cmd/compile: avoid some allocations in regalloc

Compilebench:
name      old time/op       new time/op       delta
Template        283ms ± 3%        281ms ± 4%    ~     (p=0.242 n=20+20)
Unicode         137ms ± 6%        135ms ± 6%    ~     (p=0.194 n=20+19)
GoTypes         890ms ± 2%        883ms ± 1%  -0.74%  (p=0.001 n=19+19)
Compiler        4.21s ± 2%        4.20s ± 2%  -0.40%  (p=0.033 n=20+19)
SSA             9.86s ± 2%        9.68s ± 1%  -1.80%  (p=0.000 n=20+19)
Flate           185ms ± 5%        185ms ± 7%    ~     (p=0.429 n=20+20)
GoParser        222ms ± 3%        222ms ± 4%    ~     (p=0.588 n=19+20)
Reflect         572ms ± 2%        570ms ± 3%    ~     (p=0.113 n=19+20)
Tar             263ms ± 4%        259ms ± 2%  -1.41%  (p=0.013 n=20+20)
XML             321ms ± 2%        321ms ± 4%    ~     (p=0.835 n=20+19)

name      old user-time/op  new user-time/op  delta
Template        400ms ± 5%        405ms ± 5%    ~     (p=0.096 n=20+20)
Unicode         217ms ± 8%        213ms ± 8%    ~     (p=0.242 n=20+20)
GoTypes         1.23s ± 3%        1.22s ± 3%    ~     (p=0.923 n=19+20)
Compiler        5.76s ± 6%        5.81s ± 2%    ~     (p=0.687 n=20+19)
SSA             14.2s ± 4%        14.0s ± 4%    ~     (p=0.121 n=20+20)
Flate           248ms ± 7%        251ms ±10%    ~     (p=0.369 n=20+20)
GoParser        308ms ± 5%        305ms ± 6%    ~     (p=0.336 n=19+20)
Reflect         771ms ± 2%        766ms ± 2%    ~     (p=0.113 n=20+19)
Tar             370ms ± 5%        362ms ± 7%  -2.06%  (p=0.036 n=19+20)
XML             435ms ± 4%        432ms ± 5%    ~     (p=0.369 n=20+20)

name      old alloc/op      new alloc/op      delta
Template       39.5MB ± 0%       39.4MB ± 0%  -0.20%  (p=0.000 n=20+20)
Unicode        29.1MB ± 0%       29.1MB ± 0%    ~     (p=0.064 n=20+20)
GoTypes         117MB ± 0%        117MB ± 0%  -0.17%  (p=0.000 n=20+20)
Compiler        503MB ± 0%        502MB ± 0%  -0.15%  (p=0.000 n=19+19)
SSA            1.42GB ± 0%       1.42GB ± 0%  -0.16%  (p=0.000 n=20+20)
Flate          25.3MB ± 0%       25.3MB ± 0%  -0.19%  (p=0.000 n=20+20)
GoParser       31.4MB ± 0%       31.3MB ± 0%  -0.14%  (p=0.000 n=20+18)
Reflect        78.1MB ± 0%       77.9MB ± 0%  -0.34%  (p=0.000 n=20+19)
Tar            40.1MB ± 0%       40.0MB ± 0%  -0.17%  (p=0.000 n=20+20)
XML            45.3MB ± 0%       45.2MB ± 0%  -0.13%  (p=0.000 n=20+20)

name      old allocs/op     new allocs/op     delta
Template         393k ± 0%         392k ± 0%  -0.21%  (p=0.000 n=20+19)
Unicode          337k ± 0%         337k ± 0%  -0.02%  (p=0.000 n=20+20)
GoTypes         1.22M ± 0%        1.22M ± 0%  -0.21%  (p=0.000 n=20+20)
Compiler        4.77M ± 0%        4.76M ± 0%  -0.16%  (p=0.000 n=20+20)
SSA             11.8M ± 0%        11.8M ± 0%  -0.12%  (p=0.000 n=20+20)
Flate            242k ± 0%         241k ± 0%  -0.20%  (p=0.000 n=20+20)
GoParser         324k ± 0%         324k ± 0%  -0.14%  (p=0.000 n=20+20)
Reflect          985k ± 0%         981k ± 0%  -0.38%  (p=0.000 n=20+20)
Tar              403k ± 0%         402k ± 0%  -0.19%  (p=0.000 n=20+20)
XML              424k ± 0%         424k ± 0%  -0.16%  (p=0.000 n=19+20)

Change-Id: I131e382b64cd6db11a9263a477d45d80c180c499
Reviewed-on: https://go-review.googlesource.com/102421
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

diff --git a/src/cmd/compile/internal/ssa/regalloc.go b/src/cmd/compile/internal/ssa/regalloc.go
index 09efc90e72a63c3d7b1c521f95f03e9d4e78a9af..291ca91c6129fa11d02cf19039906290f8329ba4 100644 (file)
--- a/src/cmd/compile/internal/ssa/regalloc.go
+++ b/src/cmd/compile/internal/ssa/regalloc.go
@@ -987,9 +987,13 @@ func (s *regAllocState) regalloc(f *Func) {
                }
 
                // Allocate space to record the desired registers for each value.
-               dinfo = dinfo[:0]
-               for i := 0; i < len(oldSched); i++ {
-                       dinfo = append(dinfo, dentry{})
+               if l := len(oldSched); cap(dinfo) < l {
+                       dinfo = make([]dentry, l)
+               } else {
+                       dinfo = dinfo[:l]
+                       for i := range dinfo {
+                               dinfo[i] = dentry{}
+                       }
                }
 
                // Load static desired register info at the end of the block.