go/src/cmd/compile/internal/gc/cplx.go

476 lines
7.7 KiB
Go
Raw Normal View History

// Copyright 2009 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.
package gc
import "cmd/internal/obj"
func overlap_cplx(f *Node, t *Node) bool {
// check whether f and t could be overlapping stack references.
// not exact, because it's hard to check for the stack register
// in portable code. close enough: worst case we will allocate
// an extra temporary and the registerizer will clean it up.
return f.Op == OINDREG && t.Op == OINDREG && f.Xoffset+f.Type.Width >= t.Xoffset && t.Xoffset+t.Type.Width >= f.Xoffset
}
func complexbool(op Op, nl, nr, res *Node, wantTrue bool, likely int, to *obj.Prog) {
// make both sides addable in ullman order
if nr != nil {
if nl.Ullman > nr.Ullman && !nl.Addable {
nl = CgenTemp(nl)
}
if !nr.Addable {
nr = CgenTemp(nr)
}
}
if !nl.Addable {
nl = CgenTemp(nl)
}
// Break nl and nr into real and imaginary components.
var lreal, limag, rreal, rimag Node
subnode(&lreal, &limag, nl)
subnode(&rreal, &rimag, nr)
// build tree
cmd/internal/gc, cmd/6g: generate boolean values without jumps Use SETcc instructions instead of Jcc to generate boolean values. This generates shorter, jump-free code, which may in turn enable other peephole optimizations. For example, given func f(i, j int) bool { return i == j } Before "".f t=1 size=32 value=0 args=0x18 locals=0x0 0x0000 00000 (x.go:3) TEXT "".f(SB), $0-24 0x0000 00000 (x.go:3) FUNCDATA $0, gclocals·b4c25e9b09fd0cf9bb429dcefe91c353(SB) 0x0000 00000 (x.go:3) FUNCDATA $1, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB) 0x0000 00000 (x.go:4) MOVQ "".i+8(FP), BX 0x0005 00005 (x.go:4) MOVQ "".j+16(FP), BP 0x000a 00010 (x.go:4) CMPQ BX, BP 0x000d 00013 (x.go:4) JEQ 21 0x000f 00015 (x.go:4) MOVB $0, "".~r2+24(FP) 0x0014 00020 (x.go:4) RET 0x0015 00021 (x.go:4) MOVB $1, "".~r2+24(FP) 0x001a 00026 (x.go:4) JMP 20 After "".f t=1 size=32 value=0 args=0x18 locals=0x0 0x0000 00000 (x.go:3) TEXT "".f(SB), $0-24 0x0000 00000 (x.go:3) FUNCDATA $0, gclocals·b4c25e9b09fd0cf9bb429dcefe91c353(SB) 0x0000 00000 (x.go:3) FUNCDATA $1, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB) 0x0000 00000 (x.go:4) MOVQ "".i+8(FP), BX 0x0005 00005 (x.go:4) MOVQ "".j+16(FP), BP 0x000a 00010 (x.go:4) CMPQ BX, BP 0x000d 00013 (x.go:4) SETEQ "".~r2+24(FP) 0x0012 00018 (x.go:4) RET regexp benchmarks, best of 12 runs: benchmark old ns/op new ns/op delta BenchmarkNotOnePassShortB 782 733 -6.27% BenchmarkLiteral 180 171 -5.00% BenchmarkNotLiteral 2855 2721 -4.69% BenchmarkMatchHard_32 2672 2557 -4.30% BenchmarkMatchHard_1K 80182 76732 -4.30% BenchmarkMatchEasy1_32M 76440180 73304748 -4.10% BenchmarkMatchEasy1_32K 68798 66350 -3.56% BenchmarkAnchoredLongMatch 482 465 -3.53% BenchmarkMatchEasy1_1M 2373042 2292692 -3.39% BenchmarkReplaceAll 2776 2690 -3.10% BenchmarkNotOnePassShortA 1397 1360 -2.65% BenchmarkMatchClass_InRange 3842 3742 -2.60% BenchmarkMatchEasy0_32 125 122 -2.40% BenchmarkMatchEasy0_32K 11414 11164 -2.19% BenchmarkMatchEasy0_1K 668 654 -2.10% BenchmarkAnchoredShortMatch 260 255 -1.92% BenchmarkAnchoredLiteralShortNonMatch 164 161 -1.83% BenchmarkOnePassShortB 623 612 -1.77% BenchmarkOnePassShortA 801 788 -1.62% BenchmarkMatchClass 4094 4033 -1.49% BenchmarkMatchEasy0_32M 14078800 13890704 -1.34% BenchmarkMatchHard_32K 4095844 4045820 -1.22% BenchmarkMatchEasy1_1K 1663 1643 -1.20% BenchmarkMatchHard_1M 131261708 129708215 -1.18% BenchmarkMatchHard_32M 4210112412 4169292003 -0.97% BenchmarkMatchMedium_32K 2460752 2438611 -0.90% BenchmarkMatchEasy0_1M 422914 419672 -0.77% BenchmarkMatchMedium_1M 78581121 78040160 -0.69% BenchmarkMatchMedium_32M 2515287278 2498464906 -0.67% BenchmarkMatchMedium_32 1754 1746 -0.46% BenchmarkMatchMedium_1K 52105 52106 +0.00% BenchmarkAnchoredLiteralLongNonMatch 185 185 +0.00% BenchmarkMatchEasy1_32 107 107 +0.00% BenchmarkOnePassLongNotPrefix 505 505 +0.00% BenchmarkOnePassLongPrefix 147 147 +0.00% The godoc binary is ~0.12% smaller after this CL. Updates #5729. toolstash -cmp passes for all architectures other than amd64 and amd64p32. Other architectures can be done in follow-up CLs. Change-Id: I0e167e259274b722958567fc0af83a17ca002da7 Reviewed-on: https://go-review.googlesource.com/2284 Reviewed-by: Russ Cox <rsc@golang.org>
2015-04-08 09:54:15 -07:00
// if branching:
// real(l) == real(r) && imag(l) == imag(r)
// if generating a value, use a branch-free version:
// real(l) == real(r) & imag(l) == imag(r)
realeq := Node{
Op: OEQ,
Left: &lreal,
Right: &rreal,
Type: Types[TBOOL],
}
imageq := Node{
Op: OEQ,
Left: &limag,
Right: &rimag,
Type: Types[TBOOL],
}
and := Node{
Op: OANDAND,
Left: &realeq,
Right: &imageq,
Type: Types[TBOOL],
}
cmd/internal/gc, cmd/6g: generate boolean values without jumps Use SETcc instructions instead of Jcc to generate boolean values. This generates shorter, jump-free code, which may in turn enable other peephole optimizations. For example, given func f(i, j int) bool { return i == j } Before "".f t=1 size=32 value=0 args=0x18 locals=0x0 0x0000 00000 (x.go:3) TEXT "".f(SB), $0-24 0x0000 00000 (x.go:3) FUNCDATA $0, gclocals·b4c25e9b09fd0cf9bb429dcefe91c353(SB) 0x0000 00000 (x.go:3) FUNCDATA $1, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB) 0x0000 00000 (x.go:4) MOVQ "".i+8(FP), BX 0x0005 00005 (x.go:4) MOVQ "".j+16(FP), BP 0x000a 00010 (x.go:4) CMPQ BX, BP 0x000d 00013 (x.go:4) JEQ 21 0x000f 00015 (x.go:4) MOVB $0, "".~r2+24(FP) 0x0014 00020 (x.go:4) RET 0x0015 00021 (x.go:4) MOVB $1, "".~r2+24(FP) 0x001a 00026 (x.go:4) JMP 20 After "".f t=1 size=32 value=0 args=0x18 locals=0x0 0x0000 00000 (x.go:3) TEXT "".f(SB), $0-24 0x0000 00000 (x.go:3) FUNCDATA $0, gclocals·b4c25e9b09fd0cf9bb429dcefe91c353(SB) 0x0000 00000 (x.go:3) FUNCDATA $1, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB) 0x0000 00000 (x.go:4) MOVQ "".i+8(FP), BX 0x0005 00005 (x.go:4) MOVQ "".j+16(FP), BP 0x000a 00010 (x.go:4) CMPQ BX, BP 0x000d 00013 (x.go:4) SETEQ "".