gpu: Add a new extension CHROMIUM_deschedule.

gpu: Add a new extension CHROMIUM_deschedule.

The extension adds a function DescheduleUntilFinishedCHROMIUM that allows
clients to throttle themselves. From the perspective of the client, the command
performs the equivalent of a service side glFinish.

BUG=581777
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel

Committed: https://crrev.com/13338b6e1d09f38f5ba7dbdaca9f765281deba2f
Cr-Commit-Position: refs/heads/master@{#397047}

[erikchen, 2016-05-28 01:01:46]

Description was changed from

==========
Implement glDescheduleUntilFinished.

BUG=
==========

to

==========
Implement glDescheduleUntilFinished.

BUG=
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[erikchen, 2016-05-28 01:02:30]

The CQ bit was checked by erikchen@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-05-28 01:02:42]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2021603002/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2021603002/1

[erikchen, 2016-05-28 01:14:19]

The CQ bit was checked by erikchen@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-05-28 01:14:36]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2021603002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2021603002/20001

[erikchen, 2016-05-28 01:17:21]

Description was changed from

==========
Implement glDescheduleUntilFinished.

BUG=
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
gpu: Add a new extension CHROMIUM_deschedule.

The extension adds a function DescheduleUntilFinishedCHROMIUM that allows
clients to throttle themselves. From the perspective fo the client, the command
performs the equivalent of a service side glFinish.

BUG=581777
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[erikchen, 2016-05-28 01:17:54]

Description was changed from

==========
gpu: Add a new extension CHROMIUM_deschedule.

The extension adds a function DescheduleUntilFinishedCHROMIUM that allows
clients to throttle themselves. From the perspective fo the client, the command
performs the equivalent of a service side glFinish.

BUG=581777
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
gpu: Add a new extension CHROMIUM_deschedule.

The extension adds a function DescheduleUntilFinishedCHROMIUM that allows
clients to throttle themselves. From the perspective fo the client, the command
performs the equivalent of a service side glFinish.

BUG=581777
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[commit-bot: I haz the power, 2016-05-28 01:25:05]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-05-28 01:25:06]

Dry run: Try jobs failed on following builders:
  android_compile_dbg on tryserver.chromium.android (JOB_FAILED,
https://build.chromium.org/p/tryserver.chromium.android/builders/android_comp...)

[erikchen, 2016-05-31 17:39:42]

The CQ bit was checked by erikchen@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-05-31 17:40:03]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2021603002/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2021603002/60001

[erikchen, 2016-05-31 17:43:26]

Description was changed from

==========
gpu: Add a new extension CHROMIUM_deschedule.

The extension adds a function DescheduleUntilFinishedCHROMIUM that allows
clients to throttle themselves. From the perspective fo the client, the command
performs the equivalent of a service side glFinish.

BUG=581777
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
gpu: Add a new extension CHROMIUM_deschedule.

The extension adds a function DescheduleUntilFinishedCHROMIUM that allows
clients to throttle themselves. From the perspective of the client, the command
performs the equivalent of a service side glFinish.

BUG=581777
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[erikchen, 2016-05-31 17:55:12]

erikchen@chromium.org changed reviewers:
+ piman@chromium.org

[erikchen, 2016-05-31 17:55:13]

piman: Please review.

While this CL adds a new test, the interesting logic lies in the interaction
between GpuCommandBufferStub and GLES2CmdDecoder, which is untested. The best
way to do this is with some type of end-to-end test, not a unit test. 

My next CL will use this functionality in WebGL, which will thoroughly test this
path in the WebGL conformance tests. I would like to keep these two CLs
separate.

[commit-bot: I haz the power, 2016-05-31 19:12:06]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-05-31 19:12:07]

Dry run: This issue passed the CQ dry run.

