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/5g/ggen.go

497 lines
11 KiB
Go
Raw Normal View History

[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
// 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 main
import (
cmd/5g etc: tweak import blocks Remove blank lines and merge lone imports into blocks. Change-Id: Ib46dad584456909969f1ba3a2c7f5667abf336ae Reviewed-on: https://go-review.googlesource.com/6792 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> Reviewed-by: Rob Pike <r@golang.org>
2015-03-03 22:20:44 -05:00
"cmd/internal/gc"
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
"cmd/internal/obj"
"cmd/internal/obj/arm"
)
func defframe(ptxt *obj.Prog) {
var n *gc.Node
// fill in argument size, stack size
ptxt.To.Type = obj.TYPE_TEXTSIZE
cmd/internal/obj: replace Addr.U struct {...} with Val interface{} An interface{} is more in the spirit of the original union. By my calculations, on 64-bit systems this reduces Addr from 120 to 80 bytes, and Prog from 592 to 424 bytes. Change-Id: I0d7b0981513c2a3c94c9ac76bb4f8816485b5a3c Reviewed-on: https://go-review.googlesource.com/7744 Reviewed-by: Rob Pike <r@golang.org>
2015-03-16 15:54:44 -04:00
ptxt.To.Val = int32(gc.Rnd(gc.Curfn.Type.Argwid, int64(gc.Widthptr)))
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
frame := uint32(gc.Rnd(gc.Stksize+gc.Maxarg, int64(gc.Widthreg)))
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
ptxt.To.Offset = int64(frame)
// insert code to contain ambiguously live variables
// so that garbage collector only sees initialized values
// when it looks for pointers.
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
p := ptxt
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
hi := int64(0)
lo := hi
r0 := uint32(0)
cmd/internal/gc: unembed Node.Func This is a follow-up to CL 7360. It was generated with eg and gofmt -r. The only manual changes are the unembedding in syntax.go and backporting changes from y.go to go.y. Passes toolstash -cmp. Change-Id: I3d6d06ecb659809a4bc8592395d5b9a18967218e Reviewed-on: https://go-review.googlesource.com/8053 Reviewed-by: Russ Cox <rsc@golang.org> Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
2015-03-25 19:33:01 -07:00
for l := gc.Curfn.Func.Dcl; l != nil; l = l.Next {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
n = l.N
cmd/internal/gc: make Node.Needzero a bool Node.Needzero only has two values and acts as a bool, so make it a bool. Change-Id: Ica46e5ebafbe478017ea52ce6bb335f404059677 Reviewed-on: https://go-review.googlesource.com/6800 Reviewed-by: Ian Lance Taylor <iant@golang.org> Run-TryBot: Dave Cheney <dave@cheney.net> TryBot-Result: Gobot Gobot <gobot@golang.org>
2015-03-05 07:11:40 +11:00
if !n.Needzero {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
continue
}
if n.Class != gc.PAUTO {
gc.Fatal("needzero class %d", n.Class)
}
if n.Type.Width%int64(gc.Widthptr) != 0 || n.Xoffset%int64(gc.Widthptr) != 0 || n.Type.Width == 0 {
gc.Fatal("var %v has size %d offset %d", gc.Nconv(n, obj.FmtLong), int(n.Type.Width), int(n.Xoffset))
}
if lo != hi && n.Xoffset+n.Type.Width >= lo-int64(2*gc.Widthptr) {
// merge with range we already have
lo = gc.Rnd(n.Xoffset, int64(gc.Widthptr))
continue
}
// zero old range
p = zerorange(p, int64(frame), lo, hi, &r0)
// set new range
hi = n.Xoffset + n.Type.Width
lo = n.Xoffset
}
// zero final range
zerorange(p, int64(frame), lo, hi, &r0)
}
func zerorange(p *obj.Prog, frame int64, lo int64, hi int64, r0 *uint32) *obj.Prog {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
