach Frontend via Config. Passing a Frontend as an argument to the rewrite rules causes a 2-3% regression in compile times. This CL takes a different approach: It treats the variable names "config" and "fe" as special and calculates them as needed. The "as needed part" is also important to performance: If they are calculated eagerly, the nilchecks themselves cause a regression. This introduces a little bit of magic into the rewrite generator. However, from the perspective of the rules, the config variable was already more or less magic. And it makes the upcoming changes much clearer. Passes toolstash -cmp. Change-Id: I173f2bcc124cba43d53138bfa3775e21316a9107 Reviewed-on: https://go-review.googlesource.com/38326 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2017-03-17 10:50:20 -07:00
(Load <t> ptr mem) && t.IsStruct() && t.NumFields() == 1 && fe.CanSSA(t) ->
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(StructMake1
cmd/compile: emit OffPtr for first field in SSA'd structs Given (Store [c] (OffPtr <T1> [0] (Addr <T> _)) _ (Store [c] (Addr <T> _) _ _)) dead store elimination doesn't eliminate the inner Store, because it addresses a type of a different width than the first store. When decomposing StructMake operations, always generate an OffPtr to address struct fields so that dead stores to the first field of the struct can be optimized away. benchmarks affected on darwin/amd64: HTTPClientServer-8 73.2µs ± 1% 72.7µs ± 1% -0.69% (p=0.022 n=9+10) TimeParse-8 304ns ± 1% 300ns ± 0% -1.61% (p=0.000 n=9+9) RegexpMatchEasy1_32-8 80.1ns ± 0% 79.5ns ± 1% -0.84% (p=0.000 n=8+9) GobDecode-8 6.78ms ± 0% 6.81ms ± 1% +0.46% (p=0.000 n=9+10) Gunzip-8 36.1ms ± 1% 36.2ms ± 0% +0.37% (p=0.019 n=10+10) JSONEncode-8 15.6ms ± 0% 15.7ms ± 0% +0.69% (p=0.000 n=9+10) Change-Id: Ia80d73fd047f9400c616ca64fdee4f438a0e7f21 Reviewed-on: https://go-review.googlesource.com/37769 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-04 16:17:12 -08:00
(Load <t.FieldType(0)> (OffPtr <t.FieldType(0).PtrTo()> [0] ptr) mem))
cmd/compile: evaluate config as needed in rewrite rules Prior to this CL, config was an explicit argument to the SSA rewrite rules, and rules that needed a Frontend got at it via config. An upcoming CL moves Frontend from Config to Func, so rules can no longer reach Frontend via Config. Passing a Frontend as an argument to the rewrite rules causes a 2-3% regression in compile times. This CL takes a different approach: It treats the variable names "config" and "fe" as special and calculates them as needed. The "as needed part" is also important to performance: If they are calculated eagerly, the nilchecks themselves cause a regression. This introduces a little bit of magic into the rewrite generator. However, from the perspective of the rules, the config variable was already more or less magic. And it makes the upcoming changes much clearer. Passes toolstash -cmp. Change-Id: I173f2bcc124cba43d53138bfa3775e21316a9107 Reviewed-on: https://go-review.googlesource.com/38326 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2017-03-17 10:50:20 -07:00
(Load <t> ptr mem) && t.IsStruct() && t.NumFields() == 2 && fe.CanSSA(t) ->
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(StructMake2
cmd/compile: emit OffPtr for first field in SSA'd structs Given (Store [c] (OffPtr <T1> [0] (Addr <T> _)) _ (Store [c] (Addr <T> _) _ _)) dead store elimination doesn't eliminate the inner Store, because it addresses a type of a different width than the first store. When decomposing StructMake operations, always generate an OffPtr to address struct fields so that dead stores to the first field of the struct can be optimized away. benchmarks affected on darwin/amd64: HTTPClientServer-8 73.2µs ± 1% 72.7µs ± 1% -0.69% (p=0.022 n=9+10) TimeParse-8 304ns ± 1% 300ns ± 0% -1.61% (p=0.000 n=9+9) RegexpMatchEasy1_32-8 80.1ns ± 0% 79.5ns ± 1% -0.84% (p=0.000 n=8+9) GobDecode-8 6.78ms ± 0% 6.81ms ± 1% +0.46% (p=0.000 n=9+10) Gunzip-8 36.1ms ± 1% 36.2ms ± 0% +0.37% (p=0.019 n=10+10) JSONEncode-8 15.6ms ± 0% 15.7ms ± 0% +0.69% (p=0.000 n=9+10) Change-Id: Ia80d73fd047f9400c616ca64fdee4f438a0e7f21 Reviewed-on: https://go-review.googlesource.com/37769 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-04 16:17:12 -08:00
(Load <t.FieldType(0)> (OffPtr <t.FieldType(0).PtrTo()> [0] ptr) mem)
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(Load <t.FieldType(1)> (OffPtr <t.FieldType(1).PtrTo()> [t.FieldOff(1)] ptr) mem))
cmd/compile: evaluate config as needed in rewrite rules Prior to this CL, config was an explicit argument to the SSA rewrite rules, and rules that needed a Frontend got at it via config. An upcoming CL moves Frontend from Config to Func, so rules can no longer reach Frontend via Config. Passing a Frontend as an argument to the rewrite rules causes a 2-3% regression in compile times. This CL takes a different approach: It treats the variable names "config" and "fe" as special and calculates them as needed. The "as needed part" is also important to performance: If they are calculated eagerly, the nilchecks themselves cause a regression. This introduces a little bit of magic into the rewrite generator. However, from the perspective of the rules, the config variable was already more or less magic. And it makes the upcoming changes much clearer. Passes toolstash -cmp. Change-Id: I173f2bcc124cba43d53138bfa3775e21316a9107 Reviewed-on: https://go-review.googlesource.com/38326 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2017-03-17 10:50:20 -07:00
(Load <t> ptr mem) && t.IsStruct() && t.NumFields() == 3 && fe.CanSSA(t) ->
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(StructMake3
cmd/compile: emit OffPtr for first field in SSA'd structs Given (Store [c] (OffPtr <T1> [0] (Addr <T> _)) _ (Store [c] (Addr <T> _) _ _)) dead store elimination doesn't eliminate the inner Store, because it addresses a type of a different width than the first store. When decomposing StructMake operations, always generate an OffPtr to address struct fields so that dead stores to the first field of the struct can be optimized away. benchmarks affected on darwin/amd64: HTTPClientServer-8 73.2µs ± 1% 72.7µs ± 1% -0.69% (p=0.022 n=9+10) TimeParse-8 304ns ± 1% 300ns ± 0% -1.61% (p=0.000 n=9+9) RegexpMatchEasy1_32-8 80.1ns ± 0% 79.5ns ± 1% -0.84% (p=0.000 n=8+9) GobDecode-8 6.78ms ± 0% 6.81ms ± 1% +0.46% (p=0.000 n=9+10) Gunzip-8 36.1ms ± 1% 36.2ms ± 0% +0.37% (p=0.019 n=10+10) JSONEncode-8 15.6ms ± 0% 15.7ms ± 0% +0.69% (p=0.000 n=9+10) Change-Id: Ia80d73fd047f9400c616ca64fdee4f438a0e7f21 Reviewed-on: https://go-review.googlesource.com/37769 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-04 16:17:12 -08:00
(Load <t.FieldType(0)> (OffPtr <t.FieldType(0).PtrTo()> [0] ptr) mem)
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(Load <t.FieldType(1)> (OffPtr <t.FieldType(1).PtrTo()> [t.FieldOff(1)] ptr) mem)
(Load <t.FieldType(2)> (OffPtr <t.FieldType(2).PtrTo()> [t.FieldOff(2)] ptr) mem))
cmd/compile: evaluate config as needed in rewrite rules Prior to this CL, config was an explicit argument to the SSA rewrite rules, and rules that needed a Frontend got at it via config. An upcoming CL moves Frontend from Config to Func, so rules can no longer reach Frontend via Config. Passing a Frontend as an argument to the rewrite rules causes a 2-3% regression in compile times. This CL takes a different approach: It treats the variable names "config" and "fe" as special and calculates them as needed. The "as needed part" is also important to performance: If they are calculated eagerly, the nilchecks themselves cause a regression. This introduces a little bit of magic into the rewrite generator. However, from the perspective of the rules, the config variable was already more or less magic. And it makes the upcoming changes much clearer. Passes toolstash -cmp. Change-Id: I173f2bcc124cba43d53138bfa3775e21316a9107 Reviewed-on: https://go-review.googlesource.com/38326 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2017-03-17 10:50:20 -07:00
(Load <t> ptr mem) && t.IsStruct() && t.NumFields() == 4 && fe.CanSSA(t) ->
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(StructMake4
cmd/compile: emit OffPtr for first field in SSA'd structs Given (Store [c] (OffPtr <T1> [0] (Addr <T> _)) _ (Store [c] (Addr <T> _) _ _)) dead store elimination doesn't eliminate the inner Store, because it addresses a type of a different width than the first store. When decomposing StructMake operations, always generate an OffPtr to address struct fields so that dead stores to the first field of the struct can be optimized away. benchmarks affected on darwin/amd64: HTTPClientServer-8 73.2µs ± 1% 72.7µs ± 1% -0.69% (p=0.022 n=9+10) TimeParse-8 304ns ± 1% 300ns ± 0% -1.61% (p=0.000 n=9+9) RegexpMatchEasy1_32-8 80.1ns ± 0% 79.5ns ± 1% -0.84% (p=0.000 n=8+9) GobDecode-8 6.78ms ± 0% 6.81ms ± 1% +0.46% (p=0.000 n=9+10) Gunzip-8 36.1ms ± 1% 36.2ms ± 0% +0.37% (p=0.019 n=10+10) JSONEncode-8 15.6ms ± 0% 15.7ms ± 0% +0.69% (p=0.000 n=9+10) Change-Id: Ia80d73fd047f9400c616ca64fdee4f438a0e7f21 Reviewed-on: https://go-review.googlesource.com/37769 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-04 16:17:12 -08:00
(Load <t.FieldType(0)> (OffPtr <t.FieldType(0).PtrTo()> [0] ptr) mem)
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(Load <t.FieldType(1)> (OffPtr <t.FieldType(1).PtrTo()> [t.FieldOff(1)] ptr) mem)
(Load <t.FieldType(2)> (OffPtr <t.FieldType(2).PtrTo()> [t.FieldOff(2)] ptr) mem)
(Load <t.FieldType(3)> (OffPtr <t.FieldType(3).PtrTo()> [t.FieldOff(3)] ptr) mem))
cmd/compile: evaluate config as needed in rewrite rules Prior to this CL, config was an explicit argument to the SSA rewrite rules, and rules that needed a Frontend got at it via config. An upcoming CL moves Frontend from Config to Func, so rules can no longer reach Frontend via Config. Passing a Frontend as an argument to the rewrite rules causes a 2-3% regression in compile times. This CL takes a different approach: It treats the variable names "config" and "fe" as special and calculates them as needed. The "as needed part" is also important to performance: If they are calculated eagerly, the nilchecks themselves cause a regression. This introduces a little bit of magic into the rewrite generator. However, from the perspective of the rules, the config variable was already more or less magic. And it makes the upcoming changes much clearer. Passes toolstash -cmp. Change-Id: I173f2bcc124cba43d53138bfa3775e21316a9107 Reviewed-on: https://go-review.googlesource.com/38326 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2017-03-17 10:50:20 -07:00
(StructSelect [i] x:(Load <t> ptr mem)) && !fe.CanSSA(t) ->
cmd/compile: allow naming of subexpressions Allow names to be used for subexpressions of match rules. For example: (OpA x:(OpB y)) -> ..use x here to refer to the OpB value.. This gets rid of the .Args[0].Args[0]... way of naming we used to use. While we're here, give all subexpression matches names instead of recomputing them with .Args[i] sequences each time they are referenced. Makes the generated rule code a bit smaller. Change-Id: Ie42139f6f208933b75bd2ae8bd34e95419bc0e4e Reviewed-on: https://go-review.googlesource.com/20997 Run-TryBot: Todd Neal <todd@tneal.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-03-21 16:18:45 -07:00
@x.Block (Load <v.Type> (OffPtr <v.Type.PtrTo()> [t.FieldOff(int(i))] ptr) mem)
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(Store _ (StructMake0) mem) -> mem
(Store dst (StructMake1 <t> f0) mem) ->
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(0)} (OffPtr <t.FieldType(0).PtrTo()> [0] dst) f0 mem)
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(Store dst (StructMake2 <t> f0 f1) mem) ->
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(1)}
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(OffPtr <t.FieldType(1).PtrTo()> [t.FieldOff(1)] dst)
f1
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(0)}
cmd/compile: fix OffPtr type in 2-field struct Store rule The type of the OffPtr for the first field was incorrect. It should have been a pointer to the field type, rather than the field type itself. Fixes #19475. Change-Id: I3960b404da0f4bee759331126cce6140d2ce1df7 Reviewed-on: https://go-review.googlesource.com/37869 Run-TryBot: Michael Munday <munday@ca.ibm.com> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-09 13:03:07 -05:00
(OffPtr <t.FieldType(0).PtrTo()> [0] dst)
cmd/compile: emit OffPtr for first field in SSA'd structs Given (Store [c] (OffPtr <T1> [0] (Addr <T> _)) _ (Store [c] (Addr <T> _) _ _)) dead store elimination doesn't eliminate the inner Store, because it addresses a type of a different width than the first store. When decomposing StructMake operations, always generate an OffPtr to address struct fields so that dead stores to the first field of the struct can be optimized away. benchmarks affected on darwin/amd64: HTTPClientServer-8 73.2µs ± 1% 72.7µs ± 1% -0.69% (p=0.022 n=9+10) TimeParse-8 304ns ± 1% 300ns ± 0% -1.61% (p=0.000 n=9+9) RegexpMatchEasy1_32-8 80.1ns ± 0% 79.5ns ± 1% -0.84% (p=0.000 n=8+9) GobDecode-8 6.78ms ± 0% 6.81ms ± 1% +0.46% (p=0.000 n=9+10) Gunzip-8 36.1ms ± 1% 36.2ms ± 0% +0.37% (p=0.019 n=10+10) JSONEncode-8 15.6ms ± 0% 15.7ms ± 0% +0.69% (p=0.000 n=9+10) Change-Id: Ia80d73fd047f9400c616ca64fdee4f438a0e7f21 Reviewed-on: https://go-review.googlesource.com/37769 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-04 16:17:12 -08:00
f0 mem))
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(Store dst (StructMake3 <t> f0 f1 f2) mem) ->
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(2)}
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(OffPtr <t.FieldType(2).PtrTo()> [t.FieldOff(2)] dst)
f2
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(1)}
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(OffPtr <t.FieldType(1).PtrTo()> [t.FieldOff(1)] dst)
f1
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(0)}
cmd/compile: emit OffPtr for first field in SSA'd structs Given (Store [c] (OffPtr <T1> [0] (Addr <T> _)) _ (Store [c] (Addr <T> _) _ _)) dead store elimination doesn't eliminate the inner Store, because it addresses a type of a different width than the first store. When decomposing StructMake operations, always generate an OffPtr to address struct fields so that dead stores to the first field of the struct can be optimized away. benchmarks affected on darwin/amd64: HTTPClientServer-8 73.2µs ± 1% 72.7µs ± 1% -0.69% (p=0.022 n=9+10) TimeParse-8 304ns ± 1% 300ns ± 0% -1.61% (p=0.000 n=9+9) RegexpMatchEasy1_32-8 80.1ns ± 0% 79.5ns ± 1% -0.84% (p=0.000 n=8+9) GobDecode-8 6.78ms ± 0% 6.81ms ± 1% +0.46% (p=0.000 n=9+10) Gunzip-8 36.1ms ± 1% 36.2ms ± 0% +0.37% (p=0.019 n=10+10) JSONEncode-8 15.6ms ± 0% 15.7ms ± 0% +0.69% (p=0.000 n=9+10) Change-Id: Ia80d73fd047f9400c616ca64fdee4f438a0e7f21 Reviewed-on: https://go-review.googlesource.com/37769 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-04 16:17:12 -08:00
(OffPtr <t.FieldType(0).PtrTo()> [0] dst)
f0 mem)))
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(Store dst (StructMake4 <t> f0 f1 f2 f3) mem) ->
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(3)}
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(OffPtr <t.FieldType(3).PtrTo()> [t.FieldOff(3)] dst)
f3
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(2)}
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(OffPtr <t.FieldType(2).PtrTo()> [t.FieldOff(2)] dst)
f2
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(1)}
[dev.ssa] cmd/compile: Allow structs to be SSAd Break small structs up into their components so they can be registerized. Change StructSelect to use field indexes instead of field offsets, as field offsets aren't unique in the presence of zero-sized fields. Change-Id: I2f1dc89f7fa58e1cf58aa1a32b238959d53f62e4 Reviewed-on: https://go-review.googlesource.com/18570 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-01-11 21:05:33 -08:00
(OffPtr <t.FieldType(1).PtrTo()> [t.FieldOff(1)] dst)
f1
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store {t.FieldType(0)}
cmd/compile: emit OffPtr for first field in SSA'd structs Given (Store [c] (OffPtr <T1> [0] (Addr <T> _)) _ (Store [c] (Addr <T> _) _ _)) dead store elimination doesn't eliminate the inner Store, because it addresses a type of a different width than the first store. When decomposing StructMake operations, always generate an OffPtr to address struct fields so that dead stores to the first field of the struct can be optimized away. benchmarks affected on darwin/amd64: HTTPClientServer-8 73.2µs ± 1% 72.7µs ± 1% -0.69% (p=0.022 n=9+10) TimeParse-8 304ns ± 1% 300ns ± 0% -1.61% (p=0.000 n=9+9) RegexpMatchEasy1_32-8 80.1ns ± 0% 79.5ns ± 1% -0.84% (p=0.000 n=8+9) GobDecode-8 6.78ms ± 0% 6.81ms ± 1% +0.46% (p=0.000 n=9+10) Gunzip-8 36.1ms ± 1% 36.2ms ± 0% +0.37% (p=0.019 n=10+10) JSONEncode-8 15.6ms ± 0% 15.7ms ± 0% +0.69% (p=0.000 n=9+10) Change-Id: Ia80d73fd047f9400c616ca64fdee4f438a0e7f21 Reviewed-on: https://go-review.googlesource.com/37769 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-04 16:17:12 -08:00
(OffPtr <t.FieldType(0).PtrTo()> [0] dst)
f0 mem))))
[dev.ssa] cmd/internal/ssa: Handle more instructions + some cleanup Add & as an input op. Add several output ops (loads & stores, TESTB, LEAQglobal, branches, memcopy) Some other small things: - Add exprAddr to builder to generate addresses of expressions. Use it in various places that had ad-hoc code. - Separate out nil & bounds check generation to separate functions. - Add explicit FP and SP ops so we dont need specialized *FP and *SP opcodes. - Fix fallthrough at end of functions with no return values. - rematerialization of more opcodes. Change-Id: I781decfcef9770fb15f0cd6b061547f7824a2d5e Reviewed-on: https://go-review.googlesource.com/10213 Reviewed-by: Alan Donovan <adonovan@google.com>
2015-05-18 16:44:20 -07:00
cmd/compile: do more type conversion inline The code to do the conversion is smaller than the call to the runtime. The 1-result asserts need to call panic if they fail, but that code is out of line. The only conversions left in the runtime are those which might allocate and those which might need to generate an itab. Given the following types: type E interface{} type I interface { foo() } type I2 iterface { foo(); bar() } type Big [10]int func (b Big) foo() { ... } This CL inlines the following conversions: was assertE2T var e E = ... b := i.(Big) was assertE2T2 var e E = ... b, ok := i.(Big) was assertI2T var i I = ... b := i.(Big) was assertI2T2 var i I = ... b, ok := i.(Big) was assertI2E var i I = ... e := i.(E) was assertI2E2 var i I = ... e, ok := i.(E) These are the remaining runtime calls: convT2E: var b Big = ... var e E = b convT2I: var b Big = ... var i I = b convI2I: var i2 I2 = ... var i I = i2 assertE2I: var e E = ... i := e.(I) assertE2I2: var e E = ... i, ok := e.(I) assertI2I: var i I = ... i2 := i.(I2) assertI2I2: var i I = ... i2, ok := i.(I2) Fixes #17405 Fixes #8422 Change-Id: Ida2367bf8ce3cd2c6bb599a1814f1d275afabe21 Reviewed-on: https://go-review.googlesource.com/32313 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-10-28 11:37:45 -07:00
// Putting struct{*byte} and similar into direct interfaces.
