Stowage/go
2
0
Fork 0
mirror of https://github.com/golang/go.git synced 2025-12-08 06:10:04 +00:00
Code Issues Projects Releases Packages Wiki Activity
go/src/cmd/compile/internal/ssa/rewritedec.go

508 lines
12 KiB
Go
Raw Normal View History

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
// autogenerated from gen/dec.rules: do not edit!
// generated with: cd gen; go run *.go
package ssa
import "math"
var _ = math.MinInt8 // in case not otherwise used
func rewriteValuedec(v *Value, config *Config) bool {
switch v.Op {
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
case OpComplexImag:
return rewriteValuedec_OpComplexImag(v, config)
case OpComplexReal:
return rewriteValuedec_OpComplexReal(v, config)
case OpIData:
return rewriteValuedec_OpIData(v, config)
case OpITab:
return rewriteValuedec_OpITab(v, config)
case OpLoad:
return rewriteValuedec_OpLoad(v, config)
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
case OpSliceCap:
return rewriteValuedec_OpSliceCap(v, config)
case OpSliceLen:
return rewriteValuedec_OpSliceLen(v, config)
case OpSlicePtr:
return rewriteValuedec_OpSlicePtr(v, config)
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
case OpStore:
return rewriteValuedec_OpStore(v, config)
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
case OpStringLen:
return rewriteValuedec_OpStringLen(v, config)
case OpStringPtr:
return rewriteValuedec_OpStringPtr(v, config)
}
return false
}
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
func rewriteValuedec_OpComplexImag(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (ComplexImag (ComplexMake _ imag ))
// cond:
// result: imag
for {
v_0 := v.Args[0]
if v_0.Op != OpComplexMake {
break
}
imag := v_0.Args[1]
v.reset(OpCopy)
v.Type = imag.Type
v.AddArg(imag)
return true
}
return false
}
func rewriteValuedec_OpComplexReal(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (ComplexReal (ComplexMake real _ ))
// cond:
// result: real
for {
v_0 := v.Args[0]
if v_0.Op != OpComplexMake {
break
}
real := v_0.Args[0]
v.reset(OpCopy)
v.Type = real.Type
v.AddArg(real)
return true
}
return false
}
func rewriteValuedec_OpIData(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (IData (IMake _ data))
// cond:
// result: data
for {
v_0 := v.Args[0]
if v_0.Op != OpIMake {
break
}
data := v_0.Args[1]
v.reset(OpCopy)
v.Type = data.Type
v.AddArg(data)
return true
}
return false
}
func rewriteValuedec_OpITab(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (ITab (IMake itab _))
// cond:
// result: itab
for {
v_0 := v.Args[0]
if v_0.Op != OpIMake {
break
}
itab := v_0.Args[0]
v.reset(OpCopy)
v.Type = itab.Type
v.AddArg(itab)
return true
}
return false
}
func rewriteValuedec_OpLoad(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (Load <t> ptr mem)
// cond: t.IsComplex() && t.Size() == 8
// result: (ComplexMake (Load <config.fe.TypeFloat32()> ptr mem) (Load <config.fe.TypeFloat32()> (OffPtr <config.fe.TypeFloat32().PtrTo()> [4] ptr) mem) )
for {
t := v.Type
ptr := v.Args[0]
mem := v.Args[1]
if !(t.IsComplex() && t.Size() == 8) {
break
}
v.reset(OpComplexMake)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeFloat32())
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
v0.AddArg(ptr)
v0.AddArg(mem)
v.AddArg(v0)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeFloat32())
v2 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeFloat32().PtrTo())
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
v2.AuxInt = 4
v2.AddArg(ptr)
v1.AddArg(v2)
v1.AddArg(mem)
v.AddArg(v1)
return true
}
// match: (Load <t> ptr mem)
// cond: t.IsComplex() && t.Size() == 16
// result: (ComplexMake (Load <config.fe.TypeFloat64()> ptr mem) (Load <config.fe.TypeFloat64()> (OffPtr <config.fe.TypeFloat64().PtrTo()> [8] ptr) mem) )
for {
t := v.Type
ptr := v.Args[0]
mem := v.Args[1]
if !(t.IsComplex() && t.Size() == 16) {
break
}
v.reset(OpComplexMake)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeFloat64())
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
v0.AddArg(ptr)
v0.AddArg(mem)
v.AddArg(v0)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeFloat64())
v2 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeFloat64().PtrTo())
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
v2.AuxInt = 8
v2.AddArg(ptr)
v1.AddArg(v2)
v1.AddArg(mem)
v.AddArg(v1)
return true
}
// match: (Load <t> ptr mem)
// cond: t.IsString()
// result: (StringMake (Load <config.fe.TypeBytePtr()> ptr mem) (Load <config.fe.TypeInt()> (OffPtr <config.fe.TypeInt().PtrTo()> [config.PtrSize] ptr) mem))
for {
t := v.Type
ptr := v.Args[0]
mem := v.Args[1]
if !(t.IsString()) {
break
}
v.reset(OpStringMake)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeBytePtr())
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
v0.AddArg(ptr)
v0.AddArg(mem)
v.AddArg(v0)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeInt())
v2 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeInt().PtrTo())
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
v2.AuxInt = config.PtrSize
v2.AddArg(ptr)