[piman, 2016-05-31 19:25:21]

https://codereview.chromium.org/2021603002/diff/60001/gpu/command_buffer/serv...
File gpu/command_buffer/service/gles2_cmd_decoder_autogen.h (right):

https://codereview.chromium.org/2021603002/diff/60001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_autogen.h:4684: return
error::kNoError;
You need to return kDeferCommandUntilLater here, otherwise the CommandExecutor
will happily continue executing the rest of the batch of commands and only later
realize that the command buffer is unscheduled and stop processing.

One thing though, kDeferCommandUntilLater is intended to break out before the
command executes, so it will not progress the "get" pointer, which means, when
rescheduled, it will try to re-execute the same command. 2 ways out: either keep
an extra state on GLES2DecoderImpl, or introduce another error state (e.g.
kDeferLaterCommands or kYieldCommandProcessing) to get out of the inner loop in
the CommandExecutor/CommandParser while still increasing the "get" pointer.

[erikchen, 2016-05-31 22:08:26]

piman: PTAL

[erikchen, 2016-05-31 22:08:32]

The CQ bit was checked by erikchen@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-05-31 22:09:18]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2021603002/100001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2021603002/100001

[piman, 2016-05-31 22:39:21]

One last thing then I think it's good to go.

https://codereview.chromium.org/2021603002/diff/100001/gpu/command_buffer/ser...
File gpu/command_buffer/service/command_executor.cc (right):

https://codereview.chromium.org/2021603002/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/command_executor.cc:64: error ==
error::kDeferLaterCommands) {
I think we want to update the Get offset (since we did execute the command and
don't want to get back to it), and possibly call the command_processed_callback_
(so that the watchdog gets pinged). How about just following through the rest,
we should exit the loop on l.83 ?

[erikchen, 2016-05-31 22:52:01]

https://codereview.chromium.org/2021603002/diff/100001/gpu/command_buffer/ser...
File gpu/command_buffer/service/command_executor.cc (right):

https://codereview.chromium.org/2021603002/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/command_executor.cc:64: error ==
error::kDeferLaterCommands) {
On 2016/05/31 22:39:21, piman OOO back 2016-5-31 wrote:
> I think we want to update the Get offset (since we did execute the command and
> don't want to get back to it), and possibly call the
command_processed_callback_
> (so that the watchdog gets pinged). How about just following through the rest,
> we should exit the loop on l.83 ?

Good point. Done.

[erikchen, 2016-05-31 22:52:08]

The CQ bit was checked by erikchen@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-05-31 22:52:44]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2021603002/120001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2021603002/120001

[piman, 2016-05-31 23:02:58]

lgtm

[erikchen, 2016-05-31 23:18:56]

erikchen@chromium.org changed reviewers:
+ fsamuel@chromium.org

[erikchen, 2016-05-31 23:18:56]

fsamuel: Please review components/mus/gles2.

[Fady Samuel, 2016-05-31 23:21:37]

mus lgtm

[commit-bot: I haz the power, 2016-05-31 23:57:51]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-05-31 23:57:52]

Dry run: Try jobs failed on following builders:
  mac_chromium_rel_ng on tryserver.chromium.mac (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_...)

[erikchen, 2016-06-01 00:41:59]

https://codereview.chromium.org/2021603002/diff/120001/gpu/command_buffer/ser...
File gpu/command_buffer/service/gles2_cmd_decoder.cc (right):

https://codereview.chromium.org/2021603002/diff/120001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder.cc:13686: return
error::kUnknownCommand;
I thought I was being clever by returning error::kUnknownCommand. This broke by
newly added gl_test, and I don't want to make GLManager  support
multithreading...so changing this to be kNoError again.

piman: If you think it's important that this stay as kUnknownCommand, say so and
I'll follow up with a CL that fixes this and GLManager.

[erikchen, 2016-06-01 00:42:05]

The CQ bit was checked by erikchen@chromium.org

[erikchen, 2016-06-01 00:42:06]

The patchset sent to the CQ was uploaded after l-g-t-m from
fsamuel@chromium.org, piman@chromium.org
Link to the patchset: https://codereview.chromium.org/2021603002/#ps140001
(title: "Beh.")

