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

355 lines
9.3 KiB
Go
Raw Normal View History

// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package gc
// select
func typecheckselect(sel *Node) {
var ncase *Node
var n *Node
var def *Node
lno := setlineno(sel)
count := 0
typecheckslice(sel.Ninit.Slice(), Etop)
for _, n1 := range sel.List.Slice() {
count++
ncase = n1
setlineno(ncase)
if ncase.Op != OXCASE {
Fatalf("typecheckselect %v", oconv(ncase.Op, 0))
}
if ncase.List.Len() == 0 {
// default
if def != nil {
Yyerror("multiple defaults in select (first at %v)", def.Line())
} else {
def = ncase
}
} else if ncase.List.Len() > 1 {
Yyerror("select cases cannot be lists")
} else {
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
ncase.List.SetIndex(0, typecheck(ncase.List.Index(0), Etop))
n = ncase.List.Index(0)
ncase.Left = n
ncase.List.Set(nil)
setlineno(n)
switch n.Op {
default:
Yyerror("select case must be receive, send or assign recv")
// convert x = <-c into OSELRECV(x, <-c).
// remove implicit conversions; the eventual assignment
// will reintroduce them.
case OAS:
if (n.Right.Op == OCONVNOP || n.Right.Op == OCONVIFACE) && n.Right.Implicit {
n.Right = n.Right.Left
}
if n.Right.Op != ORECV {
Yyerror("select assignment must have receive on right hand side")
break
}
n.Op = OSELRECV
// convert x, ok = <-c into OSELRECV2(x, <-c) with ntest=ok
case OAS2RECV:
if n.Rlist.First().Op != ORECV {
Yyerror("select assignment must have receive on right hand side")
break
}
n.Op = OSELRECV2
n.Left = n.List.First()
n.List.Set1(n.List.Second())
n.Right = n.Rlist.First()
n.Rlist.Set(nil)
// convert <-c into OSELRECV(N, <-c)
case ORECV:
n = Nod(OSELRECV, nil, n)
n.Typecheck = 1
ncase.Left = n
case OSEND:
break
}
}
typecheckslice(ncase.Nbody.Slice(), Etop)
}
sel.Xoffset = int64(count)
lineno = lno
}
func walkselect(sel *Node) {
if sel.List.Len() == 0 && sel.Xoffset != 0 {
Fatalf("double walkselect") // already rewrote
}
lno := setlineno(sel)
i := sel.List.Len()
// optimization: zero-case select
var init []*Node
var r *Node
var n *Node
var var_ *Node
var selv *Node
if i == 0 {
sel.Nbody.Set1(mkcall("block", nil, nil))
goto out
}
// optimization: one-case select: single op.
// TODO(rsc): Reenable optimization once order.go can handle it.
// golang.org/issue/7672.
if i == 1 {
cas := sel.List.First()
setlineno(cas)
l := cas.Ninit.Slice()
if cas.Left != nil { // not default:
n := cas.Left
l = append(l, n.Ninit.Slice()...)
n.Ninit.Set(nil)
var ch *Node
switch n.Op {
default:
Fatalf("select %v", oconv(n.Op, 0))
// ok already
case OSEND:
ch = n.Left
case OSELRECV, OSELRECV2:
ch = n.Right.Left
if n.Op == OSELRECV || n.List.Len() == 0 {
if n.Left == nil {
n = n.Right
} else {
n.Op = OAS
}
break
}
if n.Left == nil {
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
nblank = typecheck(nblank, Erv|Easgn)
n.Left = nblank
}
n.Op = OAS2
n.List.Set(append([]*Node{n.Left}, n.List.Slice()...))
n.Rlist.Set1(n.Right)
n.Right = nil
n.Left = nil
n.Typecheck = 0
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
n = typecheck(n, Etop)
}
// if ch == nil { block() }; n;
a := Nod(OIF, nil, nil)
a.Left = Nod(OEQ, ch, nodnil())
var ln Nodes
ln.Set(l)
a.Nbody.Set1(mkcall("block", nil, &ln))
l = ln.Slice()
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
a = typecheck(a, Etop)
l = append(l, a)
l = append(l, n)
}
l = append(l, cas.Nbody.Slice()...)
sel.Nbody.Set(l)
goto out
}
// convert case value arguments to addresses.