~r2+24(FP) 0x0012 00018 (x.go:4) RET regexp benchmarks, best of 12 runs: benchmark old ns/op new ns/op delta BenchmarkNotOnePassShortB 782 733 -6.27% BenchmarkLiteral 180 171 -5.00% BenchmarkNotLiteral 2855 2721 -4.69% BenchmarkMatchHard_32 2672 2557 -4.30% BenchmarkMatchHard_1K 80182 76732 -4.30% BenchmarkMatchEasy1_32M 76440180 73304748 -4.10% BenchmarkMatchEasy1_32K 68798 66350 -3.56% BenchmarkAnchoredLongMatch 482 465 -3.53% BenchmarkMatchEasy1_1M 2373042 2292692 -3.39% BenchmarkReplaceAll 2776 2690 -3.10% BenchmarkNotOnePassShortA 1397 1360 -2.65% BenchmarkMatchClass_InRange 3842 3742 -2.60% BenchmarkMatchEasy0_32 125 122 -2.40% BenchmarkMatchEasy0_32K 11414 11164 -2.19% BenchmarkMatchEasy0_1K 668 654 -2.10% BenchmarkAnchoredShortMatch 260 255 -1.92% BenchmarkAnchoredLiteralShortNonMatch 164 161 -1.83% BenchmarkOnePassShortB 623 612 -1.77% BenchmarkOnePassShortA 801 788 -1.62% BenchmarkMatchClass 4094 4033 -1.49% BenchmarkMatchEasy0_32M 14078800 13890704 -1.34% BenchmarkMatchHard_32K 4095844 4045820 -1.22% BenchmarkMatchEasy1_1K 1663 1643 -1.20% BenchmarkMatchHard_1M 131261708 129708215 -1.18% BenchmarkMatchHard_32M 4210112412 4169292003 -0.97% BenchmarkMatchMedium_32K 2460752 2438611 -0.90% BenchmarkMatchEasy0_1M 422914 419672 -0.77% BenchmarkMatchMedium_1M 78581121 78040160 -0.69% BenchmarkMatchMedium_32M 2515287278 2498464906 -0.67% BenchmarkMatchMedium_32 1754 1746 -0.46% BenchmarkMatchMedium_1K 52105 52106 +0.00% BenchmarkAnchoredLiteralLongNonMatch 185 185 +0.00% BenchmarkMatchEasy1_32 107 107 +0.00% BenchmarkOnePassLongNotPrefix 505 505 +0.00% BenchmarkOnePassLongPrefix 147 147 +0.00% The godoc binary is ~0.12% smaller after this CL. Updates #5729. toolstash -cmp passes for all architectures other than amd64 and amd64p32. Other architectures can be done in follow-up CLs. Change-Id: I0e167e259274b722958567fc0af83a17ca002da7 Reviewed-on: https://go-review.googlesource.com/2284 Reviewed-by: Russ Cox <rsc@golang.org>
2015-04-08 09:54:15 -07:00
if res != nil {
// generating a value
and.Op = OAND
if op == ONE {
and.Op = OOR
realeq.Op = ONE
imageq.Op = ONE
}
Bvgen(&and, res, true)
return
}
// generating a branch
if op == ONE {
wantTrue = !wantTrue
}
Bgen(&and, wantTrue, likely, to)
}
// break addable nc-complex into nr-real and ni-imaginary
func subnode(nr *Node, ni *Node, nc *Node) {
if !nc.Addable {
Fatalf("subnode not addable")
}
tc := Simsimtype(nc.Type)
tc = cplxsubtype(tc)
t := Types[tc]
if nc.Op == OLITERAL {
u := nc.Val().U.(*Mpcplx)
nodfconst(nr, t, &u.Real)
nodfconst(ni, t, &u.Imag)
return
}
*nr = *nc
nr.Type = t
*ni = *nc
ni.Type = t
ni.Xoffset += t.Width
}
// generate code res = -nl
func minus(nl *Node, res *Node) {
var ra Node
ra.Op = OMINUS
ra.Left = nl
ra.Type = nl.Type
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(&ra, res)