cmd/compile: make [0]T and [1]T SSAable types We used to have to keep on-stack copies of these types. Now they can be registerized. [0]T is kind of trivial but might as well handle it. This change enables another change I'm working on to improve how x.(T) expressions are handled (#17405). This CL helps because now all types that are direct interface types are registerizeable (e.g. [1]*byte). No higher-degree arrays for now because non-constant indexes are hard. Update #17405 Change-Id: I2399940965d17b3969ae66f6fe447a8cefdd6edd Reviewed-on: https://go-review.googlesource.com/32416 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-10-30 21:10:03 -07:00
(IMake typ (StructMake1 val)) -> (IMake typ val)
cmd/compile: fix typing of IData opcodes The rules for extracting the interface data word don't leave the result typed correctly. If I do i.([1]*int)[0], the result should have type *int, not [1]*int. Using (IData x) for the result keeps the typing of the original top-level Value. I don't think this would ever cause a real codegen bug, bug fixing it at least makes the typing shown in ssa.html more consistent. Change-Id: I239d821c394e58347639387981b0510d13b2f7b7 Reviewed-on: https://go-review.googlesource.com/c/go/+/204042 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2019-10-29 13:14:58 -07:00
(StructSelect [0] (IData x)) -> (IData x)
cmd/compile: make [0]T and [1]T SSAable types We used to have to keep on-stack copies of these types. Now they can be registerized. [0]T is kind of trivial but might as well handle it. This change enables another change I'm working on to improve how x.(T) expressions are handled (#17405). This CL helps because now all types that are direct interface types are registerizeable (e.g. [1]*byte). No higher-degree arrays for now because non-constant indexes are hard. Update #17405 Change-Id: I2399940965d17b3969ae66f6fe447a8cefdd6edd Reviewed-on: https://go-review.googlesource.com/32416 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-10-30 21:10:03 -07:00
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
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
(Store {t} dst (Load src mem) mem) && !fe.CanSSA(t.(*types.Type)) ->
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} [sizeof(t)] dst src mem)
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
(Store {t} dst (Load src mem) (VarDef {x} mem)) && !fe.CanSSA(t.(*types.Type)) ->
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} [sizeof(t)] dst src (VarDef {x} mem))
[dev.ssa] cmd/compile: add compose/decompose for complex, phi, constants Still to do: arithmetic Change-Id: I31fd23b34980c9ed4b4e304b8597134b2ba6ca5c Reviewed-on: https://go-review.googlesource.com/14024 Reviewed-by: Keith Randall <khr@golang.org>
2015-08-28 14:24:10 -04:00
cmd/compile: make [0]T and [1]T SSAable types We used to have to keep on-stack copies of these types. Now they can be registerized. [0]T is kind of trivial but might as well handle it. This change enables another change I'm working on to improve how x.(T) expressions are handled (#17405). This CL helps because now all types that are direct interface types are registerizeable (e.g. [1]*byte). No higher-degree arrays for now because non-constant indexes are hard. Update #17405 Change-Id: I2399940965d17b3969ae66f6fe447a8cefdd6edd Reviewed-on: https://go-review.googlesource.com/32416 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-10-30 21:10:03 -07:00
// array ops
(ArraySelect (ArrayMake1 x)) -> x
(Load <t> _ _) && t.IsArray() && t.NumElem() == 0 ->
(ArrayMake0)
cmd/compile: evaluate config as needed in rewrite rules Prior to this CL, config was an explicit argument to the SSA rewrite rules, and rules that needed a Frontend got at it via config. An upcoming CL moves Frontend from Config to Func, so rules can no longer reach Frontend via Config. Passing a Frontend as an argument to the rewrite rules causes a 2-3% regression in compile times. This CL takes a different approach: It treats the variable names "config" and "fe" as special and calculates them as needed. The "as needed part" is also important to performance: If they are calculated eagerly, the nilchecks themselves cause a regression. This introduces a little bit of magic into the rewrite generator. However, from the perspective of the rules, the config variable was already more or less magic. And it makes the upcoming changes much clearer. Passes toolstash -cmp. Change-Id: I173f2bcc124cba43d53138bfa3775e21316a9107 Reviewed-on: https://go-review.googlesource.com/38326 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2017-03-17 10:50:20 -07:00
(Load <t> ptr mem) && t.IsArray() && t.NumElem() == 1 && fe.CanSSA(t) ->
cmd/compile/internal/types: remove ElemType wrapper This was an artifact from when we had a separate ssa.Type interface to break circular dependency between packages ssa and gc. It's no longer needed now that package ssa directly uses package types. Change-Id: I6a93e5d79082815f7f0eb89507381969cc6cb403 Reviewed-on: https://go-review.googlesource.com/109137 Run-TryBot: Matthew Dempsky <mdempsky@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-04-24 13:39:51 -07:00
(ArrayMake1 (Load <t.Elem()> ptr mem))
cmd/compile: make [0]T and [1]T SSAable types We used to have to keep on-stack copies of these types. Now they can be registerized. [0]T is kind of trivial but might as well handle it. This change enables another change I'm working on to improve how x.(T) expressions are handled (#17405). This CL helps because now all types that are direct interface types are registerizeable (e.g. [1]*byte). No higher-degree arrays for now because non-constant indexes are hard. Update #17405 Change-Id: I2399940965d17b3969ae66f6fe447a8cefdd6edd Reviewed-on: https://go-review.googlesource.com/32416 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-10-30 21:10:03 -07:00
(Store _ (ArrayMake0) mem) -> mem
cmd/compile: use type information in Aux for Store size Remove size AuxInt in Store, and alignment in Move/Zero. We still pass size AuxInt to Move/Zero, as it is used for partial Move/Zero lowering (e.g. cmd/compile/internal/ssa/gen/386.rules:288). SizeAndAlign is gone. Passes "toolstash -cmp" on std. Change-Id: I1ca34652b65dd30de886940e789fcf41d521475d Reviewed-on: https://go-review.googlesource.com/38150 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-13 21:51:08 -04:00
(Store dst (ArrayMake1 e) mem) -> (Store {e.Type} dst e mem)
cmd/compile: make [0]T and [1]T SSAable types We used to have to keep on-stack copies of these types. Now they can be registerized. [0]T is kind of trivial but might as well handle it. This change enables another change I'm working on to improve how x.(T) expressions are handled (#17405). This CL helps because now all types that are direct interface types are registerizeable (e.g. [1]*byte). No higher-degree arrays for now because non-constant indexes are hard. Update #17405 Change-Id: I2399940965d17b3969ae66f6fe447a8cefdd6edd Reviewed-on: https://go-review.googlesource.com/32416 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-10-30 21:10:03 -07:00
cmd/compile: fix typing of IData opcodes The rules for extracting the interface data word don't leave the result typed correctly. If I do i.([1]*int)[0], the result should have type *int, not [1]*int. Using (IData x) for the result keeps the typing of the original top-level Value. I don't think this would ever cause a real codegen bug, bug fixing it at least makes the typing shown in ssa.html more consistent. Change-Id: I239d821c394e58347639387981b0510d13b2f7b7 Reviewed-on: https://go-review.googlesource.com/c/go/+/204042 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2019-10-29 13:14:58 -07:00
// Putting [1]*byte and similar into direct interfaces.
cmd/compile: make [0]T and [1]T SSAable types We used to have to keep on-stack copies of these types. Now they can be registerized. [0]T is kind of trivial but might as well handle it. This change enables another change I'm working on to improve how x.(T) expressions are handled (#17405). This CL helps because now all types that are direct interface types are registerizeable (e.g. [1]*byte). No higher-degree arrays for now because non-constant indexes are hard. Update #17405 Change-Id: I2399940965d17b3969ae66f6fe447a8cefdd6edd Reviewed-on: https://go-review.googlesource.com/32416 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-10-30 21:10:03 -07:00
(IMake typ (ArrayMake1 val)) -> (IMake typ val)
cmd/compile: fix typing of IData opcodes The rules for extracting the interface data word don't leave the result typed correctly. If I do i.([1]*int)[0], the result should have type *int, not [1]*int. Using (IData x) for the result keeps the typing of the original top-level Value. I don't think this would ever cause a real codegen bug, bug fixing it at least makes the typing shown in ssa.html more consistent. Change-Id: I239d821c394e58347639387981b0510d13b2f7b7 Reviewed-on: https://go-review.googlesource.com/c/go/+/204042 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2019-10-29 13:14:58 -07:00
(ArraySelect [0] (IData x)) -> (IData x)
cmd/compile: make [0]T and [1]T SSAable types We used to have to keep on-stack copies of these types. Now they can be registerized. [0]T is kind of trivial but might as well handle it. This change enables another change I'm working on to improve how x.(T) expressions are handled (#17405). This CL helps because now all types that are direct interface types are registerizeable (e.g. [1]*byte). No higher-degree arrays for now because non-constant indexes are hard. Update #17405 Change-Id: I2399940965d17b3969ae66f6fe447a8cefdd6edd Reviewed-on: https://go-review.googlesource.com/32416 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-10-30 21:10:03 -07:00
[dev.ssa] cmd/internal/ssa: add string operations Add ops to load, store, select ptr & len, and build constant strings. A few other minor cleanups. Change-Id: I6f0f7419d641b119b613ed44561cd308a466051c Reviewed-on: https://go-review.googlesource.com/10449 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-05-27 14:52:22 -07:00
// string ops
cmd/compile: fix no-opt build after moving decomposing user functions decompose-builtin pass requires an opt pass, but -N disables late-opt, the only opt pass (out of two) that happens after decompose-builtin. This CL enables both 'opt' and 'late opt' passes. The extra compile time for 'late opt' in negligible since most rewrites were already done in the first 'opt' (also measured before). We should put some effort in splitting the generic rules into required and optional. Also update generic.rules comments about lowering of StringMake and SliceMake. Tested with GO_GCFLAGS=-N ./all.bash Change-Id: I92999681aaa02587b6dc6e32ce997a91f1fc9499 Reviewed-on: https://go-review.googlesource.com/20682 Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2016-03-14 19:11:19 +01:00
// 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.
cmd/compile: make []byte("...") more efficient Do []byte(string) conversions more efficiently when the string is a constant. Instead of calling stringtobyteslice, allocate just the space we need and encode the initialization directly. []byte("foo") rewrites to the following pseudocode: var s [3]byte // on heap or stack, depending on whether b escapes s = *(*[3]byte)(&"foo"[0]) // initialize s from the string b = s[:] which generates this assembly: 0x001d 00029 (tmp1.go:9) LEAQ type.[3]uint8(SB), AX 0x0024 00036 (tmp1.go:9) MOVQ AX, (SP) 0x0028 00040 (tmp1.go:9) CALL runtime.newobject(SB) 0x002d 00045 (tmp1.go:9) MOVQ 8(SP), AX 0x0032 00050 (tmp1.go:9) MOVBLZX go.string."foo"+2(SB), CX 0x0039 00057 (tmp1.go:9) MOVWLZX go.string."foo"(SB), DX 0x0040 00064 (tmp1.go:9) MOVW DX, (AX) 0x0043 00067 (tmp1.go:9) MOVB CL, 2(AX) // Then the slice is b = {AX, 3, 3} The generated code is still not optimal, as it still does load/store from read-only memory instead of constant stores. Next CL... Update #26498 Fixes #10170 Change-Id: I4b990b19f9a308f60c8f4f148934acffefe0a5bd Reviewed-on: https://go-review.googlesource.com/c/140698 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2018-10-08 17:46:45 -07:00
(StringPtr (StringMake (Addr <t> {s} base) _)) -> (Addr <t> {s} base)
cmd/compile/internal/ssa: lower builtins much later * Move lowering into a separate pass. * SliceLen/SliceCap is now available to various intermediate passes which use useful for bounds checking. * Add a second opt pass to handle the new opportunities Decreases the code size of binaries in pkg/tool/linux_amd64 by ~45K. Updates #14564 #14606 Change-Id: I5b2bd6202181c50623a3585fbf15c0d6db6d4685 Reviewed-on: https://go-review.googlesource.com/20172 Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-03-03 11:13:43 +01:00
(StringLen (StringMake _ (Const64 <t> [c]))) -> (Const64 <t> [c])
[dev.ssa] cmd/compile: some SSA optimizations Some optimizations of things I've seen looking at generated code. (x+y)-x == y x-0 == x The ptr portion of the constant string "" can be nil. Also update TODO with recent changes. Change-Id: I02c41ca2f9e9e178bf889058d3e083b446672dbe Reviewed-on: https://go-review.googlesource.com/16771 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-09 20:54:34 -08:00
(ConstString {s}) && config.PtrSize == 4 && s.(string) == "" ->
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]))
[dev.ssa] cmd/compile: some SSA optimizations Some optimizations of things I've seen looking at generated code. (x+y)-x == y x-0 == x The ptr portion of the constant string "" can be nil. Also update TODO with recent changes. Change-Id: I02c41ca2f9e9e178bf889058d3e083b446672dbe Reviewed-on: https://go-review.googlesource.com/16771 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-09 20:54:34 -08:00
(ConstString {s}) && config.PtrSize == 8 && s.(string) == "" ->
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]))
[dev.ssa] cmd/compile: some SSA optimizations Some optimizations of things I've seen looking at generated code. (x+y)-x == y x-0 == x The ptr portion of the constant string "" can be nil. Also update TODO with recent changes. Change-Id: I02c41ca2f9e9e178bf889058d3e083b446672dbe Reviewed-on: https://go-review.googlesource.com/16771 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-09 20:54:34 -08:00
(ConstString {s}) && config.PtrSize == 4 && s.(string) != "" ->
[dev.ssa] cmd/compile: Do pointer arithmetic with int, not uintptr Be more consistent about this. There's no reason to do the pointer arithmetic on a different type, as sizeof(int) >= sizeof(ptr) on all of our platforms. It simplifies our rewrite rules also, except for a few that need duplication. Add some more constant folding to get constant indexing and slicing to fold down to nothing. Change-Id: I3e56cdb14b3dc1a6a0514f0333e883f92c19e3c7 Reviewed-on: https://go-review.googlesource.com/16586 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-02 21:28:13 -08:00
(StringMake
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
(Addr <typ.BytePtr> {fe.StringData(s.(string))}
[dev.ssa] cmd/compile: Do pointer arithmetic with int, not uintptr Be more consistent about this. There's no reason to do the pointer arithmetic on a different type, as sizeof(int) >= sizeof(ptr) on all of our platforms. It simplifies our rewrite rules also, except for a few that need duplication. Add some more constant folding to get constant indexing and slicing to fold down to nothing. Change-Id: I3e56cdb14b3dc1a6a0514f0333e883f92c19e3c7 Reviewed-on: https://go-review.googlesource.com/16586 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-02 21:28:13 -08:00
(SB))
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> [int64(len(s.(string)))]))
[dev.ssa] cmd/compile: some SSA optimizations Some optimizations of things I've seen looking at generated code. (x+y)-x == y x-0 == x The ptr portion of the constant string "" can be nil. Also update TODO with recent changes. Change-Id: I02c41ca2f9e9e178bf889058d3e083b446672dbe Reviewed-on: https://go-review.googlesource.com/16771 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-09 20:54:34 -08:00
(ConstString {s}) && config.PtrSize == 8 && s.(string) != "" ->
[dev.ssa] cmd/compile: add decompose pass Decompose breaks compound objects up into pieces that can be operated on by the target architecture. The decompose pass only does phi ops, the rest is done by the rewrite rules in generic.rules. Compound objects include strings,slices,interfaces,structs,arrays. Arrays aren't decomposed because of indexing (we could support constant indexes, but dynamic indexes can't be handled using SSA). Structs will come in a subsequent CL. TODO: after this pass we have lost the association between, e.g., a string's pointer and its size. It would be nice if we could keep that information around for debugging info somehow. Change-Id: I6379ab962a7beef62297d0f68c421f22aa0a0901 Reviewed-on: https://go-review.googlesource.com/13683 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-08-18 10:26:28 -07:00
(StringMake
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
(Addr <typ.BytePtr> {fe.StringData(s.(string))}
[dev.ssa] cmd/compile/internal/ssa: allow ops to have a default type Specifying types in rewrites for all subexpressions gets verbose quickly. Allow opcodes to specify a default type which is used when none is supplied explicitly. Provide default types for a few easy opcodes. There are probably more we can do, but this is a good start. Change-Id: Iedc2a1a423cc3e2d4472640433982f9aa76a9f18 Reviewed-on: https://go-review.googlesource.com/14128 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-09-01 09:16:58 -07:00
(SB))
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> [int64(len(s.(string)))]))
[dev.ssa] cmd/compile: add decompose pass Decompose breaks compound objects up into pieces that can be operated on by the target architecture. The decompose pass only does phi ops, the rest is done by the rewrite rules in generic.rules. Compound objects include strings,slices,interfaces,structs,arrays. Arrays aren't decomposed because of indexing (we could support constant indexes, but dynamic indexes can't be handled using SSA). Structs will come in a subsequent CL. TODO: after this pass we have lost the association between, e.g., a string's pointer and its size. It would be nice if we could keep that information around for debugging info somehow. Change-Id: I6379ab962a7beef62297d0f68c421f22aa0a0901 Reviewed-on: https://go-review.googlesource.com/13683 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-08-18 10:26:28 -07:00
// 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.
