cc: Avoid issuing COMMANDS_COMPLETED queries unless necessary.

cc: Avoid issuing COMMANDS_COMPLETED queries unless necessary.

Minor change to the Fence API that allows GLRenderer to not
issue sync queries unless some resource that depend on them
was used to produce a frame.

Also includes some minor renaming cleanup related to usage of
Fences in ResourceProvider code.

BUG=

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=284318

[piman, 2014-07-18 22:19:42]

lgtm

https://codereview.chromium.org/398543004/diff/1/cc/output/gl_renderer.cc
File cc/output/gl_renderer.cc (right):

https://codereview.chromium.org/398543004/diff/1/cc/output/gl_renderer.cc#new...
cc/output/gl_renderer.cc:259: gl_->BeginQueryEXT(GL_COMMANDS_COMPLETED_CHROMIUM,
query_id_);
nit: it's a bit of an abuse. Maybe add a comment saying that BeginQueryEXT on
GL_COMMANDS_COMPLETED_CHROMIUM is effectively a noop relative to GL, so it
doesn't matter where it happens.

It's conceivable that in the future, for out-of-order GPUs (e.g. tilers), a more
optimal fence (with another extension) would need to bracket draw calls as to
not force ordering. Then we'd want to revisit this.

[reveman, 2014-07-18 23:51:06]

https://codereview.chromium.org/398543004/diff/1/cc/output/gl_renderer.cc
File cc/output/gl_renderer.cc (right):

https://codereview.chromium.org/398543004/diff/1/cc/output/gl_renderer.cc#new...
cc/output/gl_renderer.cc:259: gl_->BeginQueryEXT(GL_COMMANDS_COMPLETED_CHROMIUM,
query_id_);
On 2014/07/18 22:19:42, piman (slow to review) wrote:
> nit: it's a bit of an abuse. Maybe add a comment saying that BeginQueryEXT on
> GL_COMMANDS_COMPLETED_CHROMIUM is effectively a noop relative to GL, so it
> doesn't matter where it happens.
> 
> It's conceivable that in the future, for out-of-order GPUs (e.g. tilers), a
more
> optimal fence (with another extension) would need to bracket draw calls as to
> not force ordering. Then we'd want to revisit this.

Good point. I've updated the patch so we're still issuing BeginQueryEXT before
any drawing commands that depend on the query (seems like a good idea to keep
that behavior if possible) and I've added some comments to make this more clear.

[piman, 2014-07-19 00:16:42]

ok, great, LGTM.

[vmiura, 2014-07-19 00:24:23]

LGTM

[reveman, 2014-07-19 00:46:27]

The CQ bit was checked by reveman@chromium.org

[commit-bot: I haz the power, 2014-07-19 00:47:42]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/reveman@chromium.org/398543004/20001

[commit-bot: I haz the power, 2014-07-19 03:33:38]

FYI, CQ is re-trying this CL (attempt #1).
The failing builders are:
  win_chromium_compile_dbg on tryserver.chromium
(http://build.chromium.org/p/tryserver.chromium/builders/win_chromium_compile_...)
  mac_gpu_retina_triggered_tests on tryserver.chromium.gpu
(http://build.chromium.org/p/tryserver.chromium.gpu/builders/mac_gpu_retina_tr...)
  mac_gpu_triggered_tests on tryserver.chromium.gpu
(http://build.chromium.org/p/tryserver.chromium.gpu/builders/mac_gpu_triggered...)

[commit-bot: I haz the power, 2014-07-19 05:07:59]

Change committed as 284318