}
// build and execute tree
// real(res) = -real(nl)
// imag(res) = -imag(nl)
func complexminus(nl *Node, res *Node) {
var n1 Node
var n2 Node
var n5 Node
var n6 Node
subnode(&n1, &n2, nl)
subnode(&n5, &n6, res)
minus(&n1, &n5)
minus(&n2, &n6)
}
// build and execute tree
// real(res) = real(nl) op real(nr)
// imag(res) = imag(nl) op imag(nr)
func complexadd(op Op, nl *Node, nr *Node, res *Node) {
var n1 Node
var n2 Node
var n3 Node
var n4 Node
var n5 Node
var n6 Node
subnode(&n1, &n2, nl)
subnode(&n3, &n4, nr)
subnode(&n5, &n6, res)
var ra Node
ra.Op = op
ra.Left = &n1
ra.Right = &n3
ra.Type = n1.Type
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(&ra, &n5)
ra = Node{}
ra.Op = op
ra.Left = &n2
ra.Right = &n4
ra.Type = n2.Type
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(&ra, &n6)
}
// build and execute tree
// tmp = real(nl)*real(nr) - imag(nl)*imag(nr)
// imag(res) = real(nl)*imag(nr) + imag(nl)*real(nr)
// real(res) = tmp
func complexmul(nl *Node, nr *Node, res *Node) {
var n1 Node
var n2 Node
var n3 Node
var n4 Node
var n5 Node
var n6 Node
var tmp Node
subnode(&n1, &n2, nl)
subnode(&n3, &n4, nr)
subnode(&n5, &n6, res)
Tempname(&tmp, n5.Type)
// real part -> tmp
var rm1 Node
rm1.Op = OMUL
rm1.Left = &n1
rm1.Right = &n3
rm1.Type = n1.Type
var rm2 Node
rm2.Op = OMUL
rm2.Left = &n2
rm2.Right = &n4
rm2.Type = n2.Type
var ra Node
ra.Op = OSUB
ra.Left = &rm1
ra.Right = &rm2
ra.Type = rm1.Type
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(&ra, &tmp)
// imag part
rm1 = Node{}
rm1.Op = OMUL
rm1.Left = &n1
rm1.Right = &n4
rm1.Type = n1.Type
rm2 = Node{}
rm2.Op = OMUL
rm2.Left = &n2
rm2.Right = &n3
rm2.Type = n2.Type
ra = Node{}
ra.Op = OADD
ra.Left = &rm1
ra.Right = &rm2
ra.Type = rm1.Type
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(&ra, &n6)
// tmp ->real part
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(&tmp, &n5)
}
func nodfconst(n *Node, t *Type, fval *Mpflt) {
*n = Node{}
n.Op = OLITERAL
n.Addable = true
ullmancalc(n)
n.SetVal(Val{fval})
n.Type = t
if !t.IsFloat() {
Fatalf("nodfconst: bad type %v", t)
}
}
func Complexop(n *Node, res *Node) bool {
if n != nil && n.Type != nil {
if n.Type.IsComplex() {
goto maybe
}
}
if res != nil && res.Type != nil {
if res.Type.IsComplex() {
goto maybe
}
}
if n.Op == OREAL || n.Op == OIMAG {
//dump("\ncomplex-yes", n);
return true
}
//dump("\ncomplex-no", n);
return false
maybe:
switch n.Op {
case OCONV, // implemented ops
OADD,
OSUB,
OMUL,
OMINUS,
OCOMPLEX,
OREAL,
OIMAG:
//dump("\ncomplex-yes", n);
return true
case ODOT,
ODOTPTR,
OINDEX,
OIND,
ONAME:
//dump("\ncomplex-yes", n);
return true
}
//dump("\ncomplex-no", n);
return false
}
func Complexmove(f *Node, t *Node) {
if Debug['g'] != 0 {
Dump("\ncomplexmove-f", f)
Dump("complexmove-t", t)
}
if !t.Addable {
Fatalf("complexmove: to not addable")
}
ft := Simsimtype(f.Type)
tt := Simsimtype(t.Type)
// complex to complex move/convert.