cmd/compile/internal/ssa: lower builtins much later * Move lowering into a separate pass. * SliceLen/SliceCap is now available to various intermediate passes which use useful for bounds checking. * Add a second opt pass to handle the new opportunities Decreases the code size of binaries in pkg/tool/linux_amd64 by ~45K. Updates #14564 #14606 Change-Id: I5b2bd6202181c50623a3585fbf15c0d6db6d4685 Reviewed-on: https://go-review.googlesource.com/20172 Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-03-03 11:13:43 +01:00
(SliceLen (SliceMake _ (Const64 <t> [c]) _)) -> (Const64 <t> [c])
(SliceCap (SliceMake _ _ (Const64 <t> [c]))) -> (Const64 <t> [c])
cmd/compile: Repurpose old sliceopt.go for prove phase. Adapt old test for prove's bounds check elimination. Added missing rule to generic rules that lead to differences between 32 and 64 bit platforms on sliceopt test. Added debugging to prove.go that was helpful-to-necessary to discover that missing rule. Lowered debugging level on prove.go from 3 to 1; no idea why it was previously 3. Change-Id: I09de206aeb2fced9f2796efe2bfd4a59927eda0c Reviewed-on: https://go-review.googlesource.com/23290 Run-TryBot: David Chase <drchase@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-10-19 11:47:52 -04:00
(SliceLen (SliceMake _ (Const32 <t> [c]) _)) -> (Const32 <t> [c])
(SliceCap (SliceMake _ _ (Const32 <t> [c]))) -> (Const32 <t> [c])
cmd/compile/internal/ssa: generalize prove to all booleans * Refacts a bit saving and restoring parents restrictions * Shaves ~100k from pkg/tools/linux_amd64, but most of the savings come from the rewrite rules. * Improves on the following artificial test case: func f1(a4 bool, a6 bool) bool { return a6 || (a6 || (a6 || a4)) || (a6 || (a4 || a6 || (false || a6))) } Change-Id: I714000f75a37a3a6617c6e6834c75bd23674215f Reviewed-on: https://go-review.googlesource.com/20306 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-07 18:36:16 +01:00
(SlicePtr (SliceMake (SlicePtr x) _ _)) -> (SlicePtr x)
(SliceLen (SliceMake _ (SliceLen x) _)) -> (SliceLen x)
(SliceCap (SliceMake _ _ (SliceCap x))) -> (SliceCap x)
cmd/compile: simplify SliceCap when it equals SliceLen Shows up occassionally, especially after p = p[:8:len(p)] Updates #14905 Change-Id: Iab35ef2eac57817e6a10c6aaeeb84709e8021641 Reviewed-on: https://go-review.googlesource.com/21025 Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-22 17:34:36 +01:00
(SliceCap (SliceMake _ _ (SliceLen x))) -> (SliceLen x)
[dev.ssa] cmd/compile: Do pointer arithmetic with int, not uintptr Be more consistent about this. There's no reason to do the pointer arithmetic on a different type, as sizeof(int) >= sizeof(ptr) on all of our platforms. It simplifies our rewrite rules also, except for a few that need duplication. Add some more constant folding to get constant indexing and slicing to fold down to nothing. Change-Id: I3e56cdb14b3dc1a6a0514f0333e883f92c19e3c7 Reviewed-on: https://go-review.googlesource.com/16586 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-02 21:28:13 -08:00
(ConstSlice) && config.PtrSize == 4 ->
(SliceMake
cmd/compile/internal/types: remove ElemType wrapper This was an artifact from when we had a separate ssa.Type interface to break circular dependency between packages ssa and gc. It's no longer needed now that package ssa directly uses package types. Change-Id: I6a93e5d79082815f7f0eb89507381969cc6cb403 Reviewed-on: https://go-review.googlesource.com/109137 Run-TryBot: Matthew Dempsky <mdempsky@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-04-24 13:39:51 -07:00
(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]))
[dev.ssa] cmd/compile: Do pointer arithmetic with int, not uintptr Be more consistent about this. There's no reason to do the pointer arithmetic on a different type, as sizeof(int) >= sizeof(ptr) on all of our platforms. It simplifies our rewrite rules also, except for a few that need duplication. Add some more constant folding to get constant indexing and slicing to fold down to nothing. Change-Id: I3e56cdb14b3dc1a6a0514f0333e883f92c19e3c7 Reviewed-on: https://go-review.googlesource.com/16586 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-02 21:28:13 -08:00
(ConstSlice) && config.PtrSize == 8 ->
[dev.ssa] cmd/compile: add decompose pass Decompose breaks compound objects up into pieces that can be operated on by the target architecture. The decompose pass only does phi ops, the rest is done by the rewrite rules in generic.rules. Compound objects include strings,slices,interfaces,structs,arrays. Arrays aren't decomposed because of indexing (we could support constant indexes, but dynamic indexes can't be handled using SSA). Structs will come in a subsequent CL. TODO: after this pass we have lost the association between, e.g., a string's pointer and its size. It would be nice if we could keep that information around for debugging info somehow. Change-Id: I6379ab962a7beef62297d0f68c421f22aa0a0901 Reviewed-on: https://go-review.googlesource.com/13683 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-08-18 10:26:28 -07:00
(SliceMake
cmd/compile/internal/types: remove ElemType wrapper This was an artifact from when we had a separate ssa.Type interface to break circular dependency between packages ssa and gc. It's no longer needed now that package ssa directly uses package types. Change-Id: I6a93e5d79082815f7f0eb89507381969cc6cb403 Reviewed-on: https://go-review.googlesource.com/109137 Run-TryBot: Matthew Dempsky <mdempsky@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-04-24 13:39:51 -07:00
(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]))
[dev.ssa] cmd/compile: add decompose pass Decompose breaks compound objects up into pieces that can be operated on by the target architecture. The decompose pass only does phi ops, the rest is done by the rewrite rules in generic.rules. Compound objects include strings,slices,interfaces,structs,arrays. Arrays aren't decomposed because of indexing (we could support constant indexes, but dynamic indexes can't be handled using SSA). Structs will come in a subsequent CL. TODO: after this pass we have lost the association between, e.g., a string's pointer and its size. It would be nice if we could keep that information around for debugging info somehow. Change-Id: I6379ab962a7beef62297d0f68c421f22aa0a0901 Reviewed-on: https://go-review.googlesource.com/13683 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-08-18 10:26:28 -07:00
// interface ops
(ConstInterface) ->
(IMake
cmd/compile: mark the first word of an interface as a uintptr The first word of an interface is a pointer, but for the purposes of GC we don't need to treat it as such. 1. If it is a non-empty interface, the pointer points to an itab which is always in persistentalloc space. 2. If it is an empty interface, the pointer points to a _type. a. If it is a compile-time-allocated type, it points into the read-only data section. b. If it is a reflect-allocated type, it points into the Go heap. Reflect is responsible for keeping a reference to the underlying type so it won't be GCd. If we ever have a moving GC, we need to change this for 2b (as well as scan itabs to update their itab._type fields). Write barriers on the first word of interfaces have already been removed. Change-Id: I643e91d7ac4de980ac2717436eff94097c65d959 Reviewed-on: https://go-review.googlesource.com/97518 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2018-02-27 13:46:03 -08:00
(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>))
[dev.ssa] cmd/compile/internal/ssa: autogenerate opcodes Revamp autogeneration. Get rid of gogenerate commands, they are more trouble than they are worth. (If the code won't compile, gogenerate doesn't work.) Generate opcode enums & tables. This means we only have to specify opcodes in one place instead of two. Add arch prefixes to opcodes so they will be globally unique. Change-Id: I175d0a89b701b2377bbe699f3756731b7c9f5a9f Reviewed-on: https://go-review.googlesource.com/10812 Reviewed-by: Alan Donovan <adonovan@google.com>
2015-06-06 16:03:33 -07:00
cmd/compile: redo nil checks Get rid of BlockCheck. Josh goaded me into it, and I went down a rabbithole making it happen. NilCheck now panics if the pointer is nil and returns void, as before. BlockCheck is gone, and NilCheck is no longer a Control value for any block. It just exists (and deadcode knows not to throw it away). I rewrote the nilcheckelim pass to handle this case. In particular, there can now be multiple NilCheck ops per block. I moved all of the arch-dependent nil check elimination done as part of ssaGenValue into its own proper pass, so we don't have to duplicate that code for every architecture. Making the arch-dependent nil check its own pass means I needed to add a bunch of flags to the opcode table so I could write the code without arch-dependent ops everywhere. Change-Id: I419f891ac9b0de313033ff09115c374163416a9f Reviewed-on: https://go-review.googlesource.com/29120 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-09-13 17:01:01 -07:00
(NilCheck (GetG mem) mem) -> mem
[dev.ssa] cmd/compile: implement OGETG Change-Id: I7ecf62cf399c710b4a617803c43e83fce09b8a7d Reviewed-on: https://go-review.googlesource.com/13585 Reviewed-by: Keith Randall <khr@golang.org>
2015-08-12 11:22:16 -07:00
[dev.ssa] cmd/compile: rewrite if not Rewrite if !cond by swapping the branches and removing the not. Change-Id: If3af1bac02bfc566faba872a8c7f7e5ce38e9f58 Reviewed-on: https://go-review.googlesource.com/12610 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-07-23 18:44:09 -05:00
(If (Not cond) yes no) -> (If cond no yes)
cmd/compile: allow multiple SSA block control values Control values are used to choose which successor of a block is jumped to. Typically a control value takes the form of a 'flags' value that represents the result of a comparison. Some architectures however use a variable in a register as a control value. Up until now we have managed with a single control value per block. However some architectures (e.g. s390x and riscv64) have combined compare-and-branch instructions that take two variables in registers as parameters. To generate these instructions we need to support 2 control values per block. This CL allows up to 2 control values to be used in a block in order to support the addition of compare-and-branch instructions. I have implemented s390x compare-and-branch instructions in a different CL. Passes toolstash-check -all. Results of compilebench: name old time/op new time/op delta Template 208ms ± 1% 209ms ± 1% ~ (p=0.289 n=20+20) Unicode 83.7ms ± 1% 83.3ms ± 3% -0.49% (p=0.017 n=18+18) GoTypes 748ms ± 1% 748ms ± 0% ~ (p=0.460 n=20+18) Compiler 3.47s ± 1% 3.48s ± 1% ~ (p=0.070 n=19+18) SSA 11.5s ± 1% 11.7s ± 1% +1.64% (p=0.000 n=19+18) Flate 130ms ± 1% 130ms ± 1% ~ (p=0.588 n=19+20) GoParser 160ms ± 1% 161ms ± 1% ~ (p=0.211 n=20+20) Reflect 465ms ± 1% 467ms ± 1% +0.42% (p=0.007 n=20+20) Tar 184ms ± 1% 185ms ± 2% ~ (p=0.087 n=18+20) XML 253ms ± 1% 253ms ± 1% ~ (p=0.377 n=20+18) LinkCompiler 769ms ± 2% 774ms ± 2% ~ (p=0.070 n=19+19) ExternalLinkCompiler 3.59s ±11% 3.68s ± 6% ~ (p=0.072 n=20+20) LinkWithoutDebugCompiler 446ms ± 5% 454ms ± 3% +1.79% (p=0.002 n=19+20) StdCmd 26.0s ± 2% 26.0s ± 2% ~ (p=0.799 n=20+20) name old user-time/op new user-time/op delta Template 238ms ± 5% 240ms ± 5% ~ (p=0.142 n=20+20) Unicode 105ms ±11% 106ms ±10% ~ (p=0.512 n=20+20) GoTypes 876ms ± 2% 873ms ± 4% ~ (p=0.647 n=20+19) Compiler 4.17s ± 2% 4.19s ± 1% ~ (p=0.093 n=20+18) SSA 13.9s ± 1% 14.1s ± 1% +1.45% (p=0.000 n=18+18) Flate 145ms ±13% 146ms ± 5% ~ (p=0.851 n=20+18) GoParser 185ms ± 5% 188ms ± 7% ~ (p=0.174 n=20+20) Reflect 534ms ± 3% 538ms ± 2% ~ (p=0.105 n=20+18) Tar 215ms ± 4% 211ms ± 9% ~ (p=0.079 n=19+20) XML 295ms ± 6% 295ms ± 5% ~ (p=0.968 n=20+20) LinkCompiler 832ms ± 4% 837ms ± 7% ~ (p=0.707 n=17+20) ExternalLinkCompiler 1.58s ± 8% 1.60s ± 4% ~ (p=0.296 n=20+19) LinkWithoutDebugCompiler 478ms ±12% 489ms ±10% ~ (p=0.429 n=20+20) name old object-bytes new object-bytes delta Template 559kB ± 0% 559kB ± 0% ~ (all equal) Unicode 216kB ± 0% 216kB ± 0% ~ (all equal) GoTypes 2.03MB ± 0% 2.03MB ± 0% ~ (all equal) Compiler 8.07MB ± 0% 8.07MB ± 0% -0.06% (p=0.000 n=20+20) SSA 27.1MB ± 0% 27.3MB ± 0% +0.89% (p=0.000 n=20+20) Flate 343kB ± 0% 343kB ± 0% ~ (all equal) GoParser 441kB ± 0% 441kB ± 0% ~ (all equal) Reflect 1.36MB ± 0% 1.36MB ± 0% ~ (all equal) Tar 487kB ± 0% 487kB ± 0% ~ (all equal) XML 632kB ± 0% 632kB ± 0% ~ (all equal) name old export-bytes new export-bytes delta Template 18.5kB ± 0% 18.5kB ± 0% ~ (all equal) Unicode 7.92kB ± 0% 7.92kB ± 0% ~ (all equal) GoTypes 35.0kB ± 0% 35.0kB ± 0% ~ (all equal) Compiler 109kB ± 0% 110kB ± 0% +0.72% (p=0.000 n=20+20) SSA 137kB ± 0% 138kB ± 0% +0.58% (p=0.000 n=20+20) Flate 4.89kB ± 0% 4.89kB ± 0% ~ (all equal) GoParser 8.49kB ± 0% 8.49kB ± 0% ~ (all equal) Reflect 11.4kB ± 0% 11.4kB ± 0% ~ (all equal) Tar 10.5kB ± 0% 10.5kB ± 0% ~ (all equal) XML 16.7kB ± 0% 16.7kB ± 0% ~ (all equal) name old text-bytes new text-bytes delta HelloSize 761kB ± 0% 761kB ± 0% ~ (all equal) CmdGoSize 10.8MB ± 0% 10.8MB ± 0% ~ (all equal) name old data-bytes new data-bytes delta HelloSize 10.7kB ± 0% 10.7kB ± 0% ~ (all equal) CmdGoSize 312kB ± 0% 312kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 122kB ± 0% 122kB ± 0% ~ (all equal) CmdGoSize 146kB ± 0% 146kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.13MB ± 0% 1.13MB ± 0% ~ (all equal) CmdGoSize 15.1MB ± 0% 15.1MB ± 0% ~ (all equal) Change-Id: I3cc2f9829a109543d9a68be4a21775d2d3e9801f Reviewed-on: https://go-review.googlesource.com/c/go/+/196557 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Daniel Martí <mvdan@mvdan.cc> Reviewed-by: Keith Randall <khr@golang.org>
2019-08-12 20:19:58 +01:00
(If (ConstBool [c]) yes no) && c == 1 -> (First yes no)
(If (ConstBool [c]) yes no) && c == 0 -> (First no yes)
[dev.ssa] cmd/compile: get rid of converts in unsafe.Pointer arithmetic unsafe.Pointer->uintptr, add, then uintptr->unsafe.Pointer. Do the add directly on the pointer type instead. Change-Id: I5a3a32691d0a000e16975857974ed9a1039c6d28 Reviewed-on: https://go-review.googlesource.com/16281 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-10-23 14:08:50 -07:00
// Get rid of Convert ops for pointer arithmetic on unsafe.Pointer.
cmd/compile: remove unnecessary convert ops on 32-bit archs Missing rule to fold out Convert ops on 32-bit architectures. This comes up with expressions like unsafe.Pointer(uintptr(p) + x) Change-Id: I429e968e5d1a3e13a386ddd29a08ebb6d7dd938a Reviewed-on: https://go-review.googlesource.com/112158 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> Reviewed-by: David Chase <drchase@google.com>
2018-05-08 08:53:21 -07:00
(Convert (Add(64|32) (Convert ptr mem) off) mem) -> (Add(64|32) ptr off)
[dev.ssa] cmd/compile: be safer about uintptr/unsafe.Pointer conversions Make sure that when a pointer value is live across a function call, we save it as a pointer. (And similarly a uintptr live across a function call should not be saved as a pointer.) Add a nasty test case. This is probably what is preventing the merge from master to dev.ssa. Signs point to something like this bug happening in mallocgc. Change-Id: Ib23fa1251b8d1c50d82c6a448cb4a4fc28219029 Reviewed-on: https://go-review.googlesource.com/16830 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-10 15:35:36 -08:00
(Convert (Convert ptr mem) mem) -> ptr
[dev.ssa] cmd/compile: start arguments as spilled Declare a function's arguments as having already been spilled so their use just requires a restore. Allow spill locations to be portions of larger objects the stack. Required to load portions of compound input arguments. Rename the memory input to InputMem. Use Arg for the pre-spilled argument values. Change-Id: I8fe2a03ffbba1022d98bfae2052b376b96d32dda Reviewed-on: https://go-review.googlesource.com/16536 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2015-11-02 08:10:26 -08:00
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
// strength reduction of divide by a constant.
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// See ../magic.go for a detailed description of these algorithms.
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// Unsigned divide by power of 2. Strength reduce to a shift.
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
(Div8u n (Const8 [c])) && isPowerOfTwo(c&0xff) -> (Rsh8Ux64 n (Const64 <typ.UInt64> [log2(c&0xff)]))
(Div16u n (Const16 [c])) && isPowerOfTwo(c&0xffff) -> (Rsh16Ux64 n (Const64 <typ.UInt64> [log2(c&0xffff)]))
(Div32u n (Const32 [c])) && isPowerOfTwo(c&0xffffffff) -> (Rsh32Ux64 n (Const64 <typ.UInt64> [log2(c&0xffffffff)]))
(Div64u n (Const64 [c])) && isPowerOfTwo(c) -> (Rsh64Ux64 n (Const64 <typ.UInt64> [log2(c)]))
cmd/compile: add rules handling unsigned div/mod by constant 1<<63 Normally 64-bit div/mod is turned into runtime calls on 32-bit arch, but the front end leaves power-of-two constant division and hopes the SSA backend turns into a shift or AND. The SSA rule is (Mod64u <t> n (Const64 [c])) && isPowerOfTwo(c) -> (And64 n (Const64 <t> [c-1])) But isPowerOfTwo returns true only for positive int64, which leaves out 1<<63 unhandled. Add a special case for 1<<63. Fixes #21517. Change-Id: I02d27dc7177d4af0ee8d7f5533714edecddf8c95 Reviewed-on: https://go-review.googlesource.com/56890 Reviewed-by: Keith Randall <khr@golang.org>
2017-08-18 08:37:58 -04:00
(Div64u n (Const64 [-1<<63])) -> (Rsh64Ux64 n (Const64 <typ.UInt64> [63]))
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
cmd/compile: optimize signed non-negative div/mod by a power of 2 This CL optimizes assembly for len() or cap() division by a power of 2 constants: func lenDiv(s []int) int { return len(s) / 16 } amd64 assembly before the CL: MOVQ "".s+16(SP), AX MOVQ AX, CX SARQ $63, AX SHRQ $60, AX ADDQ CX, AX SARQ $4, AX MOVQ AX, "".~r1+32(SP) RET amd64 assembly after the CL: MOVQ "".s+16(SP), AX SHRQ $4, AX MOVQ AX, "".~r1+32(SP) RET The CL relies on the fact that len() and cap() result cannot be negative. Trigger stats for the added SSA rules on linux/amd64 when running make.bash: 46 Div64 12 Mod64 The added SSA rules may trigger on more cases in the future when SSA values will be populated with the info on their lower bounds. For instance: func f(i int16) int16 { if i < 3 { return -1 } // Lower bound of i is 3 here -> i is non-negative, // so unsigned arithmetics may be used here. return i % 16 } Change-Id: I8bc6be5a03e71157ced533c01416451ff6f1a7f0 Reviewed-on: https://go-review.googlesource.com/65530 Reviewed-by: Keith Randall <khr@golang.org>
2017-09-23 00:34:37 +03:00
// Signed non-negative divide by power of 2.
