code review 115780043: runtime: improve select benchmarks

runtime: improve select benchmarks
1. Add select on sync channels benchmark.
2. Make channels in BenchmarkSelectNonblock shared.
With GOMAXPROCS=1 it is the same, but with GOMAXPROCS>1
it becomes a more interesting benchmark.

[dvyukov, 2014-07-14 11:22:11]

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

I'd like you to review this change to
https://dvyukov%40google.com@code.google.com/p/go/

[khr, 2014-07-17 04:27:11]

https://codereview.appspot.com/115780043/diff/40001/src/pkg/runtime/chan_test.go
File src/pkg/runtime/chan_test.go (right):

https://codereview.appspot.com/115780043/diff/40001/src/pkg/runtime/chan_test...
src/pkg/runtime/chan_test.go:510: myc4 := make(chan int, 1)
Now all the selects except the first few iterations will take the default
clause.  Is that what you want?

The reads always used to default.  But the writes didn't have to.

[dvyukov, 2014-07-17 04:32:39]

https://codereview.appspot.com/115780043/diff/40001/src/pkg/runtime/chan_test.go
File src/pkg/runtime/chan_test.go (right):

https://codereview.appspot.com/115780043/diff/40001/src/pkg/runtime/chan_test...
src/pkg/runtime/chan_test.go:510: myc4 := make(chan int, 1)
On 2014/07/17 04:27:11, khr wrote:
> Now all the selects except the first few iterations will take the default
> clause.  Is that what you want?
> 
> The reads always used to default.  But the writes didn't have to.

Yes, it's benchmark for default case.
Before and after this CL 2 writes succeed, but it does not matter because it's
only 2.

[khr1, 2014-07-17 04:33:47]

LGTM.



On Wed, Jul 16, 2014 at 9:32 PM, <dvyukov@google.com> wrote:

>
> https://codereview.appspot.com/115780043/diff/40001/src/
> pkg/runtime/chan_test.go
> File src/pkg/runtime/chan_test.go (right):
>
> https://codereview.appspot.com/115780043/diff/40001/src/
> pkg/runtime/chan_test.go#newcode510
> src/pkg/runtime/chan_test.go:510: myc4 := make(chan int, 1)
> On 2014/07/17 04:27:11, khr wrote:
>
>> Now all the selects except the first few iterations will take the
>>
> default
>
>> clause.  Is that what you want?
>>
>
>  The reads always used to default.  But the writes didn't have to.
>>
>
> Yes, it's benchmark for default case.
> Before and after this CL 2 writes succeed, but it does not matter
> because it's only 2.
>
> https://codereview.appspot.com/115780043/
>

[dvyukov, 2014-07-17 17:49:52]

*** Submitted as https://code.google.com/p/go/source/detail?r=f2fb9e3b91d1 ***

runtime: improve select benchmarks
1. Add select on sync channels benchmark.
2. Make channels in BenchmarkSelectNonblock shared.
With GOMAXPROCS=1 it is the same, but with GOMAXPROCS>1
it becomes a more interesting benchmark.

LGTM=khr
R=golang-codereviews, khr
CC=golang-codereviews
https://codereview.appspot.com/115780043

[gobot, 2014-07-17 18:04:40]

This CL appears to have broken the netbsd-amd64-bsiegert builder.
See http://build.golang.org/log/67a1fe57a951fb48630f573163823192639899ba