cnt := hi - lo
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
if cnt == 0 {
return p
}
if *r0 == 0 {
p = appendpp(p, arm.AMOVW, obj.TYPE_CONST, 0, 0, obj.TYPE_REG, arm.REG_R0, 0)
*r0 = 1
}
if cnt < int64(4*gc.Widthptr) {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
for i := int64(0); i < cnt; i += int64(gc.Widthptr) {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
p = appendpp(p, arm.AMOVW, obj.TYPE_REG, arm.REG_R0, 0, obj.TYPE_MEM, arm.REGSP, int32(4+frame+lo+i))
}
} else if !gc.Nacl && (cnt <= int64(128*gc.Widthptr)) {
p = appendpp(p, arm.AADD, obj.TYPE_CONST, 0, int32(4+frame+lo), obj.TYPE_REG, arm.REG_R1, 0)
p.Reg = arm.REGSP
p = appendpp(p, obj.ADUFFZERO, obj.TYPE_NONE, 0, 0, obj.TYPE_MEM, 0, 0)
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
f := gc.Sysfunc("duffzero")
cmd/internal/gc: change Naddr to take a *Addr to fill in This allows gins to let Naddr fill in p.From and p.To directly, avoiding the zeroing and copying of a temporary. Change-Id: I96d120afe266e68f94d5e82b00886bf6bd458f85 Reviewed-on: https://go-review.googlesource.com/7742 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-03-16 15:27:19 -04:00
gc.Naddr(&p.To, f)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gc.Afunclit(&p.To, f)
p.To.Offset = 4 * (128 - cnt/int64(gc.Widthptr))
} else {
p = appendpp(p, arm.AADD, obj.TYPE_CONST, 0, int32(4+frame+lo), obj.TYPE_REG, arm.REG_R1, 0)
p.Reg = arm.REGSP
p = appendpp(p, arm.AADD, obj.TYPE_CONST, 0, int32(cnt), obj.TYPE_REG, arm.REG_R2, 0)
p.Reg = arm.REG_R1
p = appendpp(p, arm.AMOVW, obj.TYPE_REG, arm.REG_R0, 0, obj.TYPE_MEM, arm.REG_R1, 4)
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
p1 := p
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
p.Scond |= arm.C_PBIT
p = appendpp(p, arm.ACMP, obj.TYPE_REG, arm.REG_R1, 0, obj.TYPE_NONE, 0, 0)
p.Reg = arm.REG_R2
p = appendpp(p, arm.ABNE, obj.TYPE_NONE, 0, 0, obj.TYPE_BRANCH, 0, 0)
gc.Patch(p, p1)
}
return p
}
func appendpp(p *obj.Prog, as int, ftype int, freg int, foffset int32, ttype int, treg int, toffset int32) *obj.Prog {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
q := gc.Ctxt.NewProg()
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gc.Clearp(q)
q.As = int16(as)
q.Lineno = p.Lineno
q.From.Type = int16(ftype)
q.From.Reg = int16(freg)
q.From.Offset = int64(foffset)
q.To.Type = int16(ttype)
q.To.Reg = int16(treg)
q.To.Offset = int64(toffset)
q.Link = p.Link
p.Link = q
return q
}
/*
* generate high multiply
* res = (nl * nr) >> wordsize
*/
func cgen_hmul(nl *gc.Node, nr *gc.Node, res *gc.Node) {
if nl.Ullman < nr.Ullman {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
tmp := nl
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
nl = nr
nr = tmp
}
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
t := nl.Type
w := int(t.Width * 8)
var n1 gc.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
gc.Regalloc(&n1, t, res)
gc.Cgen(nl, &n1)
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var n2 gc.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
gc.Regalloc(&n2, t, nil)
gc.Cgen(nr, &n2)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
switch gc.Simtype[t.Etype] {
case gc.TINT8,
gc.TINT16:
gins(optoas(gc.OMUL, t), &n2, &n1)
gshift(arm.AMOVW, &n1, arm.SHIFT_AR, int32(w), &n1)
case gc.TUINT8,
gc.TUINT16:
gins(optoas(gc.OMUL, t), &n2, &n1)
gshift(arm.AMOVW, &n1, arm.SHIFT_LR, int32(w), &n1)
// perform a long multiplication.