(Div8 n (Const8 [c])) && isNonNegative(n) && isPowerOfTwo(c&0xff) -> (Rsh8Ux64 n (Const64 <typ.UInt64> [log2(c&0xff)]))
(Div16 n (Const16 [c])) && isNonNegative(n) && isPowerOfTwo(c&0xffff) -> (Rsh16Ux64 n (Const64 <typ.UInt64> [log2(c&0xffff)]))
(Div32 n (Const32 [c])) && isNonNegative(n) && isPowerOfTwo(c&0xffffffff) -> (Rsh32Ux64 n (Const64 <typ.UInt64> [log2(c&0xffffffff)]))
(Div64 n (Const64 [c])) && isNonNegative(n) && isPowerOfTwo(c) -> (Rsh64Ux64 n (Const64 <typ.UInt64> [log2(c)]))
(Div64 n (Const64 [-1<<63])) && isNonNegative(n) -> (Const64 [0])
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
// Unsigned divide, not a power of 2. Strength reduce to a multiply.
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// For 8-bit divides, we just do a direct 9-bit by 8-bit multiply.
(Div8u x (Const8 [c])) && umagicOK(8, 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> [int64(1<<8+umagic(8,c).m)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(ZeroExt8to32 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> [8+umagic(8,c).s])))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// For 16-bit divides on 64-bit machines, we do a direct 17-bit by 16-bit multiply.
(Div16u x (Const16 [c])) && umagicOK(16, 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+umagic(16,c).m)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(ZeroExt16to64 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> [16+umagic(16,c).s])))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// For 16-bit divides on 32-bit machines
(Div16u x (Const16 [c])) && umagicOK(16, c) && config.RegSize == 4 && umagic(16,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> [int64(1<<15+umagic(16,c).m/2)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(ZeroExt16to32 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> [16+umagic(16,c).s-1])))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Div16u x (Const16 [c])) && umagicOK(16, 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> [int64(1<<15+(umagic(16,c).m+1)/2)])
(Rsh32Ux64 <typ.UInt32> (ZeroExt16to32 x) (Const64 <typ.UInt64> [1])))
(Const64 <typ.UInt64> [16+umagic(16,c).s-2])))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div16u x (Const16 [c])) && umagicOK(16, c) && config.RegSize == 4 && config.useAvg ->
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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>
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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> [int64(umagic(16,c).m)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(ZeroExt16to32 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> [16+umagic(16,c).s-1])))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// For 32-bit divides on 32-bit machines
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div32u x (Const32 [c])) && umagicOK(32, c) && config.RegSize == 4 && umagic(32,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> [int64(int32(1<<31+umagic(32,c).m/2))])
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
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> [umagic(32,c).s-1]))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div32u x (Const32 [c])) && umagicOK(32, 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> [int64(int32(1<<31+(umagic(32,c).m+1)/2))])
(Rsh32Ux64 <typ.UInt32> x (Const64 <typ.UInt64> [1])))
(Const64 <typ.UInt64> [umagic(32,c).s-2]))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div32u x (Const32 [c])) && umagicOK(32, 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>
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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> [int64(int32(umagic(32,c).m))])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
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> [umagic(32,c).s-1]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// For 32-bit divides on 64-bit machines
// We'll use a regular (non-hi) multiply for this case.
(Div32u x (Const32 [c])) && umagicOK(32, c) && config.RegSize == 8 && umagic(32,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+umagic(32,c).m/2)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(ZeroExt32to64 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> [32+umagic(32,c).s-1])))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Div32u x (Const32 [c])) && umagicOK(32, 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+(umagic(32,c).m+1)/2)])
(Rsh64Ux64 <typ.UInt64> (ZeroExt32to64 x) (Const64 <typ.UInt64> [1])))
(Const64 <typ.UInt64> [32+umagic(32,c).s-2])))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div32u x (Const32 [c])) && umagicOK(32, c) && config.RegSize == 8 && config.useAvg ->
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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>
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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(umagic(32,c).m)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(ZeroExt32to64 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> [32+umagic(32,c).s-1])))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// 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.)
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div64u x (Const64 [c])) && umagicOK(64, c) && config.RegSize == 8 && umagic(64,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+umagic(64,c).m/2)])
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
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> [umagic(64,c).s-1]))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div64u x (Const64 [c])) && umagicOK(64, 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+(umagic(64,c).m+1)/2)])
(Rsh64Ux64 <typ.UInt64> x (Const64 <typ.UInt64> [1])))
(Const64 <typ.UInt64> [umagic(64,c).s-2]))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div64u x (Const64 [c])) && umagicOK(64, 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>
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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(umagic(64,c).m)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
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> [umagic(64,c).s-1]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// 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.
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
(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]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// 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
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
(Add8 <t> n (Rsh8Ux64 <t> (Rsh8x64 <t> n (Const64 <typ.UInt64> [ 7])) (Const64 <typ.UInt64> [ 8-log2(c)])))
(Const64 <typ.UInt64> [log2(c)]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Div16 <t> n (Const16 [c])) && isPowerOfTwo(c) ->
(Rsh16x64
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
(Add16 <t> n (Rsh16Ux64 <t> (Rsh16x64 <t> n (Const64 <typ.UInt64> [15])) (Const64 <typ.UInt64> [16-log2(c)])))
(Const64 <typ.UInt64> [log2(c)]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Div32 <t> n (Const32 [c])) && isPowerOfTwo(c) ->
(Rsh32x64
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
(Add32 <t> n (Rsh32Ux64 <t> (Rsh32x64 <t> n (Const64 <typ.UInt64> [31])) (Const64 <typ.UInt64> [32-log2(c)])))
(Const64 <typ.UInt64> [log2(c)]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Div64 <t> n (Const64 [c])) && isPowerOfTwo(c) ->
(Rsh64x64
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
(Add64 <t> n (Rsh64Ux64 <t> (Rsh64x64 <t> n (Const64 <typ.UInt64> [63])) (Const64 <typ.UInt64> [64-log2(c)])))
(Const64 <typ.UInt64> [log2(c)]))
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
// Signed divide, not a power of 2. Strength reduce to a multiply.
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Div8 <t> x (Const8 [c])) && smagicOK(8,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> [int64(smagic(8,c).m)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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> [8+smagic(8,c).s]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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])))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Div16 <t> x (Const16 [c])) && smagicOK(16,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> [int64(smagic(16,c).m)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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> [16+smagic(16,c).s]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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])))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Div32 <t> x (Const32 [c])) && smagicOK(32,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(smagic(32,c).m)])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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> [32+smagic(32,c).s]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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])))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div32 <t> x (Const32 [c])) && smagicOK(32,c) && config.RegSize == 4 && smagic(32,c).m&1 == 0 && config.useHmul ->
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Sub32 <t>
(Rsh32x64 <t>
(Hmul32 <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
(Const32 <typ.UInt32> [int64(int32(smagic(32,c).m/2))])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
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> [smagic(32,c).s-1]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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])))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div32 <t> x (Const32 [c])) && smagicOK(32,c) && config.RegSize == 4 && smagic(32,c).m&1 != 0 && config.useHmul ->
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(Sub32 <t>
(Rsh32x64 <t>
(Add32 <t>
(Hmul32 <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
(Const32 <typ.UInt32> [int64(int32(smagic(32,c).m))])
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
x)
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> [smagic(32,c).s]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
(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])))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div64 <t> x (Const64 [c])) && smagicOK(64,c) && smagic(64,c).m&1 == 0 && config.useHmul ->
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
(Sub64 <t>
(Rsh64x64 <t>
(Hmul64 <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
(Const64 <typ.UInt64> [int64(smagic(64,c).m/2)])
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
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> [smagic(64,c).s-1]))
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
(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])))
cmd/compile: add SSA config options useAvg and useHmul This commit allows architectures to disable optimizations that need the Avg* and Hmul* operations. WebAssembly has no such operations, so using them as an optimization but then having to emulate them with multiple instructions makes no sense, especially since the WebAssembly compiler may do the same optimizations internally. Updates #18892 Change-Id: If57b59e3235482a9e0ec334a7312b3e3b5fc2b61 Reviewed-on: https://go-review.googlesource.com/103256 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2018-03-09 00:06:33 +01:00
(Div64 <t> x (Const64 [c])) && smagicOK(64,c) && smagic(64,c).m&1 != 0 && config.useHmul ->
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
(Sub64 <t>
(Rsh64x64 <t>
(Add64 <t>
(Hmul64 <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
(Const64 <typ.UInt64> [int64(smagic(64,c).m)])
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
x)
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> [smagic(64,c).s]))
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
(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])))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// Unsigned mod by power of 2 constant.
(Mod8u <t> n (Const8 [c])) && isPowerOfTwo(c&0xff) -> (And8 n (Const8 <t> [(c&0xff)-1]))
(Mod16u <t> n (Const16 [c])) && isPowerOfTwo(c&0xffff) -> (And16 n (Const16 <t> [(c&0xffff)-1]))
(Mod32u <t> n (Const32 [c])) && isPowerOfTwo(c&0xffffffff) -> (And32 n (Const32 <t> [(c&0xffffffff)-1]))
(Mod64u <t> n (Const64 [c])) && isPowerOfTwo(c) -> (And64 n (Const64 <t> [c-1]))
cmd/compile: add rules handling unsigned div/mod by constant 1<<63 Normally 64-bit div/mod is turned into runtime calls on 32-bit arch, but the front end leaves power-of-two constant division and hopes the SSA backend turns into a shift or AND. The SSA rule is (Mod64u <t> n (Const64 [c])) && isPowerOfTwo(c) -> (And64 n (Const64 <t> [c-1])) But isPowerOfTwo returns true only for positive int64, which leaves out 1<<63 unhandled. Add a special case for 1<<63. Fixes #21517. Change-Id: I02d27dc7177d4af0ee8d7f5533714edecddf8c95 Reviewed-on: https://go-review.googlesource.com/56890 Reviewed-by: Keith Randall <khr@golang.org>
2017-08-18 08:37:58 -04:00
(Mod64u <t> n (Const64 [-1<<63])) -> (And64 n (Const64 <t> [1<<63-1]))
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
cmd/compile: optimize signed non-negative div/mod by a power of 2 This CL optimizes assembly for len() or cap() division by a power of 2 constants: func lenDiv(s []int) int { return len(s) / 16 } amd64 assembly before the CL: MOVQ "".s+16(SP), AX MOVQ AX, CX SARQ $63, AX SHRQ $60, AX ADDQ CX, AX SARQ $4, AX MOVQ AX, "".~r1+32(SP) RET amd64 assembly after the CL: MOVQ "".s+16(SP), AX SHRQ $4, AX MOVQ AX, "".~r1+32(SP) RET The CL relies on the fact that len() and cap() result cannot be negative. Trigger stats for the added SSA rules on linux/amd64 when running make.bash: 46 Div64 12 Mod64 The added SSA rules may trigger on more cases in the future when SSA values will be populated with the info on their lower bounds. For instance: func f(i int16) int16 { if i < 3 { return -1 } // Lower bound of i is 3 here -> i is non-negative, // so unsigned arithmetics may be used here. return i % 16 } Change-Id: I8bc6be5a03e71157ced533c01416451ff6f1a7f0 Reviewed-on: https://go-review.googlesource.com/65530 Reviewed-by: Keith Randall <khr@golang.org>
2017-09-23 00:34:37 +03:00
// Signed non-negative mod by power of 2 constant.
(Mod8 <t> n (Const8 [c])) && isNonNegative(n) && isPowerOfTwo(c&0xff) -> (And8 n (Const8 <t> [(c&0xff)-1]))
(Mod16 <t> n (Const16 [c])) && isNonNegative(n) && isPowerOfTwo(c&0xffff) -> (And16 n (Const16 <t> [(c&0xffff)-1]))
(Mod32 <t> n (Const32 [c])) && isNonNegative(n) && isPowerOfTwo(c&0xffffffff) -> (And32 n (Const32 <t> [(c&0xffffffff)-1]))
(Mod64 <t> n (Const64 [c])) && isNonNegative(n) && isPowerOfTwo(c) -> (And64 n (Const64 <t> [c-1]))
(Mod64 n (Const64 [-1<<63])) && isNonNegative(n) -> n
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// 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]))
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
// All other mods by constants, do A%B = A-(A/B*B).
[dev.ssa] cmd/compile: strength-reduce 64-bit constant divides The frontend does this for 32 bits and below, but SSA needs to do it for 64 bits. The algorithms are all copied from cgen.go:cgen_div. Speeds up TimeFormat substantially: ~40% slower to ~10% slower. Change-Id: I023ea2eb6040df98ccd9105e15ca6ea695610a7a Reviewed-on: https://go-review.googlesource.com/19302 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Todd Neal <todd@tneal.org>
2016-02-05 20:26:18 -08:00
// This implements % with two * and a bunch of ancillary ops.
// One of the * is free if the user's code also computes A/B.