// this rewrite is used by both the general code and the next optimization.
for _, cas := range sel.List.Slice() {
setlineno(cas)
n = cas.Left
if n == nil {
continue
}
switch n.Op {
case OSEND:
n.Right = Nod(OADDR, n.Right, nil)
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
n.Right = typecheck(n.Right, Erv)
case OSELRECV, OSELRECV2:
if n.Op == OSELRECV2 && n.List.Len() == 0 {
n.Op = OSELRECV
}
if n.Op == OSELRECV2 {
n.List.SetIndex(0, Nod(OADDR, n.List.First(), nil))
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
n.List.SetIndex(0, typecheck(n.List.Index(0), Erv))
}
if n.Left == nil {
n.Left = nodnil()
} else {
n.Left = Nod(OADDR, n.Left, nil)
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
n.Left = typecheck(n.Left, Erv)
}
}
}
// optimization: two-case select but one is default: single non-blocking op.
if i == 2 && (sel.List.First().Left == nil || sel.List.Second().Left == nil) {
var cas *Node
var dflt *Node
if sel.List.First().Left == nil {
cas = sel.List.Second()
dflt = sel.List.First()
} else {
dflt = sel.List.Second()
cas = sel.List.First()
}
n := cas.Left
setlineno(n)
r := Nod(OIF, nil, nil)
r.Ninit.Set(cas.Ninit.Slice())
switch n.Op {
default:
Fatalf("select %v", oconv(n.Op, 0))
// if selectnbsend(c, v) { body } else { default body }
case OSEND:
ch := n.Left
r.Left = mkcall1(chanfn("selectnbsend", 2, ch.Type), Types[TBOOL], &r.Ninit, typename(ch.Type), ch, n.Right)
// if c != nil && selectnbrecv(&v, c) { body } else { default body }
case OSELRECV:
r = Nod(OIF, nil, nil)
r.Ninit.Set(cas.Ninit.Slice())
ch := n.Right.Left
r.Left = mkcall1(chanfn("selectnbrecv", 2, ch.Type), Types[TBOOL], &r.Ninit, typename(ch.Type), n.Left, ch)
// if c != nil && selectnbrecv2(&v, c) { body } else { default body }
case OSELRECV2:
r = Nod(OIF, nil, nil)
r.Ninit.Set(cas.Ninit.Slice())
ch := n.Right.Left
r.Left = mkcall1(chanfn("selectnbrecv2", 2, ch.Type), Types[TBOOL], &r.Ninit, typename(ch.Type), n.Left, n.List.First(), ch)
}
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
r.Left = typecheck(r.Left, Erv)
r.Nbody.Set(cas.Nbody.Slice())
r.Rlist.Set(append(dflt.Ninit.Slice(), dflt.Nbody.Slice()...))