case gc.TINT32,
gc.TUINT32:
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var p *obj.Prog
cmd/{5,6,8,9}g, cmd/internal/gc: use bools for is* and okfor* No functional changes. This diff was generated as follows: * Manually edit cmd/internal/gc/go.go to update types and group variables. * Manually edit initialization in cmd/internal/gc/align.go--localized s/1/true. * Manually fix the handling of sign in cmd/internal/gc/walk.go in func bounded (near line 4000). * Manually update go.y and regenerate y.go. * Run gofmt -r many times to do the rest, using https://gist.github.com/josharian/0f61dbb2dff81f938e70. toolstash -cmp on the stdlib comes back green. Change-Id: I19766ed551714e51b325133e7138818d117b3a9a Reviewed-on: https://go-review.googlesource.com/6530 Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-01 07:54:01 +00:00
if gc.Issigned[t.Etype] {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
p = gins(arm.AMULL, &n2, nil)
} else {
p = gins(arm.AMULLU, &n2, nil)
}
// n2 * n1 -> (n1 n2)
p.Reg = n1.Val.U.Reg
p.To.Type = obj.TYPE_REGREG
p.To.Reg = n1.Val.U.Reg
p.To.Offset = int64(n2.Val.U.Reg)
default:
gc.Fatal("cgen_hmul %v", gc.Tconv(t, 0))
}
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
gc.Cgen(&n1, res)
gc.Regfree(&n1)
gc.Regfree(&n2)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
}
/*
* generate shift according to op, one of:
* res = nl << nr
* res = nl >> nr
*/
[dev.cc] cmd/internal/obj, cmd/internal/gc, new6g: reconvert Reconvert using rsc.io/c2go rev 27b3f59. Changes to converter: - fatal does not return, so no fallthrough after fatal in switch - many more function results and variables identified as bool - simplification of negated boolean expressions Change-Id: I3bc67da5e46cb7ee613e230cf7e9533036cc870b Reviewed-on: https://go-review.googlesource.com/5171 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-02-17 22:13:49 -05:00
func cgen_shift(op int, bounded bool, nl *gc.Node, nr *gc.Node, res *gc.Node) {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
if nl.Type.Width > 4 {
gc.Fatal("cgen_shift %v", gc.Tconv(nl.Type, 0))
}
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
w := int(nl.Type.Width * 8)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
if op == gc.OLROT {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
v := int(gc.Mpgetfix(nr.Val.U.Xval))
var n1 gc.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
gc.Regalloc(&n1, nl.Type, res)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
if w == 32 {
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
gc.Cgen(nl, &n1)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gshift(arm.AMOVW, &n1, arm.SHIFT_RR, int32(w)-int32(v), &n1)
} else {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var n2 gc.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
gc.Regalloc(&n2, nl.Type, nil)
gc.Cgen(nl, &n2)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gshift(arm.AMOVW, &n2, arm.SHIFT_LL, int32(v), &n1)
gshift(arm.AORR, &n2, arm.SHIFT_LR, int32(w)-int32(v), &n1)
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
gc.Regfree(&n2)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
// Ensure sign/zero-extended result.