Revert "cmd/compile: add signed divisibility by power of 2 rules" This reverts CL 168038 (git 68819fb6d2bab59e4eadcdf62aa4a2a54417d640) Reason for revert: Doesn't work on 32 bit archs. Change-Id: Idec9098060dc65bc2f774c5383f0477f8eb63a3d Reviewed-on: https://go-review.googlesource.com/c/go/+/173442 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2019-04-23 20:49:05 +00:00
(Mod8 <t> x (Const8 [c])) && x.Op != OpConst8 && (c > 0 || c == -1<<7)
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
-> (Sub8 x (Mul8 <t> (Div8 <t> x (Const8 <t> [c])) (Const8 <t> [c])))
Revert "cmd/compile: add signed divisibility by power of 2 rules" This reverts CL 168038 (git 68819fb6d2bab59e4eadcdf62aa4a2a54417d640) Reason for revert: Doesn't work on 32 bit archs. Change-Id: Idec9098060dc65bc2f774c5383f0477f8eb63a3d Reviewed-on: https://go-review.googlesource.com/c/go/+/173442 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2019-04-23 20:49:05 +00:00
(Mod16 <t> x (Const16 [c])) && x.Op != OpConst16 && (c > 0 || c == -1<<15)
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
-> (Sub16 x (Mul16 <t> (Div16 <t> x (Const16 <t> [c])) (Const16 <t> [c])))
Revert "cmd/compile: add signed divisibility by power of 2 rules" This reverts CL 168038 (git 68819fb6d2bab59e4eadcdf62aa4a2a54417d640) Reason for revert: Doesn't work on 32 bit archs. Change-Id: Idec9098060dc65bc2f774c5383f0477f8eb63a3d Reviewed-on: https://go-review.googlesource.com/c/go/+/173442 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2019-04-23 20:49:05 +00:00
(Mod32 <t> x (Const32 [c])) && x.Op != OpConst32 && (c > 0 || c == -1<<31)
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
-> (Sub32 x (Mul32 <t> (Div32 <t> x (Const32 <t> [c])) (Const32 <t> [c])))
Revert "cmd/compile: add signed divisibility by power of 2 rules" This reverts CL 168038 (git 68819fb6d2bab59e4eadcdf62aa4a2a54417d640) Reason for revert: Doesn't work on 32 bit archs. Change-Id: Idec9098060dc65bc2f774c5383f0477f8eb63a3d Reviewed-on: https://go-review.googlesource.com/c/go/+/173442 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2019-04-23 20:49:05 +00:00
(Mod64 <t> x (Const64 [c])) && x.Op != OpConst64 && (c > 0 || c == -1<<63)
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
-> (Sub64 x (Mul64 <t> (Div64 <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
(Mod8u <t> x (Const8 [c])) && x.Op != OpConst8 && c > 0 && umagicOK(8, c)
cmd/compile: move constant divide strength reduction to SSA rules Currently the conversion from constant divides to multiplies is mostly done during the walk pass. This is suboptimal because SSA can determine that the value being divided by is constant more often (e.g. after inlining). Change-Id: If1a9b993edd71be37396b9167f77da271966f85f Reviewed-on: https://go-review.googlesource.com/37015 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2017-02-13 16:00:09 -08:00
-> (Sub8 x (Mul8 <t> (Div8u <t> x (Const8 <t> [c])) (Const8 <t> [c])))
(Mod16u <t> x (Const16 [c])) && x.Op != OpConst16 && c > 0 && umagicOK(16,c)
-> (Sub16 x (Mul16 <t> (Div16u <t> x (Const16 <t> [c])) (Const16 <t> [c])))
(Mod32u <t> x (Const32 [c])) && x.Op != OpConst32 && c > 0 && umagicOK(32,c)
-> (Sub32 x (Mul32 <t> (Div32u <t> x (Const32 <t> [c])) (Const32 <t> [c])))
(Mod64u <t> x (Const64 [c])) && x.Op != OpConst64 && c > 0 && umagicOK(64,c)
cmd/compile: clean up rewrite rules Break really long lines. Add spacing to line up columns. In AMD64, put all the optimization rules after all the lowering rules. Change-Id: I45cc7368bf278416e67f89e74358db1bd4326a93 Reviewed-on: https://go-review.googlesource.com/22470 Reviewed-by: David Chase <drchase@google.com>
2016-04-26 12:08:31 -07:00
-> (Sub64 x (Mul64 <t> (Div64u <t> x (Const64 <t> [c])) (Const64 <t> [c])))
[dev.ssa] cmd/compile: add floating point optimizations in SSA for ARM Add some simplification rules for floating point ops. cmd/internal/obj/arm supports instructions that compare FP register to 0, but runtime softfloat simulator does not. This CL adds these instructions to softfloat simulator as well. Updates #15365. Change-Id: I29405b2bfcb4c8cf106cb7a1a811409fec91b170 Reviewed-on: https://go-review.googlesource.com/24790 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-07-06 10:04:45 -04:00
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 && udivisibleOK(8,c) && !hasSmallRotate(config) ->
(Eq32 (Mod32u <typ.UInt32> (ZeroExt8to32 <typ.UInt32> x) (Const32 <typ.UInt32> [c&0xff])) (Const32 <typ.UInt32> [0]))
(Eq16 (Mod16u x (Const16 [c])) (Const16 [0])) && x.Op != OpConst16 && udivisibleOK(16,c) && !hasSmallRotate(config) ->
(Eq32 (Mod32u <typ.UInt32> (ZeroExt16to32 <typ.UInt32> x) (Const32 <typ.UInt32> [c&0xffff])) (Const32 <typ.UInt32> [0]))
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
(Eq8 (Mod8 x (Const8 [c])) (Const8 [0])) && x.Op != OpConst8 && sdivisibleOK(8,c) && !hasSmallRotate(config) ->
(Eq32 (Mod32 <typ.Int32> (SignExt8to32 <typ.Int32> x) (Const32 <typ.Int32> [c])) (Const32 <typ.Int32> [0]))
(Eq16 (Mod16 x (Const16 [c])) (Const16 [0])) && x.Op != OpConst16 && sdivisibleOK(16,c) && !hasSmallRotate(config) ->
(Eq32 (Mod32 <typ.Int32> (SignExt16to32 <typ.Int32> x) (Const32 <typ.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
all: fix a bunch of misspellings Change-Id: I5b909df0fd048cd66c5a27fca1b06466d3bcaac7 GitHub-Last-Rev: 778c5d21311abee09a5fbda2e4005a5fd4cc3f9f GitHub-Pull-Request: golang/go#35624 Reviewed-on: https://go-review.googlesource.com/c/go/+/207421 Run-TryBot: Ian Lance Taylor <iant@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2019-11-15 19:49:30 +00:00
// 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 == int64(1<<8+umagic(8,c).m) && s == 8+umagic(8,c).s
&& x.Op != OpConst8 && udivisibleOK(8,c)
-> (Leq8U
(RotateLeft8 <typ.UInt8>
(Mul8 <typ.UInt8>
(Const8 <typ.UInt8> [int64(int8(udivisible(8,c).m))])
x)
(Const8 <typ.UInt8> [int64(8-udivisible(8,c).k)])
)
(Const8 <typ.UInt8> [int64(int8(udivisible(8,c).max))])
)
(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+umagic(16,c).m) && s == 16+umagic(16,c).s
&& x.Op != OpConst16 && udivisibleOK(16,c)
-> (Leq16U
(RotateLeft16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int64(int16(udivisible(16,c).m))])
x)
(Const16 <typ.UInt16> [int64(16-udivisible(16,c).k)])
)
(Const16 <typ.UInt16> [int64(int16(udivisible(16,c).max))])
)
(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 == int64(1<<15+umagic(16,c).m/2) && s == 16+umagic(16,c).s-1
&& x.Op != OpConst16 && udivisibleOK(16,c)
-> (Leq16U
(RotateLeft16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int64(int16(udivisible(16,c).m))])
x)
(Const16 <typ.UInt16> [int64(16-udivisible(16,c).k)])
)
(Const16 <typ.UInt16> [int64(int16(udivisible(16,c).max))])
)
(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 == int64(1<<15+(umagic(16,c).m+1)/2) && s == 16+umagic(16,c).s-2
&& x.Op != OpConst16 && udivisibleOK(16,c)
-> (Leq16U
(RotateLeft16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int64(int16(udivisible(16,c).m))])
x)
(Const16 <typ.UInt16> [int64(16-udivisible(16,c).k)])
)
(Const16 <typ.UInt16> [int64(int16(udivisible(16,c).max))])
)
(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 == int64(umagic(16,c).m) && s == 16+umagic(16,c).s-1
&& x.Op != OpConst16 && udivisibleOK(16,c)
-> (Leq16U
(RotateLeft16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int64(int16(udivisible(16,c).m))])
x)
(Const16 <typ.UInt16> [int64(16-udivisible(16,c).k)])
)
(Const16 <typ.UInt16> [int64(int16(udivisible(16,c).max))])
)
(Eq32 x (Mul32 (Const32 [c])
(Rsh32Ux64
mul:(Hmul32u
(Const32 [m])
x)
(Const64 [s]))
)
)
&& v.Block.Func.pass.name != "opt" && mul.Uses == 1
&& m == int64(int32(1<<31+umagic(32,c).m/2)) && s == umagic(32,c).s-1
&& x.Op != OpConst32 && udivisibleOK(32,c)
-> (Leq32U
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).m))])
x)
(Const32 <typ.UInt32> [int64(32-udivisible(32,c).k)])
)
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).max))])
)
(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 == int64(int32(1<<31+(umagic(32,c).m+1)/2)) && s == umagic(32,c).s-2
&& x.Op != OpConst32 && udivisibleOK(32,c)
-> (Leq32U
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).m))])
x)
(Const32 <typ.UInt32> [int64(32-udivisible(32,c).k)])
)
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).max))])
)
(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 == int64(int32(umagic(32,c).m)) && s == umagic(32,c).s-1
&& x.Op != OpConst32 && udivisibleOK(32,c)
-> (Leq32U
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).m))])
x)
(Const32 <typ.UInt32> [int64(32-udivisible(32,c).k)])
)
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).max))])
)
(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+umagic(32,c).m/2) && s == 32+umagic(32,c).s-1
&& x.Op != OpConst32 && udivisibleOK(32,c)
-> (Leq32U
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).m))])
x)
(Const32 <typ.UInt32> [int64(32-udivisible(32,c).k)])
)
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).max))])
)
(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+(umagic(32,c).m+1)/2) && s == 32+umagic(32,c).s-2
&& x.Op != OpConst32 && udivisibleOK(32,c)
-> (Leq32U
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).m))])
x)
(Const32 <typ.UInt32> [int64(32-udivisible(32,c).k)])
)
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).max))])
)
(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(umagic(32,c).m) && s == 32+umagic(32,c).s-1
&& x.Op != OpConst32 && udivisibleOK(32,c)
-> (Leq32U
(RotateLeft32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).m))])
x)
(Const32 <typ.UInt32> [int64(32-udivisible(32,c).k)])
)
(Const32 <typ.UInt32> [int64(int32(udivisible(32,c).max))])
)
(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+umagic(64,c).m/2) && s == umagic(64,c).s-1
&& x.Op != OpConst64 && udivisibleOK(64,c)
-> (Leq64U
(RotateLeft64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(udivisible(64,c).m)])
x)
(Const64 <typ.UInt64> [int64(64-udivisible(64,c).k)])
)
(Const64 <typ.UInt64> [int64(udivisible(64,c).max)])
)
(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+(umagic(64,c).m+1)/2) && s == umagic(64,c).s-2
&& x.Op != OpConst64 && udivisibleOK(64,c)
-> (Leq64U
(RotateLeft64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(udivisible(64,c).m)])
x)
(Const64 <typ.UInt64> [int64(64-udivisible(64,c).k)])
)
(Const64 <typ.UInt64> [int64(udivisible(64,c).max)])
)
(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(umagic(64,c).m) && s == umagic(64,c).s-1
&& x.Op != OpConst64 && udivisibleOK(64,c)
-> (Leq64U
(RotateLeft64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(udivisible(64,c).m)])
x)
(Const64 <typ.UInt64> [int64(64-udivisible(64,c).k)])
)
(Const64 <typ.UInt64> [int64(udivisible(64,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
// 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 == int64(smagic(8,c).m) && s == 8+smagic(8,c).s
&& x.Op != OpConst8 && sdivisibleOK(8,c)
-> (Leq8U
(RotateLeft8 <typ.UInt8>
(Add8 <typ.UInt8>
(Mul8 <typ.UInt8>
(Const8 <typ.UInt8> [int64(int8(sdivisible(8,c).m))])
x)
(Const8 <typ.UInt8> [int64(int8(sdivisible(8,c).a))])
)
(Const8 <typ.UInt8> [int64(8-sdivisible(8,c).k)])
)
(Const8 <typ.UInt8> [int64(int8(sdivisible(8,c).max))])
)
(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 == int64(smagic(16,c).m) && s == 16+smagic(16,c).s
&& x.Op != OpConst16 && sdivisibleOK(16,c)
-> (Leq16U
(RotateLeft16 <typ.UInt16>
(Add16 <typ.UInt16>
(Mul16 <typ.UInt16>
(Const16 <typ.UInt16> [int64(int16(sdivisible(16,c).m))])
x)
(Const16 <typ.UInt16> [int64(int16(sdivisible(16,c).a))])
)
(Const16 <typ.UInt16> [int64(16-sdivisible(16,c).k)])
)
(Const16 <typ.UInt16> [int64(int16(sdivisible(16,c).max))])
)
(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(smagic(32,c).m) && s == 32+smagic(32,c).s
&& x.Op != OpConst32 && sdivisibleOK(32,c)
-> (Leq32U
(RotateLeft32 <typ.UInt32>
(Add32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int64(int32(sdivisible(32,c).m))])
x)
(Const32 <typ.UInt32> [int64(int32(sdivisible(32,c).a))])
)
(Const32 <typ.UInt32> [int64(32-sdivisible(32,c).k)])
)
(Const32 <typ.UInt32> [int64(int32(sdivisible(32,c).max))])
)
(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 == int64(int32(smagic(32,c).m/2)) && s == smagic(32,c).s-1
&& x.Op != OpConst32 && sdivisibleOK(32,c)
-> (Leq32U
(RotateLeft32 <typ.UInt32>
(Add32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int64(int32(sdivisible(32,c).m))])
x)
(Const32 <typ.UInt32> [int64(int32(sdivisible(32,c).a))])
)
(Const32 <typ.UInt32> [int64(32-sdivisible(32,c).k)])
)
(Const32 <typ.UInt32> [int64(int32(sdivisible(32,c).max))])
)
(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 == int64(int32(smagic(32,c).m)) && s == smagic(32,c).s
&& x.Op != OpConst32 && sdivisibleOK(32,c)
-> (Leq32U
(RotateLeft32 <typ.UInt32>
(Add32 <typ.UInt32>
(Mul32 <typ.UInt32>
(Const32 <typ.UInt32> [int64(int32(sdivisible(32,c).m))])
x)
(Const32 <typ.UInt32> [int64(int32(sdivisible(32,c).a))])
)
(Const32 <typ.UInt32> [int64(32-sdivisible(32,c).k)])
)
(Const32 <typ.UInt32> [int64(int32(sdivisible(32,c).max))])
)
(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(smagic(64,c).m/2) && s == smagic(64,c).s-1
&& x.Op != OpConst64 && sdivisibleOK(64,c)
-> (Leq64U
(RotateLeft64 <typ.UInt64>
(Add64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(sdivisible(64,c).m)])
x)
(Const64 <typ.UInt64> [int64(sdivisible(64,c).a)])
)
(Const64 <typ.UInt64> [int64(64-sdivisible(64,c).k)])
)
(Const64 <typ.UInt64> [int64(sdivisible(64,c).max)])
)
(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(smagic(64,c).m) && s == smagic(64,c).s
&& x.Op != OpConst64 && sdivisibleOK(64,c)
-> (Leq64U
(RotateLeft64 <typ.UInt64>
(Add64 <typ.UInt64>
(Mul64 <typ.UInt64>
(Const64 <typ.UInt64> [int64(sdivisible(64,c).m)])
x)
(Const64 <typ.UInt64> [int64(sdivisible(64,c).a)])
)
(Const64 <typ.UInt64> [int64(64-sdivisible(64,c).k)])
)
(Const64 <typ.UInt64> [int64(sdivisible(64,c).max)])
)
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.
cmd/compile: add signed indivisibility by power of 2 rules Commit 44343c777c (CL 173557) added rules for handling divisibility checks for powers of 2 for signed integers, x%c ==0. This change adds the complementary indivisibility rules, x%c != 0. Fixes #34166 Change-Id: I87379e30af7aff633371acca82db2397da9b2c07 Reviewed-on: https://go-review.googlesource.com/c/go/+/194219 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-09-08 21:50:07 -06:00
((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
cmd/compile: add signed indivisibility by power of 2 rules Commit 44343c777c (CL 173557) added rules for handling divisibility checks for powers of 2 for signed integers, x%c ==0. This change adds the complementary indivisibility rules, x%c != 0. Fixes #34166 Change-Id: I87379e30af7aff633371acca82db2397da9b2c07 Reviewed-on: https://go-review.googlesource.com/c/go/+/194219 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-09-08 21:50:07 -06:00
-> ((Eq8|Neq8) (And8 <t> n (Const8 <t> [int64(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
cmd/compile: add signed indivisibility by power of 2 rules Commit 44343c777c (CL 173557) added rules for handling divisibility checks for powers of 2 for signed integers, x%c ==0. This change adds the complementary indivisibility rules, x%c != 0. Fixes #34166 Change-Id: I87379e30af7aff633371acca82db2397da9b2c07 Reviewed-on: https://go-review.googlesource.com/c/go/+/194219 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-09-08 21:50:07 -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
cmd/compile: add signed indivisibility by power of 2 rules Commit 44343c777c (CL 173557) added rules for handling divisibility checks for powers of 2 for signed integers, x%c ==0. This change adds the complementary indivisibility rules, x%c != 0. Fixes #34166 Change-Id: I87379e30af7aff633371acca82db2397da9b2c07 Reviewed-on: https://go-review.googlesource.com/c/go/+/194219 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-09-08 21:50:07 -06:00
-> ((Eq16|Neq16) (And16 <t> n (Const16 <t> [int64(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
cmd/compile: add signed indivisibility by power of 2 rules Commit 44343c777c (CL 173557) added rules for handling divisibility checks for powers of 2 for signed integers, x%c ==0. This change adds the complementary indivisibility rules, x%c != 0. Fixes #34166 Change-Id: I87379e30af7aff633371acca82db2397da9b2c07 Reviewed-on: https://go-review.googlesource.com/c/go/+/194219 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-09-08 21:50:07 -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
cmd/compile: add signed indivisibility by power of 2 rules Commit 44343c777c (CL 173557) added rules for handling divisibility checks for powers of 2 for signed integers, x%c ==0. This change adds the complementary indivisibility rules, x%c != 0. Fixes #34166 Change-Id: I87379e30af7aff633371acca82db2397da9b2c07 Reviewed-on: https://go-review.googlesource.com/c/go/+/194219 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-09-08 21:50:07 -06:00
-> ((Eq32|Neq32) (And32 <t> n (Const32 <t> [int64(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
cmd/compile: add signed indivisibility by power of 2 rules Commit 44343c777c (CL 173557) added rules for handling divisibility checks for powers of 2 for signed integers, x%c ==0. This change adds the complementary indivisibility rules, x%c != 0. Fixes #34166 Change-Id: I87379e30af7aff633371acca82db2397da9b2c07 Reviewed-on: https://go-review.googlesource.com/c/go/+/194219 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-09-08 21:50:07 -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
cmd/compile: add signed indivisibility by power of 2 rules Commit 44343c777c (CL 173557) added rules for handling divisibility checks for powers of 2 for signed integers, x%c ==0. This change adds the complementary indivisibility rules, x%c != 0. Fixes #34166 Change-Id: I87379e30af7aff633371acca82db2397da9b2c07 Reviewed-on: https://go-review.googlesource.com/c/go/+/194219 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-09-08 21:50:07 -06:00
-> ((Eq64|Neq64) (And64 <t> n (Const64 <t> [int64(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
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
(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 bitset tests The assembly output for x & c == c, where c is power of 2: MOVQ "".set+8(SP), AX ANDQ $8, AX CMPQ AX, $8 SETEQ "".~r2+24(SP) With optimization using bitset: MOVQ "".set+8(SP), AX BTL $3, AX SETCS "".~r2+24(SP) output less than 1 instruction. However, there is no speed improvement: name old time/op new time/op delta AllBitSet-8 0.35ns ± 0% 0.35ns ± 0% ~ (all equal) Fixes #31904 Change-Id: I5dca4e410bf45716ed2145e3473979ec997e35d4 Reviewed-on: https://go-review.googlesource.com/c/go/+/175957 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-05-08 17:02:23 +07:00
// Optimize bitsets
(Eq(8|16|32|64) (And(8|16|32|64) <t> x (Const(8|16|32|64) <t> [y])) (Const(8|16|32|64) <t> [y])) && isPowerOfTwo(y)
-> (Neq(8|16|32|64) (And(8|16|32|64) <t> x (Const(8|16|32|64) <t> [y])) (Const(8|16|32|64) <t> [0]))
(Neq(8|16|32|64) (And(8|16|32|64) <t> x (Const(8|16|32|64) <t> [y])) (Const(8|16|32|64) <t> [y])) && isPowerOfTwo(y)
-> (Eq(8|16|32|64) (And(8|16|32|64) <t> x (Const(8|16|32|64) <t> [y])) (Const(8|16|32|64) <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.