v1.AddArg(v2)
v1.AddArg(mem)
v.AddArg(v1)
return true
}
// match: (Load <t> ptr mem)
// cond: t.IsSlice()
// result: (SliceMake (Load <t.ElemType().PtrTo()> ptr mem) (Load <config.fe.TypeInt()> (OffPtr <config.fe.TypeInt().PtrTo()> [config.PtrSize] ptr) mem) (Load <config.fe.TypeInt()> (OffPtr <config.fe.TypeInt().PtrTo()> [2*config.PtrSize] ptr) mem))
for {
t := v.Type
ptr := v.Args[0]
mem := v.Args[1]
if !(t.IsSlice()) {
break
}
v.reset(OpSliceMake)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpLoad, t.ElemType().PtrTo())
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
v0.AddArg(ptr)
v0.AddArg(mem)
v.AddArg(v0)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeInt())
v2 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeInt().PtrTo())
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
v2.AuxInt = config.PtrSize
v2.AddArg(ptr)
v1.AddArg(v2)
v1.AddArg(mem)
v.AddArg(v1)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v3 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeInt())
v4 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeInt().PtrTo())
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
v4.AuxInt = 2 * config.PtrSize
v4.AddArg(ptr)
v3.AddArg(v4)
v3.AddArg(mem)
v.AddArg(v3)
return true
}
// match: (Load <t> ptr mem)
// cond: t.IsInterface()
// result: (IMake (Load <config.fe.TypeBytePtr()> ptr mem) (Load <config.fe.TypeBytePtr()> (OffPtr <config.fe.TypeBytePtr().PtrTo()> [config.PtrSize] ptr) mem))
for {
t := v.Type
ptr := v.Args[0]
mem := v.Args[1]
if !(t.IsInterface()) {
break
}
v.reset(OpIMake)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeBytePtr())
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
v0.AddArg(ptr)
v0.AddArg(mem)
v.AddArg(v0)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpLoad, config.fe.TypeBytePtr())
v2 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeBytePtr().PtrTo())
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
v2.AuxInt = config.PtrSize
v2.AddArg(ptr)
v1.AddArg(v2)
v1.AddArg(mem)
v.AddArg(v1)
return true
}
return false
}
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
func rewriteValuedec_OpSliceCap(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (SliceCap (SliceMake _ _ cap))
// cond:
// result: cap
for {
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
v_0 := v.Args[0]
if v_0.Op != OpSliceMake {
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
break
}
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
cap := v_0.Args[2]
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
v.reset(OpCopy)
v.Type = cap.Type
v.AddArg(cap)
return true
}
return false
}
func rewriteValuedec_OpSliceLen(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (SliceLen (SliceMake _ len _))
// cond:
// result: len
for {
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
v_0 := v.Args[0]
if v_0.Op != OpSliceMake {
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
break
}
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
len := v_0.Args[1]
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
v.reset(OpCopy)
v.Type = len.Type
v.AddArg(len)
return true
}
return false
}
func rewriteValuedec_OpSlicePtr(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (SlicePtr (SliceMake ptr _ _ ))
// cond:
// result: ptr
for {
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
v_0 := v.Args[0]
if v_0.Op != OpSliceMake {
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
break
}
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
ptr := v_0.Args[0]
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
v.reset(OpCopy)
v.Type = ptr.Type
v.AddArg(ptr)
return true
}
return false
}
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
func rewriteValuedec_OpStore(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (Store [8] dst (ComplexMake real imag) mem)
// cond:
// result: (Store [4] (OffPtr <config.fe.TypeFloat32().PtrTo()> [4] dst) imag (Store [4] dst real mem))
for {
if v.AuxInt != 8 {
break
}
dst := v.Args[0]
v_1 := v.Args[1]
if v_1.Op != OpComplexMake {
break
}
real := v_1.Args[0]
imag := v_1.Args[1]
mem := v.Args[2]
v.reset(OpStore)
v.AuxInt = 4
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeFloat32().PtrTo())
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
v0.AuxInt = 4
v0.AddArg(dst)
v.AddArg(v0)
v.AddArg(imag)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpStore, TypeMem)
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
v1.AuxInt = 4
v1.AddArg(dst)
v1.AddArg(real)
v1.AddArg(mem)
v.AddArg(v1)
return true
}
// match: (Store [16] dst (ComplexMake real imag) mem)
// cond:
// result: (Store [8] (OffPtr <config.fe.TypeFloat64().PtrTo()> [8] dst) imag (Store [8] dst real mem))
for {
if v.AuxInt != 16 {
break
}
dst := v.Args[0]
v_1 := v.Args[1]
if v_1.Op != OpComplexMake {
break
}
real := v_1.Args[0]
imag := v_1.Args[1]
mem := v.Args[2]
v.reset(OpStore)
v.AuxInt = 8
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeFloat64().PtrTo())
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
v0.AuxInt = 8
v0.AddArg(dst)