// make f addable.
// also use temporary if possible stack overlap.
if (ft == TCOMPLEX64 || ft == TCOMPLEX128) && (tt == TCOMPLEX64 || tt == TCOMPLEX128) {
if !f.Addable || overlap_cplx(f, t) {
var tmp Node
Tempname(&tmp, f.Type)
Complexmove(f, &tmp)
f = &tmp
}
var n1 Node
var n2 Node
subnode(&n1, &n2, f)
var n4 Node
var n3 Node
subnode(&n3, &n4, t)
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(&n1, &n3)
Cgen(&n2, &n4)
} else {
Fatalf("complexmove: unknown conversion: %v -> %v\n", f.Type, t.Type)
}
}
func Complexgen(n *Node, res *Node) {
if Debug['g'] != 0 {
Dump("\ncomplexgen-n", n)
Dump("complexgen-res", res)
}
for n.Op == OCONVNOP {
n = n.Left
}
// pick off float/complex opcodes
switch n.Op {
case OCOMPLEX:
if res.Addable {
var n1 Node
var n2 Node
subnode(&n1, &n2, res)
var tmp Node
Tempname(&tmp, n1.Type)
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(n.Left, &tmp)
Cgen(n.Right, &n2)
Cgen(&tmp, &n1)
return
}
case OREAL, OIMAG:
nl := n.Left
if !nl.Addable {
var tmp Node
Tempname(&tmp, nl.Type)
Complexgen(nl, &tmp)
nl = &tmp
}
var n1 Node
var n2 Node
subnode(&n1, &n2, nl)
if n.Op == OREAL {
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(&n1, res)
return
}
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(&n2, res)
return
}
// perform conversion from n to res
tl := Simsimtype(res.Type)
tl = cplxsubtype(tl)
tr := Simsimtype(n.Type)
tr = cplxsubtype(tr)
if tl != tr {
if !n.Addable {
var n1 Node
Tempname(&n1, n.Type)
Complexmove(n, &n1)
n = &n1
}
Complexmove(n, res)
return
}
if !res.Addable {
var n1 Node
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Igen(res, &n1, nil)
Cgen(n, &n1)
Regfree(&n1)
return
}
if n.Addable {
Complexmove(n, res)
return
}
switch n.Op {
default:
Dump("complexgen: unknown op", n)
Fatalf("complexgen: unknown op %v", n.Op)
case ODOT,
ODOTPTR,
OINDEX,
OIND,
cmd/compile: fix liveness computation for heap-escaped parameters The liveness computation of parameters generally was never correct, but forcing all parameters to be live throughout the function covered up that problem. The new SSA back end is too clever: even though it currently keeps the parameter values live throughout the function, it may find optimizations that mean the current values are not written back to the original parameter stack slots immediately or ever (for example if a parameter is set to nil, SSA constant propagation may replace all later uses of the parameter with a constant nil, eliminating the need to write the nil value back to the stack slot), so the liveness code must now track the actual operations on the stack slots, exposing these problems. One small problem in the handling of arguments is that nodarg can return ONAME PPARAM nodes with adjusted offsets, so that there are actually multiple *Node pointers for the same parameter in the instruction stream. This might be possible to correct, but not in this CL. For now, we fix this by using n.Orig instead of n when considering PPARAM and PPARAMOUT nodes. The major problem in the handling of arguments is general confusion in the liveness code about the meaning of PPARAM|PHEAP and PPARAMOUT|PHEAP nodes, especially as contrasted with PAUTO|PHEAP. The difference between these two is that when a local variable "moves" to the heap, it's really just allocated there to start with; in contrast, when an argument moves to the heap, the actual data has to be copied there from the stack at the beginning of the function, and when a result "moves" to the heap the value in the heap has to be copied back to the stack when the function returns This general confusion is also present in the SSA back end. The PHEAP bit worked decently when I first introduced it 7 years ago (!) in 391425ae. The back end did nothing sophisticated, and in particular there was no analysis at all: no escape analysis, no liveness analysis, and certainly no SSA back end. But the