code review 128340043: runtime: mark with non-atomic operations when GOMAXPROCS=1

runtime: mark with non-atomic operations when GOMAXPROCS=1
Perf builders show 3-5% GC pause increase with GOMAXPROCS=1 when marking with atomic ops:
http://goperfd.appspot.com/perfdetail?commit=a8a6e765d6a87f7ccb71fd85a60eb5a821151f85&commit0=3b864e02b987171e05e2e9d0840b85b5b6476386&kind=builder&builder=linux-amd64&benchmark=http

[dvyukov, 2014-08-15 18:47:56]

Hello golang-codereviews@googlegroups.com (cc: dave@cheney.net, khr@golang.org,
rlh@golang.org, rsc@golang.org),

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

[rlh, 2014-08-15 18:52:59]

On 2014/08/15 18:47:56, dvyukov wrote:
> Hello mailto:golang-codereviews@googlegroups.com (cc: mailto:dave@cheney.net,
mailto:khr@golang.org,
> mailto:rlh@golang.org, mailto:rsc@golang.org),
> 
> I'd like you to review this change to
> https://dvyukov%40google.com@code.google.com/p/go/

Looks good.

[dvyukov, 2014-08-15 18:58:15]

On 2014/08/15 18:52:59, rlh wrote:
> On 2014/08/15 18:47:56, dvyukov wrote:
> > Hello mailto:golang-codereviews@googlegroups.com (cc:
mailto:dave@cheney.net,
> mailto:khr@golang.org,
> > mailto:rlh@golang.org, mailto:rsc@golang.org),
> > 
> > I'd like you to review this change to
> > https://dvyukov%40google.com@code.google.com/p/go/
> 
> Looks good.

If you want to approve it, please say LGTM, it's magic phrase recognized by
tools.

[rlh, 2014-08-15 21:00:42]

On 2014/08/15 18:58:15, dvyukov wrote:
> On 2014/08/15 18:52:59, rlh wrote:
> > On 2014/08/15 18:47:56, dvyukov wrote:
> > > Hello mailto:golang-codereviews@googlegroups.com (cc:
> mailto:dave@cheney.net,
> > mailto:khr@golang.org,
> > > mailto:rlh@golang.org, mailto:rsc@golang.org),
> > > 
> > > I'd like you to review this change to
> > > https://dvyukov%40google.com@code.google.com/p/go/
> > 
> > Looks good.
> 
> If you want to approve it, please say LGTM, it's magic phrase recognized by
> tools.

LGTM

[dvyukov, 2014-08-16 05:08:03]

*** Submitted as https://code.google.com/p/go/source/detail?r=55f09668f3b2 ***

runtime: mark with non-atomic operations when GOMAXPROCS=1
Perf builders show 3-5% GC pause increase with GOMAXPROCS=1 when marking with
atomic ops:
http://goperfd.appspot.com/perfdetail?commit=a8a6e765d6a87f7ccb71fd85a60eb5a8...

LGTM=rlh
R=golang-codereviews, rlh
CC=dave, golang-codereviews, khr, rsc
https://codereview.appspot.com/128340043

[gobot, 2014-08-16 05:22:41]

This CL appears to have broken the darwin-amd64 builder.
See http://build.golang.org/log/4666ee88b31ae7dbcad94421d2f9d346e5aacd24

[dave_cheney.net, 2014-08-16 05:31:59]

Unrelated, but I'm working if the networking fix that was backported
to 1.3.1 isn't related to the increase in networking errors we see.

--- FAIL: TestErrorNil (0.00s)
net_test.go:230: Dial 127.0.0.1:65535 succeeded
FAIL
FAIL net 2.913s



On Sat, Aug 16, 2014 at 3:22 PM,  <gobot@golang.org> wrote:
> This CL appears to have broken the darwin-amd64 builder.
> See http://build.golang.org/log/4666ee88b31ae7dbcad94421d2f9d346e5aacd24
>
> https://codereview.appspot.com/128340043/