v.AddArg(v0)
v.AddArg(imag)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpStore, TypeMem)
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
v1.AuxInt = 8
v1.AddArg(dst)
v1.AddArg(real)
v1.AddArg(mem)
v.AddArg(v1)
return true
}
// match: (Store [2*config.PtrSize] dst (StringMake ptr len) mem)
// cond:
// result: (Store [config.PtrSize] (OffPtr <config.fe.TypeInt().PtrTo()> [config.PtrSize] dst) len (Store [config.PtrSize] dst ptr mem))
for {
if v.AuxInt != 2*config.PtrSize {
break
}
dst := v.Args[0]
v_1 := v.Args[1]
if v_1.Op != OpStringMake {
break
}
ptr := v_1.Args[0]
len := v_1.Args[1]
mem := v.Args[2]
v.reset(OpStore)
v.AuxInt = config.PtrSize
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeInt().PtrTo())
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
v0.AuxInt = config.PtrSize
v0.AddArg(dst)
v.AddArg(v0)
v.AddArg(len)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpStore, TypeMem)
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
v1.AuxInt = config.PtrSize
v1.AddArg(dst)
v1.AddArg(ptr)
v1.AddArg(mem)
v.AddArg(v1)
return true
}
// match: (Store [3*config.PtrSize] dst (SliceMake ptr len cap) mem)
// cond:
// result: (Store [config.PtrSize] (OffPtr <config.fe.TypeInt().PtrTo()> [2*config.PtrSize] dst) cap (Store [config.PtrSize] (OffPtr <config.fe.TypeInt().PtrTo()> [config.PtrSize] dst) len (Store [config.PtrSize] dst ptr mem)))
for {
if v.AuxInt != 3*config.PtrSize {
break
}
dst := v.Args[0]
v_1 := v.Args[1]
if v_1.Op != OpSliceMake {
break
}
ptr := v_1.Args[0]
len := v_1.Args[1]
cap := v_1.Args[2]
mem := v.Args[2]
v.reset(OpStore)
v.AuxInt = config.PtrSize
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeInt().PtrTo())
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
v0.AuxInt = 2 * config.PtrSize
v0.AddArg(dst)
v.AddArg(v0)
v.AddArg(cap)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpStore, TypeMem)
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
v1.AuxInt = config.PtrSize
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v2 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeInt().PtrTo())
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
v2.AuxInt = config.PtrSize
v2.AddArg(dst)
v1.AddArg(v2)
v1.AddArg(len)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v3 := b.NewValue0(v.Pos, OpStore, TypeMem)
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
v3.AuxInt = config.PtrSize
v3.AddArg(dst)
v3.AddArg(ptr)
v3.AddArg(mem)
v1.AddArg(v3)
v.AddArg(v1)
return true
}
// match: (Store [2*config.PtrSize] dst (IMake itab data) mem)
// cond:
// result: (Store [config.PtrSize] (OffPtr <config.fe.TypeBytePtr().PtrTo()> [config.PtrSize] dst) data (Store [config.PtrSize] dst itab mem))
for {
if v.AuxInt != 2*config.PtrSize {
break
}
dst := v.Args[0]
v_1 := v.Args[1]
if v_1.Op != OpIMake {
break
}
itab := v_1.Args[0]
data := v_1.Args[1]
mem := v.Args[2]
v.reset(OpStore)
v.AuxInt = config.PtrSize
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v0 := b.NewValue0(v.Pos, OpOffPtr, config.fe.TypeBytePtr().PtrTo())
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
v0.AuxInt = config.PtrSize
v0.AddArg(dst)
v.AddArg(v0)
v.AddArg(data)
[dev.inline] cmd/compile/internal/ssa: rename various fields from Line to Pos This is a mostly mechanical rename followed by manual fixes where necessary. Change-Id: Ie5c670b133db978f15dc03e50dc2da0c80fc8842 Reviewed-on: https://go-review.googlesource.com/34137 Reviewed-by: David Lazar <lazard@golang.org>
2016-12-07 18:14:35 -08:00
v1 := b.NewValue0(v.Pos, OpStore, TypeMem)
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
v1.AuxInt = config.PtrSize
v1.AddArg(dst)
v1.AddArg(itab)
v1.AddArg(mem)
v.AddArg(v1)
return true
}
return false
}
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
func rewriteValuedec_OpStringLen(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (StringLen (StringMake _ len))
// cond:
// result: len
for {
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
v_0 := v.Args[0]
if v_0.Op != OpStringMake {
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
break
}
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
len := v_0.Args[1]
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
v.reset(OpCopy)
v.Type = len.Type
v.AddArg(len)
return true
}
return false
}
func rewriteValuedec_OpStringPtr(v *Value, config *Config) bool {
b := v.Block
_ = b
// match: (StringPtr (StringMake ptr _))
// cond:
// result: ptr
for {
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
v_0 := v.Args[0]
if v_0.Op != OpStringMake {
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
break
}
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
ptr := v_0.Args[0]
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
v.reset(OpCopy)
v.Type = ptr.Type
v.AddArg(ptr)
return true
}
return false
}
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
func rewriteBlockdec(b *Block, config *Config) bool {
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
switch b.Kind {
}
return false
}
Reference in a new issue Copy permalink