complications caused in the various downstream consumers suggest that this should be a detail kept mainly in the front end. This CL therefore eliminates both the PHEAP bit and even the idea of "heap variables" from the back ends. First, it replaces the PPARAM|PHEAP, PPARAMOUT|PHEAP, and PAUTO|PHEAP variable classes with the single PAUTOHEAP, a pseudo-class indicating a variable maintained on the heap and available by indirecting a local variable kept on the stack (a plain PAUTO). Second, walkexpr replaces all references to PAUTOHEAP variables with indirections of the corresponding PAUTO variable. The back ends and the liveness code now just see plain indirected variables. This may actually produce better code, but the real goal here is to eliminate these little-used and somewhat suspect code paths in the back end analyses. The OPARAM node type goes away too. A followup CL will do the same to PPARAMREF. I'm not sure that the back ends (SSA in particular) are handling those right either, and with the framework established in this CL that change is trivial and the result clearly more correct. Fixes #15747. Change-Id: I2770b1ce3cbc93981bfc7166be66a9da12013d74 Reviewed-on: https://go-review.googlesource.com/23393 Reviewed-by: Keith Randall <khr@golang.org> Run-TryBot: Russ Cox <rsc@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-05-25 01:33:24 -04:00
ONAME, // PPARAMREF var
OCALLFUNC,
OCALLMETH,
OCALLINTER:
var n1 Node
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Igen(n, &n1, res)
Complexmove(&n1, res)
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Regfree(&n1)
return
case OCONV,
OADD,
OSUB,
OMUL,
OMINUS,
OCOMPLEX,
OREAL,
OIMAG:
break
}
nl := n.Left
if nl == nil {
return
}
nr := n.Right
// make both sides addable in ullman order
var tnl Node
if nr != nil {
if nl.Ullman > nr.Ullman && !nl.Addable {
Tempname(&tnl, nl.Type)
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(nl, &tnl)
nl = &tnl
}
if !nr.Addable {
var tnr Node
Tempname(&tnr, nr.Type)
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(nr, &tnr)
nr = &tnr
}
}
if !nl.Addable {
Tempname(&tnl, nl.Type)
cmd/internal/gc: move cgen, regalloc, et al to portable code This CL moves the bulk of the code that has been copy-and-pasted since the initial 386 port back into a shared place, cutting 5 copies to 1. The motivation here is not cleanup per se but instead to reduce the cost of introducing changes in shared concepts like regalloc or general expression evaluation. For example, a change after this one will implement x.(*T) without a call into the runtime. This CL makes that followup work 5x easier. The single copy still has more special cases for architecture details than I'd like, but having them called out explicitly like this at least opens the door to generalizing the conditions and smoothing out the distinctions in the future. This is a LARGE CL. I started by trying to pull in one function at a time in a sequence of CLs and it became clear that everything was so interrelated that it had to be moved as a whole. Apologies for the size. It is not clear how many more releases this code will matter for; eventually it will be replaced by Keith's SSA work. But as noted above, the deduplication was necessary to reduce the cost of working on the current code while we have it. Passes tests on amd64, 386, arm, and ppc64le. Can build arm64 binaries but not tested there. Being able to build binaries means it is probably very close. Change-Id: I735977f04c0614f80215fb12966dfe9bbd1f5861 Reviewed-on: https://go-review.googlesource.com/7853 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2015-03-18 17:26:36 -04:00
Cgen(nl, &tnl)
nl = &tnl
}
switch n.Op {
default:
Fatalf("complexgen: unknown op %v", n.Op)
case OCONV:
Complexmove(nl, res)
case OMINUS:
complexminus(nl, res)
case OADD, OSUB:
complexadd(n.Op, nl, nr, res)
case OMUL:
complexmul(nl, nr, res)
}
}