// 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))
// 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))
// x + (z - C) -> (x + z) - C
(Add64 (Sub64 z i:(Const64 <t>)) x) && (z.Op != OpConst64 && x.Op != OpConst64) -> (Sub64 (Add64 <t> x z) i)
(Add32 (Sub32 z i:(Const32 <t>)) x) && (z.Op != OpConst32 && x.Op != OpConst32) -> (Sub32 (Add32 <t> x z) i)
(Add16 (Sub16 z i:(Const16 <t>)) x) && (z.Op != OpConst16 && x.Op != OpConst16) -> (Sub16 (Add16 <t> x z) i)
(Add8 (Sub8 z i:(Const8 <t>)) x) && (z.Op != OpConst8 && x.Op != OpConst8) -> (Sub8 (Add8 <t> x z) i)
// 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)
// x - (z - C) -> x + (C - z) -> (x - z) + C
(Sub64 x (Sub64 z i:(Const64 <t>))) && (z.Op != OpConst64 && x.Op != OpConst64) -> (Add64 i (Sub64 <t> x z))
(Sub32 x (Sub32 z i:(Const32 <t>))) && (z.Op != OpConst32 && x.Op != OpConst32) -> (Add32 i (Sub32 <t> x z))
(Sub16 x (Sub16 z i:(Const16 <t>))) && (z.Op != OpConst16 && x.Op != OpConst16) -> (Add16 i (Sub16 <t> x z))
(Sub8 x (Sub8 z i:(Const8 <t>))) && (z.Op != OpConst8 && x.Op != OpConst8) -> (Add8 i (Sub8 <t> x z))
// 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))
// 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))
// 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))
// 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> [int64(int32(c+d))]) x)
(Add16 (Const16 <t> [c]) (Add16 (Const16 <t> [d]) x)) -> (Add16 (Const16 <t> [int64(int16(c+d))]) x)
(Add8 (Const8 <t> [c]) (Add8 (Const8 <t> [d]) x)) -> (Add8 (Const8 <t> [int64(int8(c+d))]) x)
// 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> [int64(int32(c+d))]) x)
(Add16 (Const16 <t> [c]) (Sub16 (Const16 <t> [d]) x)) -> (Sub16 (Const16 <t> [int64(int16(c+d))]) x)
(Add8 (Const8 <t> [c]) (Sub8 (Const8 <t> [d]) x)) -> (Sub8 (Const8 <t> [int64(int8(c+d))]) x)
// C + (x - D) -> (C - D) + x
(Add64 (Const64 <t> [c]) (Sub64 x (Const64 <t> [d]))) -> (Add64 (Const64 <t> [c-d]) x)
(Add32 (Const32 <t> [c]) (Sub32 x (Const32 <t> [d]))) -> (Add32 (Const32 <t> [int64(int32(c-d))]) x)
(Add16 (Const16 <t> [c]) (Sub16 x (Const16 <t> [d]))) -> (Add16 (Const16 <t> [int64(int16(c-d))]) x)
(Add8 (Const8 <t> [c]) (Sub8 x (Const8 <t> [d]))) -> (Add8 (Const8 <t> [int64(int8(c-d))]) x)
// C - (x - D) -> (C + D) - x
(Sub64 (Const64 <t> [c]) (Sub64 x (Const64 <t> [d]))) -> (Sub64 (Const64 <t> [c+d]) x)
(Sub32 (Const32 <t> [c]) (Sub32 x (Const32 <t> [d]))) -> (Sub32 (Const32 <t> [int64(int32(c+d))]) x)
(Sub16 (Const16 <t> [c]) (Sub16 x (Const16 <t> [d]))) -> (Sub16 (Const16 <t> [int64(int16(c+d))]) x)
(Sub8 (Const8 <t> [c]) (Sub8 x (Const8 <t> [d]))) -> (Sub8 (Const8 <t> [int64(int8(c+d))]) x)
// 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> [int64(int32(c-d))]) x)
(Sub16 (Const16 <t> [c]) (Sub16 (Const16 <t> [d]) x)) -> (Add16 (Const16 <t> [int64(int16(c-d))]) x)
(Sub8 (Const8 <t> [c]) (Sub8 (Const8 <t> [d]) x)) -> (Add8 (Const8 <t> [int64(int8(c-d))]) x)
// 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> [int64(int32(c&d))]) x)
(And16 (Const16 <t> [c]) (And16 (Const16 <t> [d]) x)) -> (And16 (Const16 <t> [int64(int16(c&d))]) x)
(And8 (Const8 <t> [c]) (And8 (Const8 <t> [d]) x)) -> (And8 (Const8 <t> [int64(int8(c&d))]) x)
// 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> [int64(int32(c|d))]) x)
(Or16 (Const16 <t> [c]) (Or16 (Const16 <t> [d]) x)) -> (Or16 (Const16 <t> [int64(int16(c|d))]) x)
(Or8 (Const8 <t> [c]) (Or8 (Const8 <t> [d]) x)) -> (Or8 (Const8 <t> [int64(int8(c|d))]) x)
// 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> [int64(int32(c^d))]) x)
(Xor16 (Const16 <t> [c]) (Xor16 (Const16 <t> [d]) x)) -> (Xor16 (Const16 <t> [int64(int16(c^d))]) x)
(Xor8 (Const8 <t> [c]) (Xor8 (Const8 <t> [d]) x)) -> (Xor8 (Const8 <t> [int64(int8(c^d))]) x)
// 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> [int64(int32(c*d))]) x)
(Mul16 (Const16 <t> [c]) (Mul16 (Const16 <t> [d]) x)) -> (Mul16 (Const16 <t> [int64(int16(c*d))]) x)
(Mul8 (Const8 <t> [c]) (Mul8 (Const8 <t> [d]) x)) -> (Mul8 (Const8 <t> [int64(int8(c*d))]) x)
[dev.ssa] cmd/compile: add floating point optimizations in SSA for ARM Add some simplification rules for floating point ops. cmd/internal/obj/arm supports instructions that compare FP register to 0, but runtime softfloat simulator does not. This CL adds these instructions to softfloat simulator as well. Updates #15365. Change-Id: I29405b2bfcb4c8cf106cb7a1a811409fec91b170 Reviewed-on: https://go-review.googlesource.com/24790 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-07-06 10:04:45 -04:00
// floating point optimizations
cmd/compile: avoid more float32 <-> float64 conversions in compiler Use the new custom truncate/extension code when storing or extracting float32 values from AuxInts to avoid the value being changed by the host platform's floating point conversion instructions (e.g. sNaN -> qNaN). Updates #27516. Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974 Reviewed-on: https://go-review.googlesource.com/134855 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-09-12 12:16:50 +01:00
(Mul(32|64)F x (Const(32|64)F [auxFrom64F(1)])) -> x
(Mul32F x (Const32F [auxFrom32F(-1)])) -> (Neg32F x)
(Mul64F x (Const64F [auxFrom64F(-1)])) -> (Neg64F x)
(Mul32F x (Const32F [auxFrom32F(2)])) -> (Add32F x x)
(Mul64F x (Const64F [auxFrom64F(2)])) -> (Add64F x x)
cmd/compile: use | in the last repetitive generic rules This change or-ifies the last low-hanging rules in generic. Again, this is limited at short and repetitive rules, where the use or ors does not impact readability. Ran rulegen, no change in the actual compiler code. Change-Id: I972b523bc08532f173a3645b47d6936b6e1218c8 Reviewed-on: https://go-review.googlesource.com/96335 Reviewed-by: Giovanni Bajo <rasky@develer.com>
2018-02-22 14:32:57 +01:00
cmd/compile: avoid more float32 <-> float64 conversions in compiler Use the new custom truncate/extension code when storing or extracting float32 values from AuxInts to avoid the value being changed by the host platform's floating point conversion instructions (e.g. sNaN -> qNaN). Updates #27516. Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974 Reviewed-on: https://go-review.googlesource.com/134855 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-09-12 12:16:50 +01:00
(Div32F x (Const32F <t> [c])) && reciprocalExact32(auxTo32F(c)) -> (Mul32F x (Const32F <t> [auxFrom32F(1/auxTo32F(c))]))
(Div64F x (Const64F <t> [c])) && reciprocalExact64(auxTo64F(c)) -> (Mul64F x (Const64F <t> [auxFrom64F(1/auxTo64F(c))]))
cmd/compile: precompute constant square roots If a program wants to evaluate math.Sqrt for any constant value (for example, math.Sqrt(3)), we can replace that expression with its evaluation (1.7320508075688772) at compile time, instead of generating a SQRT assembly command or equivalent. Adds tests that math.Sqrt generates the correct values. I also compiled a short program and verified that the Sqrt expression was replaced by a constant value in the "after opt" step. Adds a short doc to the top of generic.rules explaining what the file does and how other files interact with it. Fixes #15543. Change-Id: I6b6e63ac61cec50763a09ba581024adeee03d4fa Reviewed-on: https://go-review.googlesource.com/27457 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Keith Randall <khr@golang.org>
2016-08-20 22:05:47 -07:00
Revert "cmd/compile: don't allow NaNs in floating-point constant ops" This reverts CL 213477. Reason for revert: tests are failing on linux-mips*-rtrk builders. Change-Id: I8168f7450890233f1bd7e53930b73693c26d4dc0 Reviewed-on: https://go-review.googlesource.com/c/go/+/220897 Run-TryBot: Bryan C. Mills <bcmills@google.com> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2020-02-25 14:36:38 +00:00
(Sqrt (Const64F [c])) -> (Const64F [auxFrom64F(math.Sqrt(auxTo64F(c)))])
cmd/compile: remove Zero and NilCheck for newobject Recognize runtime.newobject and don't Zero or NilCheck it. Fixes #15914 (?) Updates #15390. TBD: add test Change-Id: Ia3bfa5c2ddbe2c27c92d9f68534a713b5ce95934 Reviewed-on: https://go-review.googlesource.com/27930 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-08-26 15:41:51 -04:00
// recognize runtime.newobject and don't Zero/Nilcheck it
cmd/compile: redo nil checks Get rid of BlockCheck. Josh goaded me into it, and I went down a rabbithole making it happen. NilCheck now panics if the pointer is nil and returns void, as before. BlockCheck is gone, and NilCheck is no longer a Control value for any block. It just exists (and deadcode knows not to throw it away). I rewrote the nilcheckelim pass to handle this case. In particular, there can now be multiple NilCheck ops per block. I moved all of the arch-dependent nil check elimination done as part of ssaGenValue into its own proper pass, so we don't have to duplicate that code for every architecture. Making the arch-dependent nil check its own pass means I needed to add a bunch of flags to the opcode table so I could write the code without arch-dependent ops everywhere. Change-Id: I419f891ac9b0de313033ff09115c374163416a9f Reviewed-on: https://go-review.googlesource.com/29120 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-09-13 17:01:01 -07:00
(Zero (Load (OffPtr [c] (SP)) mem) mem)
cmd/compile: fix optimization of Zero newobject on amd64p32 On amd64p32, PtrSize and RegSize don't agree, and function return value is aligned with RegSize. Fix this rule. Other architectures are not affected, where PtrSize and RegSize are the same. Change-Id: If187d3dfde3dc3b931b8e97db5eeff49a781551b Reviewed-on: https://go-review.googlesource.com/37720 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-03 13:53:13 -05:00
&& mem.Op == OpStaticCall
cmd/compile: redo nil checks Get rid of BlockCheck. Josh goaded me into it, and I went down a rabbithole making it happen. NilCheck now panics if the pointer is nil and returns void, as before. BlockCheck is gone, and NilCheck is no longer a Control value for any block. It just exists (and deadcode knows not to throw it away). I rewrote the nilcheckelim pass to handle this case. In particular, there can now be multiple NilCheck ops per block. I moved all of the arch-dependent nil check elimination done as part of ssaGenValue into its own proper pass, so we don't have to duplicate that code for every architecture. Making the arch-dependent nil check its own pass means I needed to add a bunch of flags to the opcode table so I could write the code without arch-dependent ops everywhere. Change-Id: I419f891ac9b0de313033ff09115c374163416a9f Reviewed-on: https://go-review.googlesource.com/29120 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-09-13 17:01:01 -07:00
&& isSameSym(mem.Aux, "runtime.newobject")
cmd/compile: fix optimization of Zero newobject on amd64p32 On amd64p32, PtrSize and RegSize don't agree, and function return value is aligned with RegSize. Fix this rule. Other architectures are not affected, where PtrSize and RegSize are the same. Change-Id: If187d3dfde3dc3b931b8e97db5eeff49a781551b Reviewed-on: https://go-review.googlesource.com/37720 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-03-03 13:53:13 -05:00
&& c == config.ctxt.FixedFrameSize() + config.RegSize // offset of return value
cmd/compile: remove Zero and NilCheck for newobject Recognize runtime.newobject and don't Zero or NilCheck it. Fixes #15914 (?) Updates #15390. TBD: add test Change-Id: Ia3bfa5c2ddbe2c27c92d9f68534a713b5ce95934 Reviewed-on: https://go-review.googlesource.com/27930 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-08-26 15:41:51 -04:00
-> mem
cmd/compile: remove zeroing after newobject The Zero op right after newobject has been removed. But this rule does not cover Store of constant zero (for SSA-able types). Add rules to cover Store op as well. Updates #19027. Change-Id: I5d2b62eeca0aa9ce8dc7205b264b779de01c660b Reviewed-on: https://go-review.googlesource.com/36836 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-02-03 16:18:01 -05:00
(Store (Load (OffPtr [c] (SP)) mem) x mem)
&& isConstZero(x)
&& mem.Op == OpStaticCall
&& isSameSym(mem.Aux, "runtime.newobject")
&& c == config.ctxt.FixedFrameSize() + config.RegSize // offset of return value
-> mem
(Store (OffPtr (Load (OffPtr [c] (SP)) mem)) x mem)
&& isConstZero(x)
&& mem.Op == OpStaticCall
&& isSameSym(mem.Aux, "runtime.newobject")
&& c == config.ctxt.FixedFrameSize() + config.RegSize // offset of return value
-> mem
cmd/compile: redo nil checks Get rid of BlockCheck. Josh goaded me into it, and I went down a rabbithole making it happen. NilCheck now panics if the pointer is nil and returns void, as before. BlockCheck is gone, and NilCheck is no longer a Control value for any block. It just exists (and deadcode knows not to throw it away). I rewrote the nilcheckelim pass to handle this case. In particular, there can now be multiple NilCheck ops per block. I moved all of the arch-dependent nil check elimination done as part of ssaGenValue into its own proper pass, so we don't have to duplicate that code for every architecture. Making the arch-dependent nil check its own pass means I needed to add a bunch of flags to the opcode table so I could write the code without arch-dependent ops everywhere. Change-Id: I419f891ac9b0de313033ff09115c374163416a9f Reviewed-on: https://go-review.googlesource.com/29120 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2016-09-13 17:01:01 -07:00
// nil checks just need to rewrite to something useless.
// they will be deadcode eliminated soon afterwards.
cmd/compile: remove more nil ptr checks after newobject For code like the following (where x escapes): x := []int{1} We're currently generating a nil check. The line above is really 3 operations: t := new([1]int) t[0] = 1 x := t[:] We remove the nil check for t[0] = 1, but not for t[:]. Our current nil check removal rule is too strict about the possible memory arguments of the nil check. Unlike zeroing or storing to the result of runtime.newobject, the nilness of runtime.newobject is always false, even after other stores have happened in the meantime. Change-Id: I95fad4e3a59c27effdb37c43ea215e18f30b1e5f Reviewed-on: https://go-review.googlesource.com/58711 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2017-08-24 15:23:27 -07:00
(NilCheck (Load (OffPtr [c] (SP)) (StaticCall {sym} _)) _)
&& isSameSym(sym, "runtime.newobject")
cmd/compile: re-enable nilcheck removal for newobject Also add compiler debug ouput and add a test. Fixes #15390. Change-Id: Iceba1414c29bcc213b87837387bf8ded1f3157f1 Reviewed-on: https://go-review.googlesource.com/30011 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-09-28 10:20:24 -04:00
&& c == config.ctxt.FixedFrameSize() + config.RegSize // offset of return value
cmd/compile: move v.Pos.line check to warnRule This simplifies the rewrite rules. Change-Id: Iff062297d42a23cb31ad55e8c733842ecbc07da2 Reviewed-on: https://go-review.googlesource.com/129377 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-05-28 14:47:35 -07:00
&& warnRule(fe.Debug_checknil(), v, "removed nil check")
cmd/compile: re-enable nilcheck removal for newobject Also add compiler debug ouput and add a test. Fixes #15390. Change-Id: Iceba1414c29bcc213b87837387bf8ded1f3157f1 Reviewed-on: https://go-review.googlesource.com/30011 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-09-28 10:20:24 -04:00
-> (Invalid)
cmd/compile: remove more nil ptr checks after newobject For code like the following (where x escapes): x := []int{1} We're currently generating a nil check. The line above is really 3 operations: t := new([1]int) t[0] = 1 x := t[:] We remove the nil check for t[0] = 1, but not for t[:]. Our current nil check removal rule is too strict about the possible memory arguments of the nil check. Unlike zeroing or storing to the result of runtime.newobject, the nilness of runtime.newobject is always false, even after other stores have happened in the meantime. Change-Id: I95fad4e3a59c27effdb37c43ea215e18f30b1e5f Reviewed-on: https://go-review.googlesource.com/58711 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2017-08-24 15:23:27 -07:00
(NilCheck (OffPtr (Load (OffPtr [c] (SP)) (StaticCall {sym} _))) _)
&& isSameSym(sym, "runtime.newobject")
cmd/compile: re-enable nilcheck removal for newobject Also add compiler debug ouput and add a test. Fixes #15390. Change-Id: Iceba1414c29bcc213b87837387bf8ded1f3157f1 Reviewed-on: https://go-review.googlesource.com/30011 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-09-28 10:20:24 -04:00
&& c == config.ctxt.FixedFrameSize() + config.RegSize // offset of return value
cmd/compile: move v.Pos.line check to warnRule This simplifies the rewrite rules. Change-Id: Iff062297d42a23cb31ad55e8c733842ecbc07da2 Reviewed-on: https://go-review.googlesource.com/129377 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
2018-05-28 14:47:35 -07:00
&& warnRule(fe.Debug_checknil(), v, "removed nil check")
cmd/compile: re-enable nilcheck removal for newobject Also add compiler debug ouput and add a test. Fixes #15390. Change-Id: Iceba1414c29bcc213b87837387bf8ded1f3157f1 Reviewed-on: https://go-review.googlesource.com/30011 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-09-28 10:20:24 -04:00
-> (Invalid)
cmd/compile: de-virtualize interface calls With this change, code like h := sha1.New() h.Write(buf) sum := h.Sum() gets compiled into static calls rather than interface calls, because the compiler is able to prove that 'h' is really a *sha1.digest. The InterCall re-write rule hits a few dozen times during make.bash, and hundreds of times during all.bash. The most common pattern identified by the compiler is a constructor like func New() Interface { return &impl{...} } where the constructor gets inlined into the caller, and the result is used immediately. Examples include {sha1,md5,crc32,crc64,...}.New, base64.NewEncoder, base64.NewDecoder, errors.New, net.Pipe, and so on. Some existing benchmarks that change on darwin/amd64: Crc64/ISO4KB-8 2.67µs ± 1% 2.66µs ± 0% -0.36% (p=0.015 n=10+10) Crc64/ISO1KB-8 694ns ± 0% 690ns ± 1% -0.59% (p=0.001 n=10+10) Adler32KB-8 473ns ± 1% 471ns ± 0% -0.39% (p=0.010 n=10+9) On architectures like amd64, the reduction in code size appears to contribute more to benchmark improvements than just removing the indirect call, since that branch gets predicted accurately when called in a loop. Updates #19361 Change-Id: I57d4dc21ef40a05ec0fbd55a9bb0eb74cdc67a3d Reviewed-on: https://go-review.googlesource.com/38139 Run-TryBot: Philip Hofer <phofer@umich.edu> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2017-03-13 15:03:17 -07:00
cmd/compile: generate constants for NeqPtr, EqPtr and IsNonNil ops If both inputs are constant offsets from the same pointer then we can evaluate NeqPtr and EqPtr at compile time. Triggers a few times during all.bash. Removes a conditional branch in the following code: copy(x[1:], x[:]) This branch was recently added as an optimization in CL 94596. We now skip the memmove if the pointers are equal. However, in the above code we know at compile time that they are never equal. Also, when the offset is variable, check if the offset is zero rather than if the pointers are equal. For example: copy(x[a:], x[:]) This would now skip the copy if a == 0, rather than if x + a == x. Finally I've also added a rule to make IsNonNil true for pointers to values on the stack. The nil check elimination pass will catch these anyway, but eliminating them here might eliminate branches earlier. Change-Id: If72f436fef0a96ad0f4e296d3a1f8b6c3e712085 Reviewed-on: https://go-review.googlesource.com/106635 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-12 15:24:08 +01:00
// Evaluate constant address comparisons.