gins(optoas(gc.OAS, nl.Type), &n1, &n1)
}
gmove(&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
gc.Regfree(&n1)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
return
}
if nr.Op == gc.OLITERAL {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var n1 gc.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
gc.Regalloc(&n1, nl.Type, res)
gc.Cgen(nl, &n1)
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
sc := uint64(gc.Mpgetfix(nr.Val.U.Xval))
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
if sc == 0 {
} else // nothing to do
if sc >= uint64(nl.Type.Width*8) {
cmd/{5,6,8,9}g, cmd/internal/gc: use bools for is* and okfor* No functional changes. This diff was generated as follows: * Manually edit cmd/internal/gc/go.go to update types and group variables. * Manually edit initialization in cmd/internal/gc/align.go--localized s/1/true. * Manually fix the handling of sign in cmd/internal/gc/walk.go in func bounded (near line 4000). * Manually update go.y and regenerate y.go. * Run gofmt -r many times to do the rest, using https://gist.github.com/josharian/0f61dbb2dff81f938e70. toolstash -cmp on the stdlib comes back green. Change-Id: I19766ed551714e51b325133e7138818d117b3a9a Reviewed-on: https://go-review.googlesource.com/6530 Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-01 07:54:01 +00:00
if op == gc.ORSH && gc.Issigned[nl.Type.Etype] {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gshift(arm.AMOVW, &n1, arm.SHIFT_AR, int32(w), &n1)
} else {
gins(arm.AEOR, &n1, &n1)
}
} else {
cmd/{5,6,8,9}g, cmd/internal/gc: use bools for is* and okfor* No functional changes. This diff was generated as follows: * Manually edit cmd/internal/gc/go.go to update types and group variables. * Manually edit initialization in cmd/internal/gc/align.go--localized s/1/true. * Manually fix the handling of sign in cmd/internal/gc/walk.go in func bounded (near line 4000). * Manually update go.y and regenerate y.go. * Run gofmt -r many times to do the rest, using https://gist.github.com/josharian/0f61dbb2dff81f938e70. toolstash -cmp on the stdlib comes back green. Change-Id: I19766ed551714e51b325133e7138818d117b3a9a Reviewed-on: https://go-review.googlesource.com/6530 Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-01 07:54:01 +00:00
if op == gc.ORSH && gc.Issigned[nl.Type.Etype] {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gshift(arm.AMOVW, &n1, arm.SHIFT_AR, int32(sc), &n1)
} else if op == gc.ORSH {
gshift(arm.AMOVW, &n1, arm.SHIFT_LR, int32(sc), &n1) // OLSH
} else {
gshift(arm.AMOVW, &n1, arm.SHIFT_LL, int32(sc), &n1)
}
}
if w < 32 && op == gc.OLSH {
gins(optoas(gc.OAS, nl.Type), &n1, &n1)
}
gmove(&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
gc.Regfree(&n1)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
return
}
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
tr := nr.Type
var t gc.Node
var n1 gc.Node
var n2 gc.Node
var n3 gc.Node
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
if tr.Width > 4 {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var nt gc.Node
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gc.Tempname(&nt, nr.Type)
if nl.Ullman >= nr.Ullman {
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
gc.Regalloc(&n2, nl.Type, res)
gc.Cgen(nl, &n2)
gc.Cgen(nr, &nt)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
n1 = nt
} else {
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
gc.Cgen(nr, &nt)
gc.Regalloc(&n2, nl.Type, res)
gc.Cgen(nl, &n2)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
}
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var hi gc.Node
var lo gc.Node
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
split64(&nt, &lo, &hi)
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
gc.Regalloc(&n1, gc.Types[gc.TUINT32], nil)
gc.Regalloc(&n3, gc.Types[gc.TUINT32], nil)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gmove(&lo, &n1)
gmove(&hi, &n3)
splitclean()
gins(arm.ATST, &n3, nil)
gc.Nodconst(&t, gc.Types[gc.TUINT32], int64(w))
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
p1 := gins(arm.AMOVW, &t, &n1)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
p1.Scond = arm.C_SCOND_NE
tr = gc.Types[gc.TUINT32]
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
gc.Regfree(&n3)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
} else {
if nl.Ullman >= nr.Ullman {
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
gc.Regalloc(&n2, nl.Type, res)
gc.Cgen(nl, &n2)
gc.Regalloc(&n1, nr.Type, nil)
gc.Cgen(nr, &n1)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
} else {
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
gc.Regalloc(&n1, nr.Type, nil)
gc.Cgen(nr, &n1)
gc.Regalloc(&n2, nl.Type, res)
gc.Cgen(nl, &n2)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
}
}
// test for shift being 0
gins(arm.ATST, &n1, nil)
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
p3 := gc.Gbranch(arm.ABEQ, nil, -1)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
// test and fix up large shifts
// TODO: if(!bounded), don't emit some of this.