sel.Nbody.Set1(r)
goto out
}
init = sel.Ninit.Slice()
sel.Ninit.Set(nil)
// generate sel-struct
setlineno(sel)
selv = temp(selecttype(int32(sel.Xoffset)))
r = Nod(OAS, selv, nil)
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
r = typecheck(r, Etop)
init = append(init, r)
var_ = conv(conv(Nod(OADDR, selv, nil), Types[TUNSAFEPTR]), Ptrto(Types[TUINT8]))
r = mkcall("newselect", nil, nil, var_, Nodintconst(selv.Type.Width), Nodintconst(sel.Xoffset))
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
r = typecheck(r, Etop)
init = append(init, r)
// register cases
for _, cas := range sel.List.Slice() {
setlineno(cas)
n = cas.Left
r = Nod(OIF, nil, nil)
r.Ninit.Set(cas.Ninit.Slice())
cas.Ninit.Set(nil)
if n != nil {
r.Ninit.AppendNodes(&n.Ninit)
n.Ninit.Set(nil)
}
if n == nil {
// selectdefault(sel *byte);
r.Left = mkcall("selectdefault", Types[TBOOL], &r.Ninit, var_)
} else {
switch n.Op {
default:
Fatalf("select %v", oconv(n.Op, 0))
// selectsend(sel *byte, hchan *chan any, elem *any) (selected bool);
case OSEND:
r.Left = mkcall1(chanfn("selectsend", 2, n.Left.Type), Types[TBOOL], &r.Ninit, var_, n.Left, n.Right)
// selectrecv(sel *byte, hchan *chan any, elem *any) (selected bool);
case OSELRECV:
r.Left = mkcall1(chanfn("selectrecv", 2, n.Right.Left.Type), Types[TBOOL], &r.Ninit, var_, n.Right.Left, n.Left)
// selectrecv2(sel *byte, hchan *chan any, elem *any, received *bool) (selected bool);
case OSELRECV2:
r.Left = mkcall1(chanfn("selectrecv2", 2, n.Right.Left.Type), Types[TBOOL], &r.Ninit, var_, n.Right.Left, n.Left, n.List.First())
}
}
// selv is no longer alive after use.
r.Nbody.Append(Nod(OVARKILL, selv, nil))
r.Nbody.AppendNodes(&cas.Nbody)
r.Nbody.Append(Nod(OBREAK, nil, nil))
init = append(init, r)
}
// run the select
setlineno(sel)
init = append(init, mkcall("selectgo", nil, nil, var_))
sel.Nbody.Set(init)
out:
sel.List.Set(nil)
walkstmtlist(sel.Nbody.Slice())
lineno = lno
}
// Keep in sync with src/runtime/select.go.
func selecttype(size int32) *Type {
// TODO(dvyukov): it's possible to generate Scase only once
// and then cache; and also cache Select per size.
scase := Nod(OTSTRUCT, nil, nil)
scase.List.Append(Nod(ODCLFIELD, newname(Lookup("elem")), typenod(Ptrto(Types[TUINT8]))))
scase.List.Append(Nod(ODCLFIELD, newname(Lookup("chan")), typenod(Ptrto(Types[TUINT8]))))
scase.List.Append(Nod(ODCLFIELD, newname(Lookup("pc")), typenod(Types[TUINTPTR])))
scase.List.Append(Nod(ODCLFIELD, newname(Lookup("kind")), typenod(Types[TUINT16])))
scase.List.Append(Nod(ODCLFIELD, newname(Lookup("so")), typenod(Types[TUINT16])))
scase.List.Append(Nod(ODCLFIELD, newname(Lookup("receivedp")), typenod(Ptrto(Types[TUINT8]))))
scase.List.Append(Nod(ODCLFIELD, newname(Lookup("releasetime")), typenod(Types[TUINT64])))
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
scase = typecheck(scase, Etype)
scase.Type.Noalg = true
scase.Type.Local = true
sel := Nod(OTSTRUCT, nil, nil)
sel.List.Append(Nod(ODCLFIELD, newname(Lookup("tcase")), typenod(Types[TUINT16])))
sel.List.Append(Nod(ODCLFIELD, newname(Lookup("ncase")), typenod(Types[TUINT16])))
sel.List.Append(Nod(ODCLFIELD, newname(Lookup("pollorder")), typenod(Ptrto(Types[TUINT8]))))