(EqPtr x x) -> (ConstBool [1])
cmd/compile/internal/ssa: handle symbol address comparisons consistently CL 38338 introduced SSA rules to optimize two types of pointer equality tests: a pointer compared with itself, and comparison of addresses taken of two symbols which may have the same base. This patch adds rules to apply the same optimization to pointer inequality tests, which also ensures that two pointers to zero-width types cannot be both equal and unequal at the same time. Fixes #24503. Change-Id: Ic828aeb86ae2e680caf66c35f4c247674768a9ba Reviewed-on: https://go-review.googlesource.com/102275 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2018-03-23 04:14:25 -04:00
(NeqPtr x x) -> (ConstBool [0])
cmd/compile: generate constants for NeqPtr, EqPtr and IsNonNil ops If both inputs are constant offsets from the same pointer then we can evaluate NeqPtr and EqPtr at compile time. Triggers a few times during all.bash. Removes a conditional branch in the following code: copy(x[1:], x[:]) This branch was recently added as an optimization in CL 94596. We now skip the memmove if the pointers are equal. However, in the above code we know at compile time that they are never equal. Also, when the offset is variable, check if the offset is zero rather than if the pointers are equal. For example: copy(x[a:], x[:]) This would now skip the copy if a == 0, rather than if x + a == x. Finally I've also added a rule to make IsNonNil true for pointers to values on the stack. The nil check elimination pass will catch these anyway, but eliminating them here might eliminate branches earlier. Change-Id: If72f436fef0a96ad0f4e296d3a1f8b6c3e712085 Reviewed-on: https://go-review.googlesource.com/106635 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-12 15:24:08 +01:00
(EqPtr (Addr {a} _) (Addr {b} _)) -> (ConstBool [b2i(a == b)])
cmd/compile: optimize more pointer comparisons The existing pointer comparison optimizations don't include pointer arithmetic. Add them. These rules trigger a few times in std cmd, while compiling: time.Duration.String cmd/go/internal/tlog.NodeHash crypto/tls.ticketKeyFromBytes (3 times) crypto/elliptic.(*p256Point).p256ScalarMult (15 times!) crypto/elliptic.initTable These weird comparisons occur when using the copy builtin, which does a pointer comparison between src and dst. This also happens to fix #32454, by optimizing enough early on that all values can be eliminated. Fixes #32454 Change-Id: I799d45743350bddd15a295dc1e12f8d03c11d1c6 Reviewed-on: https://go-review.googlesource.com/c/go/+/180940 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-06-06 11:40:47 -07:00
(EqPtr (Addr {a} _) (OffPtr [o] (Addr {b} _))) -> (ConstBool [b2i(a == b && o == 0)])
(EqPtr (OffPtr [o1] (Addr {a} _)) (OffPtr [o2] (Addr {b} _))) -> (ConstBool [b2i(a == b && o1 == o2)])
cmd/compile: generate constants for NeqPtr, EqPtr and IsNonNil ops If both inputs are constant offsets from the same pointer then we can evaluate NeqPtr and EqPtr at compile time. Triggers a few times during all.bash. Removes a conditional branch in the following code: copy(x[1:], x[:]) This branch was recently added as an optimization in CL 94596. We now skip the memmove if the pointers are equal. However, in the above code we know at compile time that they are never equal. Also, when the offset is variable, check if the offset is zero rather than if the pointers are equal. For example: copy(x[a:], x[:]) This would now skip the copy if a == 0, rather than if x + a == x. Finally I've also added a rule to make IsNonNil true for pointers to values on the stack. The nil check elimination pass will catch these anyway, but eliminating them here might eliminate branches earlier. Change-Id: If72f436fef0a96ad0f4e296d3a1f8b6c3e712085 Reviewed-on: https://go-review.googlesource.com/106635 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-12 15:24:08 +01:00
(NeqPtr (Addr {a} _) (Addr {b} _)) -> (ConstBool [b2i(a != b)])
cmd/compile: optimize more pointer comparisons The existing pointer comparison optimizations don't include pointer arithmetic. Add them. These rules trigger a few times in std cmd, while compiling: time.Duration.String cmd/go/internal/tlog.NodeHash crypto/tls.ticketKeyFromBytes (3 times) crypto/elliptic.(*p256Point).p256ScalarMult (15 times!) crypto/elliptic.initTable These weird comparisons occur when using the copy builtin, which does a pointer comparison between src and dst. This also happens to fix #32454, by optimizing enough early on that all values can be eliminated. Fixes #32454 Change-Id: I799d45743350bddd15a295dc1e12f8d03c11d1c6 Reviewed-on: https://go-review.googlesource.com/c/go/+/180940 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-06-06 11:40:47 -07:00
(NeqPtr (Addr {a} _) (OffPtr [o] (Addr {b} _))) -> (ConstBool [b2i(a != b || o != 0)])
(NeqPtr (OffPtr [o1] (Addr {a} _)) (OffPtr [o2] (Addr {b} _))) -> (ConstBool [b2i(a != b || o1 != o2)])
cmd/compile: add LocalAddr that takes SP,mem operands Lack of a well-defined order between VarDef and related address operations sometimes causes problems with store order and write barrier transformations; glitches in the order are made irreparable (by later optimizations) if the two parts of the glitch straddle a split in the original block caused by insertion of a write barrier diamond. Fix this by creating a LocalAddr for addresses of locals (what VarDef matters for) that takes a memory input to help make the order explicit. Addr is modified to only be legal for SB operand, so there is no overlap between Addr and LocalAddr uses (there may be some downstream cleanup from this). Changes to generic.rules and rewrite.go ensure that codegen tests continue to pass; CSE of LocalAddr is impaired, not quite sure of the cost. Fixes #26105. Change-Id: Id4192b4440aa4e9d7ba54a465c456df9b530b515 Reviewed-on: https://go-review.googlesource.com/122483 Run-TryBot: David Chase <drchase@google.com> Reviewed-by: Keith Randall <khr@golang.org>
2018-07-03 11:34:38 -04:00
(EqPtr (LocalAddr {a} _ _) (LocalAddr {b} _ _)) -> (ConstBool [b2i(a == b)])
cmd/compile: optimize more pointer comparisons The existing pointer comparison optimizations don't include pointer arithmetic. Add them. These rules trigger a few times in std cmd, while compiling: time.Duration.String cmd/go/internal/tlog.NodeHash crypto/tls.ticketKeyFromBytes (3 times) crypto/elliptic.(*p256Point).p256ScalarMult (15 times!) crypto/elliptic.initTable These weird comparisons occur when using the copy builtin, which does a pointer comparison between src and dst. This also happens to fix #32454, by optimizing enough early on that all values can be eliminated. Fixes #32454 Change-Id: I799d45743350bddd15a295dc1e12f8d03c11d1c6 Reviewed-on: https://go-review.googlesource.com/c/go/+/180940 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-06-06 11:40:47 -07:00
(EqPtr (LocalAddr {a} _ _) (OffPtr [o] (LocalAddr {b} _ _))) -> (ConstBool [b2i(a == b && o == 0)])
(EqPtr (OffPtr [o1] (LocalAddr {a} _ _)) (OffPtr [o2] (LocalAddr {b} _ _))) -> (ConstBool [b2i(a == b && o1 == o2)])
cmd/compile: add LocalAddr that takes SP,mem operands Lack of a well-defined order between VarDef and related address operations sometimes causes problems with store order and write barrier transformations; glitches in the order are made irreparable (by later optimizations) if the two parts of the glitch straddle a split in the original block caused by insertion of a write barrier diamond. Fix this by creating a LocalAddr for addresses of locals (what VarDef matters for) that takes a memory input to help make the order explicit. Addr is modified to only be legal for SB operand, so there is no overlap between Addr and LocalAddr uses (there may be some downstream cleanup from this). Changes to generic.rules and rewrite.go ensure that codegen tests continue to pass; CSE of LocalAddr is impaired, not quite sure of the cost. Fixes #26105. Change-Id: Id4192b4440aa4e9d7ba54a465c456df9b530b515 Reviewed-on: https://go-review.googlesource.com/122483 Run-TryBot: David Chase <drchase@google.com> Reviewed-by: Keith Randall <khr@golang.org>
2018-07-03 11:34:38 -04:00
(NeqPtr (LocalAddr {a} _ _) (LocalAddr {b} _ _)) -> (ConstBool [b2i(a != b)])
cmd/compile: optimize more pointer comparisons The existing pointer comparison optimizations don't include pointer arithmetic. Add them. These rules trigger a few times in std cmd, while compiling: time.Duration.String cmd/go/internal/tlog.NodeHash crypto/tls.ticketKeyFromBytes (3 times) crypto/elliptic.(*p256Point).p256ScalarMult (15 times!) crypto/elliptic.initTable These weird comparisons occur when using the copy builtin, which does a pointer comparison between src and dst. This also happens to fix #32454, by optimizing enough early on that all values can be eliminated. Fixes #32454 Change-Id: I799d45743350bddd15a295dc1e12f8d03c11d1c6 Reviewed-on: https://go-review.googlesource.com/c/go/+/180940 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-06-06 11:40:47 -07:00
(NeqPtr (LocalAddr {a} _ _) (OffPtr [o] (LocalAddr {b} _ _))) -> (ConstBool [b2i(a != b || o != 0)])
(NeqPtr (OffPtr [o1] (LocalAddr {a} _ _)) (OffPtr [o2] (LocalAddr {b} _ _))) -> (ConstBool [b2i(a != b || o1 != o2)])
cmd/compile: generate constants for NeqPtr, EqPtr and IsNonNil ops If both inputs are constant offsets from the same pointer then we can evaluate NeqPtr and EqPtr at compile time. Triggers a few times during all.bash. Removes a conditional branch in the following code: copy(x[1:], x[:]) This branch was recently added as an optimization in CL 94596. We now skip the memmove if the pointers are equal. However, in the above code we know at compile time that they are never equal. Also, when the offset is variable, check if the offset is zero rather than if the pointers are equal. For example: copy(x[a:], x[:]) This would now skip the copy if a == 0, rather than if x + a == x. Finally I've also added a rule to make IsNonNil true for pointers to values on the stack. The nil check elimination pass will catch these anyway, but eliminating them here might eliminate branches earlier. Change-Id: If72f436fef0a96ad0f4e296d3a1f8b6c3e712085 Reviewed-on: https://go-review.googlesource.com/106635 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-12 15:24:08 +01:00
(EqPtr (OffPtr [o1] p1) p2) && isSamePtr(p1, p2) -> (ConstBool [b2i(o1 == 0)])
(NeqPtr (OffPtr [o1] p1) p2) && isSamePtr(p1, p2) -> (ConstBool [b2i(o1 != 0)])
(EqPtr (OffPtr [o1] p1) (OffPtr [o2] p2)) && isSamePtr(p1, p2) -> (ConstBool [b2i(o1 == o2)])
(NeqPtr (OffPtr [o1] p1) (OffPtr [o2] p2)) && isSamePtr(p1, p2) -> (ConstBool [b2i(o1 != o2)])
(EqPtr (Const(32|64) [c]) (Const(32|64) [d])) -> (ConstBool [b2i(c == d)])
(NeqPtr (Const(32|64) [c]) (Const(32|64) [d])) -> (ConstBool [b2i(c != d)])
cmd/compile: add LocalAddr that takes SP,mem operands Lack of a well-defined order between VarDef and related address operations sometimes causes problems with store order and write barrier transformations; glitches in the order are made irreparable (by later optimizations) if the two parts of the glitch straddle a split in the original block caused by insertion of a write barrier diamond. Fix this by creating a LocalAddr for addresses of locals (what VarDef matters for) that takes a memory input to help make the order explicit. Addr is modified to only be legal for SB operand, so there is no overlap between Addr and LocalAddr uses (there may be some downstream cleanup from this). Changes to generic.rules and rewrite.go ensure that codegen tests continue to pass; CSE of LocalAddr is impaired, not quite sure of the cost. Fixes #26105. Change-Id: Id4192b4440aa4e9d7ba54a465c456df9b530b515 Reviewed-on: https://go-review.googlesource.com/122483 Run-TryBot: David Chase <drchase@google.com> Reviewed-by: Keith Randall <khr@golang.org>
2018-07-03 11:34:38 -04:00
(EqPtr (LocalAddr _ _) (Addr _)) -> (ConstBool [0])
cmd/compile: optimize more pointer comparisons The existing pointer comparison optimizations don't include pointer arithmetic. Add them. These rules trigger a few times in std cmd, while compiling: time.Duration.String cmd/go/internal/tlog.NodeHash crypto/tls.ticketKeyFromBytes (3 times) crypto/elliptic.(*p256Point).p256ScalarMult (15 times!) crypto/elliptic.initTable These weird comparisons occur when using the copy builtin, which does a pointer comparison between src and dst. This also happens to fix #32454, by optimizing enough early on that all values can be eliminated. Fixes #32454 Change-Id: I799d45743350bddd15a295dc1e12f8d03c11d1c6 Reviewed-on: https://go-review.googlesource.com/c/go/+/180940 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-06-06 11:40:47 -07:00
(EqPtr (OffPtr (LocalAddr _ _)) (Addr _)) -> (ConstBool [0])
(EqPtr (LocalAddr _ _) (OffPtr (Addr _))) -> (ConstBool [0])
(EqPtr (OffPtr (LocalAddr _ _)) (OffPtr (Addr _))) -> (ConstBool [0])
cmd/compile: add LocalAddr that takes SP,mem operands Lack of a well-defined order between VarDef and related address operations sometimes causes problems with store order and write barrier transformations; glitches in the order are made irreparable (by later optimizations) if the two parts of the glitch straddle a split in the original block caused by insertion of a write barrier diamond. Fix this by creating a LocalAddr for addresses of locals (what VarDef matters for) that takes a memory input to help make the order explicit. Addr is modified to only be legal for SB operand, so there is no overlap between Addr and LocalAddr uses (there may be some downstream cleanup from this). Changes to generic.rules and rewrite.go ensure that codegen tests continue to pass; CSE of LocalAddr is impaired, not quite sure of the cost. Fixes #26105. Change-Id: Id4192b4440aa4e9d7ba54a465c456df9b530b515 Reviewed-on: https://go-review.googlesource.com/122483 Run-TryBot: David Chase <drchase@google.com> Reviewed-by: Keith Randall <khr@golang.org>
2018-07-03 11:34:38 -04:00
(NeqPtr (LocalAddr _ _) (Addr _)) -> (ConstBool [1])
cmd/compile: optimize more pointer comparisons The existing pointer comparison optimizations don't include pointer arithmetic. Add them. These rules trigger a few times in std cmd, while compiling: time.Duration.String cmd/go/internal/tlog.NodeHash crypto/tls.ticketKeyFromBytes (3 times) crypto/elliptic.(*p256Point).p256ScalarMult (15 times!) crypto/elliptic.initTable These weird comparisons occur when using the copy builtin, which does a pointer comparison between src and dst. This also happens to fix #32454, by optimizing enough early on that all values can be eliminated. Fixes #32454 Change-Id: I799d45743350bddd15a295dc1e12f8d03c11d1c6 Reviewed-on: https://go-review.googlesource.com/c/go/+/180940 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2019-06-06 11:40:47 -07:00
(NeqPtr (OffPtr (LocalAddr _ _)) (Addr _)) -> (ConstBool [1])
(NeqPtr (LocalAddr _ _) (OffPtr (Addr _))) -> (ConstBool [1])
(NeqPtr (OffPtr (LocalAddr _ _)) (OffPtr (Addr _))) -> (ConstBool [1])
cmd/compile: add LocalAddr that takes SP,mem operands Lack of a well-defined order between VarDef and related address operations sometimes causes problems with store order and write barrier transformations; glitches in the order are made irreparable (by later optimizations) if the two parts of the glitch straddle a split in the original block caused by insertion of a write barrier diamond. Fix this by creating a LocalAddr for addresses of locals (what VarDef matters for) that takes a memory input to help make the order explicit. Addr is modified to only be legal for SB operand, so there is no overlap between Addr and LocalAddr uses (there may be some downstream cleanup from this). Changes to generic.rules and rewrite.go ensure that codegen tests continue to pass; CSE of LocalAddr is impaired, not quite sure of the cost. Fixes #26105. Change-Id: Id4192b4440aa4e9d7ba54a465c456df9b530b515 Reviewed-on: https://go-review.googlesource.com/122483 Run-TryBot: David Chase <drchase@google.com> Reviewed-by: Keith Randall <khr@golang.org>
2018-07-03 11:34:38 -04:00
cmd/compile: generate constants for NeqPtr, EqPtr and IsNonNil ops If both inputs are constant offsets from the same pointer then we can evaluate NeqPtr and EqPtr at compile time. Triggers a few times during all.bash. Removes a conditional branch in the following code: copy(x[1:], x[:]) This branch was recently added as an optimization in CL 94596. We now skip the memmove if the pointers are equal. However, in the above code we know at compile time that they are never equal. Also, when the offset is variable, check if the offset is zero rather than if the pointers are equal. For example: copy(x[a:], x[:]) This would now skip the copy if a == 0, rather than if x + a == x. Finally I've also added a rule to make IsNonNil true for pointers to values on the stack. The nil check elimination pass will catch these anyway, but eliminating them here might eliminate branches earlier. Change-Id: If72f436fef0a96ad0f4e296d3a1f8b6c3e712085 Reviewed-on: https://go-review.googlesource.com/106635 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-04-12 15:24:08 +01:00
// 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 [0])
(IsNonNil (Const(32|64) [c])) -> (ConstBool [b2i(c != 0)])
(IsNonNil (Addr _)) -> (ConstBool [1])
cmd/compile: add LocalAddr that takes SP,mem operands Lack of a well-defined order between VarDef and related address operations sometimes causes problems with store order and write barrier transformations; glitches in the order are made irreparable (by later optimizations) if the two parts of the glitch straddle a split in the original block caused by insertion of a write barrier diamond. Fix this by creating a LocalAddr for addresses of locals (what VarDef matters for) that takes a memory input to help make the order explicit. Addr is modified to only be legal for SB operand, so there is no overlap between Addr and LocalAddr uses (there may be some downstream cleanup from this). Changes to generic.rules and rewrite.go ensure that codegen tests continue to pass; CSE of LocalAddr is impaired, not quite sure of the cost. Fixes #26105. Change-Id: Id4192b4440aa4e9d7ba54a465c456df9b530b515 Reviewed-on: https://go-review.googlesource.com/122483 Run-TryBot: David Chase <drchase@google.com> Reviewed-by: Keith Randall <khr@golang.org>
2018-07-03 11:34:38 -04:00
(IsNonNil (LocalAddr _ _)) -> (ConstBool [1])
cmd/compile/internal/ssa: recognize constant pointer comparison Teach the backend to recognize that the address of a symbol is equal with itself, and that the addresses of two different symbols are different. Some examples of where this rule hits in the standard library: - inlined uses of (*time.Time).setLoc (e.g. time.UTC) - inlined uses of bufio.NewReader (via type assertion) Change-Id: I23dcb068c2ec333655c1292917bec13bbd908c24 Reviewed-on: https://go-review.googlesource.com/38338 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-03-15 15:34:52 -07:00
cmd/compile: convert memmove call into Move when arguments are disjoint Move ops can be faster than memmove calls because the number of bytes to be moved is fixed and they don't incur the overhead of a call. This change allows memmove to be converted into a Move op when the arguments are disjoint. The optimization is only enabled on s390x at the moment, however other architectures may also benefit from it in the future. The memmove inlining rule triggers an extra 12 times when compiling the standard library. It will most likely make more of a difference as the disjoint function is improved over time (to recognize fresh heap allocations for example). Change-Id: I9af570dcfff28257b8e59e0ff584a46d8e248310 Reviewed-on: https://go-review.googlesource.com/110064 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
2018-04-29 15:12:50 +01:00
// Inline small or disjoint runtime.memmove calls with constant length.