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
gc.Regalloc(&n3, tr, nil)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gc.Nodconst(&t, gc.Types[gc.TUINT32], int64(w))
gmove(&t, &n3)
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
gins(arm.ACMP, &n1, &n3)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
if op == gc.ORSH {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var p1 *obj.Prog
var p2 *obj.Prog
cmd/{5,6,8,9}g, cmd/internal/gc: use bools for is* and okfor* No functional changes. This diff was generated as follows: * Manually edit cmd/internal/gc/go.go to update types and group variables. * Manually edit initialization in cmd/internal/gc/align.go--localized s/1/true. * Manually fix the handling of sign in cmd/internal/gc/walk.go in func bounded (near line 4000). * Manually update go.y and regenerate y.go. * Run gofmt -r many times to do the rest, using https://gist.github.com/josharian/0f61dbb2dff81f938e70. toolstash -cmp on the stdlib comes back green. Change-Id: I19766ed551714e51b325133e7138818d117b3a9a Reviewed-on: https://go-review.googlesource.com/6530 Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-01 07:54:01 +00:00
if gc.Issigned[nl.Type.Etype] {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
p1 = gshift(arm.AMOVW, &n2, arm.SHIFT_AR, int32(w)-1, &n2)
p2 = gregshift(arm.AMOVW, &n2, arm.SHIFT_AR, &n1, &n2)
} else {
p1 = gins(arm.AEOR, &n2, &n2)
p2 = gregshift(arm.AMOVW, &n2, arm.SHIFT_LR, &n1, &n2)
}
p1.Scond = arm.C_SCOND_HS
p2.Scond = arm.C_SCOND_LO
} else {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
p1 := gins(arm.AEOR, &n2, &n2)
p2 := gregshift(arm.AMOVW, &n2, arm.SHIFT_LL, &n1, &n2)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
p1.Scond = arm.C_SCOND_HS
p2.Scond = arm.C_SCOND_LO
}
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
gc.Regfree(&n3)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gc.Patch(p3, gc.Pc)
// Left-shift of smaller word must be sign/zero-extended.
if w < 32 && op == gc.OLSH {
gins(optoas(gc.OAS, nl.Type), &n2, &n2)
}
gmove(&n2, 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
gc.Regfree(&n1)
gc.Regfree(&n2)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
}
func clearfat(nl *gc.Node) {
/* clear a fat object */
if gc.Debug['g'] != 0 {
gc.Dump("\nclearfat", nl)
}
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
w := uint32(nl.Type.Width)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
// Avoid taking the address for simple enough types.
cmd/internal/gc: move componentgen into portable code Change-Id: I652cc7a33a186d1041f62f6e7581421496832a27 Reviewed-on: https://go-review.googlesource.com/7747 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
2015-03-18 12:29:40 -04:00
if gc.Componentgen(nil, nl) {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
return
}
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
c := w % 4 // bytes
q := w / 4 // quads
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var r0 gc.Node
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
r0.Op = gc.OREGISTER
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
r0.Val.U.Reg = arm.REG_R0
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var r1 gc.Node
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
r1.Op = gc.OREGISTER
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
r1.Val.U.Reg = arm.REG_R1
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var dst gc.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
gc.Regalloc(&dst, gc.Types[gc.Tptr], &r1)
gc.Agen(nl, &dst)
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var nc gc.Node
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gc.Nodconst(&nc, gc.Types[gc.TUINT32], 0)
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var nz gc.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
gc.Regalloc(&nz, gc.Types[gc.TUINT32], &r0)
gc.Cgen(&nc, &nz)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
if q > 128 {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var end gc.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
gc.Regalloc(&end, gc.Types[gc.Tptr], nil)
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
p := gins(arm.AMOVW, &dst, &end)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
p.From.Type = obj.TYPE_ADDR
p.From.Offset = int64(q) * 4
p = gins(arm.AMOVW, &nz, &dst)
p.To.Type = obj.TYPE_MEM
p.To.Offset = 4
p.Scond |= arm.C_PBIT
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
pl := p
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
p = gins(arm.ACMP, &dst, nil)
raddr(&end, p)
gc.Patch(gc.Gbranch(arm.ABNE, nil, 0), pl)
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
gc.Regfree(&end)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
} else if q >= 4 && !gc.Nacl {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
f := gc.Sysfunc("duffzero")
p := gins(obj.ADUFFZERO, nil, f)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
gc.Afunclit(&p.To, f)
// 4 and 128 = magic constants: see ../../runtime/asm_arm.s
p.To.Offset = 4 * (128 - int64(q))
} else {
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var p *obj.Prog
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
for q > 0 {
p = gins(arm.AMOVW, &nz, &dst)
p.To.Type = obj.TYPE_MEM
p.To.Offset = 4
p.Scond |= arm.C_PBIT
//print("1. %P\n", p);
q--
}
}
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
var p *obj.Prog
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
for c > 0 {
p = gins(arm.AMOVB, &nz, &dst)
p.To.Type = obj.TYPE_MEM
p.To.Offset = 1
p.Scond |= arm.C_PBIT
//print("2. %P\n", p);
c--
}
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
gc.Regfree(&dst)
gc.Regfree(&nz)
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
}
// Called after regopt and peep have run.