[commit-bot: I haz the power, 2016-06-01 00:43:03]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2021603002/140001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2021603002/140001

[piman, 2016-06-01 01:35:47]

https://codereview.chromium.org/2021603002/diff/120001/gpu/command_buffer/ser...
File gpu/command_buffer/service/gles2_cmd_decoder.cc (right):

https://codereview.chromium.org/2021603002/diff/120001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder.cc:13686: return
error::kUnknownCommand;
On 2016/06/01 00:41:58, erikchen wrote:
> I thought I was being clever by returning error::kUnknownCommand. This broke
by
> newly added gl_test, and I don't want to make GLManager  support
> multithreading...so changing this to be kNoError again.
> 
> piman: If you think it's important that this stay as kUnknownCommand, say so
and
> I'll follow up with a CL that fixes this and GLManager.

kNoError is fine, but could you generate a GL_INVALID_OPERATION error? Using
LOCAL_SET_GL_ERROR.
Follow-up is ok if this is already through the CQ.

[erikchen, 2016-06-01 01:37:45]

The CQ bit was unchecked by erikchen@chromium.org

[erikchen, 2016-06-01 02:00:12]

On 2016/06/01 01:35:47, piman OOO back 2016-6-2 wrote:
>
https://codereview.chromium.org/2021603002/diff/120001/gpu/command_buffer/ser...
> File gpu/command_buffer/service/gles2_cmd_decoder.cc (right):
> 
>
https://codereview.chromium.org/2021603002/diff/120001/gpu/command_buffer/ser...
> gpu/command_buffer/service/gles2_cmd_decoder.cc:13686: return
> error::kUnknownCommand;
> On 2016/06/01 00:41:58, erikchen wrote:
> > I thought I was being clever by returning error::kUnknownCommand. This broke
> by
> > newly added gl_test, and I don't want to make GLManager  support
> > multithreading...so changing this to be kNoError again.
> > 
> > piman: If you think it's important that this stay as kUnknownCommand, say so
> and
> > I'll follow up with a CL that fixes this and GLManager.
> 
> kNoError is fine, but could you generate a GL_INVALID_OPERATION error? Using
> LOCAL_SET_GL_ERROR.
> Follow-up is ok if this is already through the CQ.

Done.

[erikchen, 2016-06-01 02:00:17]

The CQ bit was checked by erikchen@chromium.org

[erikchen, 2016-06-01 02:00:17]

The patchset sent to the CQ was uploaded after l-g-t-m from
fsamuel@chromium.org, piman@chromium.org
Link to the patchset: https://codereview.chromium.org/2021603002/#ps150001
(title: " ")

[commit-bot: I haz the power, 2016-06-01 02:00:42]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2021603002/150001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2021603002/150001

[commit-bot: I haz the power, 2016-06-01 03:58:31]

Committed patchset #9 (id:150001)

[commit-bot: I haz the power, 2016-06-01 04:00:13]

Description was changed from

==========
gpu: Add a new extension CHROMIUM_deschedule.

The extension adds a function DescheduleUntilFinishedCHROMIUM that allows
clients to throttle themselves. From the perspective of the client, the command
performs the equivalent of a service side glFinish.

BUG=581777
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
gpu: Add a new extension CHROMIUM_deschedule.

The extension adds a function DescheduleUntilFinishedCHROMIUM that allows
clients to throttle themselves. From the perspective of the client, the command
performs the equivalent of a service side glFinish.

BUG=581777
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel

Committed: https://crrev.com/13338b6e1d09f38f5ba7dbdaca9f765281deba2f
Cr-Commit-Position: refs/heads/master@{#397047}
==========

[commit-bot: I haz the power, 2016-06-01 04:00:14]

Patchset 9 (id:??) landed as
https://crrev.com/13338b6e1d09f38f5ba7dbdaca9f765281deba2f
Cr-Commit-Position: refs/heads/master@{#397047}