sel.List.Append(Nod(ODCLFIELD, newname(Lookup("lockorder")), typenod(Ptrto(Types[TUINT8]))))
arr := Nod(OTARRAY, Nodintconst(int64(size)), scase)
sel.List.Append(Nod(ODCLFIELD, newname(Lookup("scase")), arr))
arr = Nod(OTARRAY, Nodintconst(int64(size)), typenod(Types[TUINT16]))
sel.List.Append(Nod(ODCLFIELD, newname(Lookup("lockorderarr")), arr))
arr = Nod(OTARRAY, Nodintconst(int64(size)), typenod(Types[TUINT16]))
sel.List.Append(Nod(ODCLFIELD, newname(Lookup("pollorderarr")), arr))
cmd/compile: reduce use of **Node parameters Escape analysis has a hard time with tree-like structures (see #13493 and #14858). This is unlikely to change. As a result, when invoking a function that accepts a **Node parameter, we usually allocate a *Node on the heap. This happens a whole lot. This CL changes functions from taking a **Node to acting more like append: It both modifies the input and returns a replacement for it. Because of the cascading nature of escape analysis, in order to get the benefits, I had to modify almost all such functions. The remaining functions are in racewalk and the backend. I would be happy to update them as well in a separate CL. This CL was created by manually updating the function signatures and the directly impacted bits of code. The callsites were then automatically updated using a bespoke script: https://gist.github.com/josharian/046b1be7aceae244de39 For ease of reviewing and future understanding, this CL is also broken down into four CLs, mailed separately, which show the manual and the automated changes separately. They are CLs 20990, 20991, 20992, and 20993. Passes toolstash -cmp. name old time/op new time/op delta Template 335ms ± 5% 324ms ± 5% -3.35% (p=0.000 n=23+24) Unicode 176ms ± 9% 165ms ± 6% -6.12% (p=0.000 n=23+24) GoTypes 1.10s ± 4% 1.07s ± 2% -2.77% (p=0.000 n=24+24) Compiler 5.31s ± 3% 5.15s ± 3% -2.95% (p=0.000 n=24+24) MakeBash 41.6s ± 1% 41.7s ± 2% ~ (p=0.586 n=23+23) name old alloc/op new alloc/op delta Template 63.3MB ± 0% 62.4MB ± 0% -1.36% (p=0.000 n=25+23) Unicode 42.4MB ± 0% 41.6MB ± 0% -1.99% (p=0.000 n=24+25) GoTypes 220MB ± 0% 217MB ± 0% -1.11% (p=0.000 n=25+25) Compiler 994MB ± 0% 973MB ± 0% -2.08% (p=0.000 n=24+25) name old allocs/op new allocs/op delta Template 681k ± 0% 574k ± 0% -15.71% (p=0.000 n=24+25) Unicode 518k ± 0% 413k ± 0% -20.34% (p=0.000 n=25+24) GoTypes 2.08M ± 0% 1.78M ± 0% -14.62% (p=0.000 n=25+25) Compiler 9.26M ± 0% 7.64M ± 0% -17.48% (p=0.000 n=25+25) name old text-bytes new text-bytes delta HelloSize 578k ± 0% 578k ± 0% ~ (all samples are equal) CmdGoSize 6.46M ± 0% 6.46M ± 0% ~ (all samples are equal) name old data-bytes new data-bytes delta HelloSize 128k ± 0% 128k ± 0% ~ (all samples are equal) CmdGoSize 281k ± 0% 281k ± 0% ~ (all samples are equal) name old exe-bytes new exe-bytes delta HelloSize 921k ± 0% 921k ± 0% ~ (all samples are equal) CmdGoSize 9.86M ± 0% 9.86M ± 0% ~ (all samples are equal) Change-Id: I277d95bd56d51c166ef7f560647aeaa092f3f475 Reviewed-on: https://go-review.googlesource.com/20959 Reviewed-by: Dave Cheney <dave@cheney.net> Reviewed-by: Ian Lance Taylor <iant@golang.org>
2016-03-20 08:03:31 -07:00
sel = typecheck(sel, Etype)
sel.Type.Noalg = true
sel.Type.Local = true
return sel.Type
}