// Expand CHECKNIL pseudo-op into actual nil pointer check.
func expandchecks(firstp *obj.Prog) {
var reg int
var p1 *obj.Prog
[dev.cc] cmd/5g etc: code cleanup: delay var decls and eliminate dead code Ran rsc.io/grind rev 6f0e601 on the source files. The cleanups move var declarations as close to the use as possible, splitting disjoint uses of the var into separate variables. They also remove dead code (especially in func sudoaddable), which helps with the var moving. There's more cleanup to come, but this alone cuts the time spent compiling html/template on my 2013 MacBook Pro from 3.1 seconds to 2.3 seconds. Change-Id: I4de499f47b1dd47a560c310bbcde6b08d425cfd6 Reviewed-on: https://go-review.googlesource.com/5637 Reviewed-by: Rob Pike <r@golang.org>
2015-02-23 16:07:24 -05:00
for p := firstp; p != nil; p = p.Link {
[dev.cc] cmd/internal/gc, cmd/new6g etc: convert from cmd/gc, cmd/6g etc First draft of converted Go compiler, using rsc.io/c2go rev 83d795a. Change-Id: I29f4c7010de07d2ff1947bbca9865879d83c32c3 Reviewed-on: https://go-review.googlesource.com/4851 Reviewed-by: Rob Pike <r@golang.org>
2015-02-13 14:40:36 -05:00
if p.As != obj.ACHECKNIL {
continue
}
if gc.Debug_checknil != 0 && p.Lineno > 1 { // p->lineno==1 in generated wrappers
gc.Warnl(int(p.Lineno), "generated nil check")
}
if p.From.Type != obj.TYPE_REG {
gc.Fatal("invalid nil check %v", p)
}
reg = int(p.From.Reg)
// check is
// CMP arg, $0
// MOV.EQ arg, 0(arg)
p1 = gc.Ctxt.NewProg()
gc.Clearp(p1)
p1.Link = p.Link
p.Link = p1
p1.Lineno = p.Lineno
p1.Pc = 9999
p1.As = arm.AMOVW
p1.From.Type = obj.TYPE_REG
p1.From.Reg = int16(reg)
p1.To.Type = obj.TYPE_MEM
p1.To.Reg = int16(reg)
p1.To.Offset = 0
p1.Scond = arm.C_SCOND_EQ
p.As = arm.ACMP
p.From.Type = obj.TYPE_CONST
p.From.Reg = 0
p.From.Offset = 0
p.Reg = int16(reg)
}
}
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
func ginsnop() {
var r gc.Node
gc.Nodreg(&r, gc.Types[gc.TINT], arm.REG_R0)
p := gins(arm.AAND, &r, &r)
p.Scond = arm.C_SCOND_EQ
}
/*
* generate
* as $c, n
*/
func ginscon(as int, c int64, n *gc.Node) {
var n1 gc.Node
gc.Nodconst(&n1, gc.Types[gc.TINT32], c)
var n2 gc.Node
gc.Regalloc(&n2, gc.Types[gc.TINT32], nil)
gmove(&n1, &n2)
gins(as, &n2, n)
gc.Regfree(&n2)
}
// addr += index*width if possible.
func addindex(index *gc.Node, width int64, addr *gc.Node) bool {
switch width {
case 2:
gshift(arm.AADD, index, arm.SHIFT_LL, 1, addr)
return true
case 4:
gshift(arm.AADD, index, arm.SHIFT_LL, 2, addr)
return true
case 8:
gshift(arm.AADD, index, arm.SHIFT_LL, 3, addr)
return true
}
return false
}
Reference in a new issue Copy permalink