code review 115160043: runtime: add Benchmark[Clear|Copy]Fat[8|12]

runtime: add Benchmark[Clear|Copy]Fat[8|12]

These correspond to 2 and 3 word fat copies/clears on 8g, which dominate usage in the stdlib. (70% of copies and 46% of clears are for 2 or 3 words.) I missed these in CL 111350043, which added 2 and 3 word benchmarks for 6g. A follow-up CL will optimize these cases.

[josharian, 2014-07-21 02:09:18]

Hello khr@golang.org (cc: golang-codereviews@googlegroups.com),

I'd like you to review this change to
https://code.google.com/p/go

[khr1, 2014-07-21 04:01:15]

LGTM.



On Sun, Jul 20, 2014 at 7:09 PM, <josharian@gmail.com> wrote:

> Reviewers: khr,
>
> Message:
> Hello khr@golang.org (cc: golang-codereviews@googlegroups.com),
>
> I'd like you to review this change to
> https://code.google.com/p/go
>
>
> Description:
> runtime: add Benchmark[Clear|Copy]Fat[8|12]
>
> These correspond to 2 and 3 word fat copies/clears on 8g, which dominate
> usage in the stdlib. (70% of copies and 46% of clears are for 2 or 3
> words.) I missed these in CL 111350043, which added 2 and 3 word
> benchmarks for 6g. A follow-up CL will optimize these cases.
>
> Please review this at https://codereview.appspot.com/115160043/
>
> Affected files (+26, -0 lines):
>   M src/pkg/runtime/memmove_test.go
>
>
> Index: src/pkg/runtime/memmove_test.go
> ===================================================================
> --- a/src/pkg/runtime/memmove_test.go
> +++ b/src/pkg/runtime/memmove_test.go
> @@ -162,6 +162,18 @@
>  func BenchmarkMemclr4096(b *testing.B)  { bmMemclr(b, 4096) }
>  func BenchmarkMemclr65536(b *testing.B) { bmMemclr(b, 65536) }
>
> +func BenchmarkClearFat8(b *testing.B) {
> +       for i := 0; i < b.N; i++ {
> +               var x [8 / 4]uint32
> +               _ = x
> +       }
> +}
> +func BenchmarkClearFat12(b *testing.B) {
> +       for i := 0; i < b.N; i++ {
> +               var x [12 / 4]uint32
> +               _ = x
> +       }
> +}
>  func BenchmarkClearFat16(b *testing.B) {
>         for i := 0; i < b.N; i++ {
>                 var x [16 / 4]uint32
> @@ -211,6 +223,20 @@
>         }
>  }
>
> +func BenchmarkCopyFat8(b *testing.B) {
> +       var x [8 / 4]uint32
> +       for i := 0; i < b.N; i++ {
> +               y := x
> +               _ = y
> +       }
> +}
> +func BenchmarkCopyFat12(b *testing.B) {
> +       var x [12 / 4]uint32
> +       for i := 0; i < b.N; i++ {
> +               y := x
> +               _ = y
> +       }
> +}
>  func BenchmarkCopyFat16(b *testing.B) {
>         var x [16 / 4]uint32
>         for i := 0; i < b.N; i++ {
>
>
>

[josharian, 2014-07-21 18:23:33]

*** Submitted as https://code.google.com/p/go/source/detail?r=e955fe8b2ed8 ***

runtime: add Benchmark[Clear|Copy]Fat[8|12]

These correspond to 2 and 3 word fat copies/clears on 8g, which dominate usage
in the stdlib. (70% of copies and 46% of clears are for 2 or 3 words.) I missed
these in CL 111350043, which added 2 and 3 word benchmarks for 6g. A follow-up
CL will optimize these cases.

LGTM=khr
R=khr
CC=golang-codereviews
https://codereview.appspot.com/115160043

[gobot, 2014-07-21 18:41:42]

This CL appears to have broken the linux-arm-cheney-panda builder.
See http://build.golang.org/log/7bcfdbf06810e32e2cce0d89c187769e9eea1c84

[josharian, 2014-07-21 18:44:40]

On 2014/07/21 18:41:42, gobot wrote:
> This CL appears to have broken the linux-arm-cheney-panda builder.
> See http://build.golang.org/log/7bcfdbf06810e32e2cce0d89c187769e9eea1c84

Nope. This is issue 8298.