(StaticCall {sym} s1:(Store _ (Const(64|32) [sz]) s2:(Store _ src s3:(Store {t} _ dst mem))))
cmd/compile: avoid memmove -> SSA move rewrite when size is negative We should panic in this situation. Rewriting to a SSA op just leads to a compiler panic. Fixes #36259 Change-Id: I6e0bccbed7dd0fdac7ebae76b98a211947947386 Reviewed-on: https://go-review.googlesource.com/c/go/+/212405 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2019-12-23 18:16:22 -08:00
&& sz >= 0
cmd/compile: convert memmove call into Move when arguments are disjoint Move ops can be faster than memmove calls because the number of bytes to be moved is fixed and they don't incur the overhead of a call. This change allows memmove to be converted into a Move op when the arguments are disjoint. The optimization is only enabled on s390x at the moment, however other architectures may also benefit from it in the future. The memmove inlining rule triggers an extra 12 times when compiling the standard library. It will most likely make more of a difference as the disjoint function is improved over time (to recognize fresh heap allocations for example). Change-Id: I9af570dcfff28257b8e59e0ff584a46d8e248310 Reviewed-on: https://go-review.googlesource.com/110064 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
2018-04-29 15:12:50 +01:00
&& isSameSym(sym,"runtime.memmove")
cmd/compile: fix crash when memmove argument is not the right type Make sure the argument to memmove is of pointer type before we try to get the element type. This has been noticed for code that uses unsafe+linkname so it can call runtime.memmove. Probably not the best thing to allow, but the code is out there and we'd rather not break it unnecessarily. Fixes #30061 Change-Id: I334a8453f2e293959fd742044c43fbe93f0b3d31 Reviewed-on: https://go-review.googlesource.com/c/160826 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2019-02-01 15:27:53 -08:00
&& t.(*types.Type).IsPtr() // avoids TUINTPTR, see issue 30061
cmd/compile: convert memmove call into Move when arguments are disjoint Move ops can be faster than memmove calls because the number of bytes to be moved is fixed and they don't incur the overhead of a call. This change allows memmove to be converted into a Move op when the arguments are disjoint. The optimization is only enabled on s390x at the moment, however other architectures may also benefit from it in the future. The memmove inlining rule triggers an extra 12 times when compiling the standard library. It will most likely make more of a difference as the disjoint function is improved over time (to recognize fresh heap allocations for example). Change-Id: I9af570dcfff28257b8e59e0ff584a46d8e248310 Reviewed-on: https://go-review.googlesource.com/110064 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
2018-04-29 15:12:50 +01:00
&& s1.Uses == 1 && s2.Uses == 1 && s3.Uses == 1
&& isInlinableMemmove(dst,src,sz,config)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(s1, s2, s3)
cmd/compile: convert memmove call into Move when arguments are disjoint Move ops can be faster than memmove calls because the number of bytes to be moved is fixed and they don't incur the overhead of a call. This change allows memmove to be converted into a Move op when the arguments are disjoint. The optimization is only enabled on s390x at the moment, however other architectures may also benefit from it in the future. The memmove inlining rule triggers an extra 12 times when compiling the standard library. It will most likely make more of a difference as the disjoint function is improved over time (to recognize fresh heap allocations for example). Change-Id: I9af570dcfff28257b8e59e0ff584a46d8e248310 Reviewed-on: https://go-review.googlesource.com/110064 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
2018-04-29 15:12:50 +01:00
-> (Move {t.(*types.Type).Elem()} [sz] dst src mem)
cmd/compile/internal/ssa: inline memmove with known size Replace calls to memmove with known (constant) size, with OpMove. Do it only if it is safe from aliasing point of view. Helps with code like this: append(buf,"const str"...) In strconv this provides nice benefit: Quote-6 731ns ± 2% 647ns ± 3% -11.41% (p=0.000 n=10+10) QuoteRune-6 117ns ± 5% 111ns ± 1% -4.54% (p=0.000 n=10+10) AppendQuote-6 475ns ± 0% 396ns ± 0% -16.59% (p=0.000 n=9+10) AppendQuoteRune-6 32.0ns ± 0% 27.4ns ± 0% -14.41% (p=0.000 n=8+9) Change-Id: I7704f5c51b46aed2d8f033de74c75140fc35036c Reviewed-on: https://go-review.googlesource.com/54394 Run-TryBot: Ilya Tocar <ilya.tocar@intel.com> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-08-09 14:00:38 -05:00
cmd/compile: de-virtualize interface calls With this change, code like h := sha1.New() h.Write(buf) sum := h.Sum() gets compiled into static calls rather than interface calls, because the compiler is able to prove that 'h' is really a *sha1.digest. The InterCall re-write rule hits a few dozen times during make.bash, and hundreds of times during all.bash. The most common pattern identified by the compiler is a constructor like func New() Interface { return &impl{...} } where the constructor gets inlined into the caller, and the result is used immediately. Examples include {sha1,md5,crc32,crc64,...}.New, base64.NewEncoder, base64.NewDecoder, errors.New, net.Pipe, and so on. Some existing benchmarks that change on darwin/amd64: Crc64/ISO4KB-8 2.67µs ± 1% 2.66µs ± 0% -0.36% (p=0.015 n=10+10) Crc64/ISO1KB-8 694ns ± 0% 690ns ± 1% -0.59% (p=0.001 n=10+10) Adler32KB-8 473ns ± 1% 471ns ± 0% -0.39% (p=0.010 n=10+9) On architectures like amd64, the reduction in code size appears to contribute more to benchmark improvements than just removing the indirect call, since that branch gets predicted accurately when called in a loop. Updates #19361 Change-Id: I57d4dc21ef40a05ec0fbd55a9bb0eb74cdc67a3d Reviewed-on: https://go-review.googlesource.com/38139 Run-TryBot: Philip Hofer <phofer@umich.edu> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
2017-03-13 15:03:17 -07:00
// De-virtualize interface calls into static calls.
// Note that (ITab (IMake)) doesn't get
// rewritten until after the first opt pass,
// so this rule should trigger reliably.
(InterCall [argsize] (Load (OffPtr [off] (ITab (IMake (Addr {itab} (SB)) _))) _) mem) && devirt(v, itab, off) != nil ->
(StaticCall [argsize] {devirt(v, itab, off)} 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 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)
(Move {t} [n] dst1 src mem:(VarDef (Zero {t} [n] dst0 _))) && isSamePtr(src, dst0)
-> (Zero {t} [n] dst1 mem)
// 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 + sizeof(t2)
&& clobber(store)
-> (Zero {t1} [n] p1 mem)
(Move {t1} [n] dst1 src1 store:(Store {t2} op:(OffPtr [o2] dst2) _ mem))
&& isSamePtr(dst1, dst2) && store.Uses == 1
&& n >= o2 + sizeof(t2)
&& disjoint(src1, n, op, sizeof(t2))
&& clobber(store)
-> (Move {t1} [n] dst1 src1 mem)
// 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)
(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)
(Zero {t} [n] dst1 vardef:(VarDef {x} move:(Move {t} [n] dst2 _ mem)))
&& move.Uses == 1 && vardef.Uses == 1
&& isSamePtr(dst1, dst2)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(move, vardef)
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 {x} mem))
(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)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(move, vardef)
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 {x} mem))
(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 == sizeof(t2)
&& n == sizeof(t2) + sizeof(t1)
&& isSamePtr(p1, p2) && isSamePtr(p2, p3)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(m2, m3)
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 d1 (Store {t2} op2 d2 mem))
(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 == sizeof(t3)
&& o1-o2 == sizeof(t2)
&& n == sizeof(t3) + sizeof(t2) + sizeof(t1)
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(m2, m3, m4)
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 d1 (Store {t2} op2 d2 (Store {t3} op3 d3 mem)))
(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 == sizeof(t4)
&& o2-o3 == sizeof(t3)
&& o1-o2 == sizeof(t2)
&& n == sizeof(t4) + sizeof(t3) + sizeof(t2) + sizeof(t1)
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(m2, m3, m4, m5)
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 d1 (Store {t2} op2 d2 (Store {t3} op3 d3 (Store {t4} op4 d4 mem))))
// 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)
(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)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(zero, vardef)
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 {x} mem))
(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 == sizeof(t2)
&& n == sizeof(t2) + sizeof(t1)
&& isSamePtr(p1, p2) && isSamePtr(p2, p3)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(m2, m3)
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 d1 (Store {t2} op2 d2 mem))
(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 == sizeof(t3)
&& o1-o2 == sizeof(t2)
&& n == sizeof(t3) + sizeof(t2) + sizeof(t1)
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(m2, m3, m4)
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 d1 (Store {t2} op2 d2 (Store {t3} op3 d3 mem)))
(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 == sizeof(t4)
&& o2-o3 == sizeof(t3)
&& o1-o2 == sizeof(t2)
&& n == sizeof(t4) + sizeof(t3) + sizeof(t2) + sizeof(t1)
&& isSamePtr(p1, p2) && isSamePtr(p2, p3) && isSamePtr(p3, p4) && isSamePtr(p4, p5)
cmd/compile: make clobber variadic There are often many values to clobber. Allow passing them all in at once. The goal is increased rule readability. As a bonus, it shrinks cmd/compile by ~97k, almost half a percent. Package SSA requires 1.2% less memory to compile. The single-line changes were make via regex, and the remaining multi-line clobbers were manually combined. Passes toolstash-check -all. Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19 Reviewed-on: https://go-review.googlesource.com/c/go/+/220691 Reviewed-by: Michael Munday <mike.munday@ibm.com>
2020-02-24 17:30:44 -08:00
&& clobber(m2, m3, m4, m5)
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 d1 (Store {t2} op2 d2 (Store {t3} op3 d3 (Store {t4} op4 d4 mem))))
// Don't Move from memory if the values are likely to already be
// in registers.
(Move {t1} [n] dst p1
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& o2 == sizeof(t3)
&& n == sizeof(t2) + sizeof(t3)
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
-> (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
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& o3 == sizeof(t4)
&& o2-o3 == sizeof(t3)
&& n == sizeof(t2) + sizeof(t3) + sizeof(t4)
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
-> (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
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& alignof(t5) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& registerizable(b, t5)
&& o4 == sizeof(t5)
&& o3-o4 == sizeof(t4)
&& o2-o3 == sizeof(t3)
&& n == sizeof(t2) + sizeof(t3) + sizeof(t4) + sizeof(t5)
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
-> (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
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
(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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& o2 == sizeof(t3)
&& n == sizeof(t2) + sizeof(t3)
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
-> (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
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
(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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& o3 == sizeof(t4)
&& o2-o3 == sizeof(t3)
&& n == sizeof(t2) + sizeof(t3) + sizeof(t4)
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
-> (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
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
(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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& alignof(t5) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& registerizable(b, t5)
&& o4 == sizeof(t5)
&& o3-o4 == sizeof(t4)
&& o2-o3 == sizeof(t3)
&& n == sizeof(t2) + sizeof(t3) + sizeof(t4) + sizeof(t5)
cmd/compile: fix type of OffPtr in some optimization rules In some optimization rules the type of generated OffPtr was incorrectly set to the type of the pointee, instead of the pointer. When the OffPtr value is spilled, this may generate a spill of the wrong type, e.g. a floating point spill of an integer (pointer) value. On Wasm, this leads to invalid bytecode. Fixes #27961. Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc Reviewed-on: https://go-review.googlesource.com/c/139257 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com>
2018-10-02 22:04:45 -04:00
-> (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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& registerizable(b, t2)
&& n >= o2 + sizeof(t2)
-> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& n >= o2 + sizeof(t2)
&& n >= o3 + sizeof(t3)
-> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& alignof(t5) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& n >= o2 + sizeof(t2)
&& n >= o3 + sizeof(t3)
&& n >= o4 + sizeof(t4)
-> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& alignof(t5) <= alignof(t1)
&& alignof(t6) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& registerizable(b, t5)
&& n >= o2 + sizeof(t2)
&& n >= o3 + sizeof(t3)
&& n >= o4 + sizeof(t4)
&& n >= o5 + sizeof(t5)
-> (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> [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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& registerizable(b, t2)
&& n >= o2 + sizeof(t2)
-> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& n >= o2 + sizeof(t2)
&& n >= o3 + sizeof(t3)
-> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& alignof(t5) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& n >= o2 + sizeof(t2)
&& n >= o3 + sizeof(t3)
&& n >= o4 + sizeof(t4)
-> (Store {t2} (OffPtr <tt2> [o2] dst) d1
(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)
&& alignof(t2) <= alignof(t1)
&& alignof(t3) <= alignof(t1)
&& alignof(t4) <= alignof(t1)
&& alignof(t5) <= alignof(t1)
&& alignof(t6) <= alignof(t1)
&& registerizable(b, t2)
&& registerizable(b, t3)
&& registerizable(b, t4)
&& registerizable(b, t5)
&& n >= o2 + sizeof(t2)
&& n >= o3 + sizeof(t3)
&& n >= o4 + sizeof(t4)
&& n >= o5 + sizeof(t5)
-> (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> [o5] dst) d4
(Zero {t1} [n] dst mem)))))
cmd/compile: omit racefuncentry/exit when they are not needed When compiling with -race, we insert calls to racefuncentry, into every function. Add a rule that removes them in leaf functions, without instrumented loads/stores. Shaves ~30kb from "-race" version of go tool: file difference: go_old 15626192 go_new 15597520 [-28672 bytes] section differences: global text (code) = -24513 bytes (-0.358598%) read-only data = -5849 bytes (-0.167064%) Total difference -30362 bytes (-0.097928%) Fixes #24662 Change-Id: Ia63bf1827f4cf2c25e3e28dcd097c150994ade0a Reviewed-on: https://go-review.googlesource.com/121235 Run-TryBot: Ilya Tocar <ilya.tocar@intel.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2018-06-27 11:40:24 -05:00
(StaticCall {sym} x) && needRaceCleanup(sym,v) -> 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
// 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.
cmd/compile: only optimize chained Moves on disjoint stack mem This optimization is not sound if A, B, or C might overlap with each other. Thanks to Michael Munday for pointing this out during the review of CL 143479. This reduces the number of times this optimization triggers during make.bash from 386 to 74. This is unfortunate, but I don't see an obvious way around it, short of souping up the disjointness analysis. name old object-bytes new object-bytes delta Template 507kB ± 0% 507kB ± 0% +0.13% (p=0.008 n=5+5) Unicode 225kB ± 0% 225kB ± 0% ~ (all equal) GoTypes 1.85MB ± 0% 1.85MB ± 0% +0.02% (p=0.008 n=5+5) Flate 328kB ± 0% 328kB ± 0% ~ (all equal) GoParser 402kB ± 0% 402kB ± 0% ~ (all equal) Reflect 1.41MB ± 0% 1.41MB ± 0% ~ (all equal) Tar 457kB ± 0% 458kB ± 0% +0.20% (p=0.008 n=5+5) XML 600kB ± 0% 601kB ± 0% +0.03% (p=0.008 n=5+5) Change-Id: Ida408cb627145ba9faf473a78606f050c2f3f51c Reviewed-on: https://go-review.googlesource.com/c/145208 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Michael Munday <mike.munday@ibm.com>
2018-10-28 12:01:11 -07:00
// 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 _))
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
&& t1.(*types.Type).Compare(t2.(*types.Type)) == types.CMPeq
&& isSamePtr(tmp1, tmp2)
cmd/compile: only optimize chained Moves on disjoint stack mem This optimization is not sound if A, B, or C might overlap with each other. Thanks to Michael Munday for pointing this out during the review of CL 143479. This reduces the number of times this optimization triggers during make.bash from 386 to 74. This is unfortunate, but I don't see an obvious way around it, short of souping up the disjointness analysis. name old object-bytes new object-bytes delta Template 507kB ± 0% 507kB ± 0% +0.13% (p=0.008 n=5+5) Unicode 225kB ± 0% 225kB ± 0% ~ (all equal) GoTypes 1.85MB ± 0% 1.85MB ± 0% +0.02% (p=0.008 n=5+5) Flate 328kB ± 0% 328kB ± 0% ~ (all equal) GoParser 402kB ± 0% 402kB ± 0% ~ (all equal) Reflect 1.41MB ± 0% 1.41MB ± 0% ~ (all equal) Tar 457kB ± 0% 458kB ± 0% +0.20% (p=0.008 n=5+5) XML 600kB ± 0% 601kB ± 0% +0.03% (p=0.008 n=5+5) Change-Id: Ida408cb627145ba9faf473a78606f050c2f3f51c Reviewed-on: https://go-review.googlesource.com/c/145208 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Michael Munday <mike.munday@ibm.com>
2018-10-28 12:01:11 -07:00
&& isStackPtr(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
cmd/compile: optimize A->B->C Moves that include VarDefs We have an existing optimization that recognizes memory moves of the form A -> B -> C and converts them into A -> C, in the hopes that the store to B will be end up being dead and thus eliminated. However, when A, B, and C are large types, the front end sometimes emits VarDef ops for the moves. This change adds an optimization to match that pattern. This required changing an old compiler test. The test assumed that a temporary was required to deal with a large return value. With this optimization in place, that temporary ended up being eliminated. Triggers 649 times during 'go build -a std cmd'. Cuts 16k off cmd/go. name old object-bytes new object-bytes delta Template 507kB ± 0% 507kB ± 0% -0.15% (p=0.008 n=5+5) Unicode 225kB ± 0% 225kB ± 0% ~ (all equal) GoTypes 1.85MB ± 0% 1.85MB ± 0% ~ (all equal) Flate 328kB ± 0% 328kB ± 0% ~ (all equal) GoParser 402kB ± 0% 402kB ± 0% -0.00% (p=0.008 n=5+5) Reflect 1.41MB ± 0% 1.41MB ± 0% -0.20% (p=0.008 n=5+5) Tar 458kB ± 0% 458kB ± 0% ~ (all equal) XML 601kB ± 0% 599kB ± 0% -0.21% (p=0.008 n=5+5) Change-Id: I9b5f25c8663a0b772ad1ee51fa61f74b74d26dd3 Reviewed-on: https://go-review.googlesource.com/c/143479 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Michael Munday <mike.munday@ibm.com>
2018-10-20 19:59:27 -07:00
// Same, but for large types that require VarDefs.
(Move {t1} [s] dst tmp1 midmem:(VarDef (Move {t2} [s] tmp2 src _)))
&& t1.(*types.Type).Compare(t2.(*types.Type)) == types.CMPeq
&& isSamePtr(tmp1, tmp2)
&& isStackPtr(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
// 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
Reference in a new issue Copy permalink