Implement glGetUniformBlocksCHROMIUM: service side data packing

Implement glGetUniformBlocksCHROMIUM: service side data packing

The plan is to implement a Chromium internal command glGetUniformBlocksCHROMIUM,
like what we did with glGetProgramInfoCHROMIUM. So we could obtain all uniform
block related information from a linked program in one command. This is to
optimize the below commands:
  * glGetUniformBlockIndex
  * glGetActiveUniformBlockiv
  * glGetActiveUniformBlockName

By separating glGetUniformBlocksCHROMIUM from glGetProgramInfoCHROMIUM,
command buffer doesn't need to respond differently for glGetProgramInfoCHROMIUM
depending on whether it's ES2 or ES3 command. The drawback is we add an extra
client/service side round trip for some ES3 programs.  If this becomes a
critical perf issue, we can easily merge these two commands into one at a future
time.

BUG=453127
TEST=gpu_unittests
R=piman@chromium.org

Committed: https://crrev.com/b2285c12f91b4ded610cde29763417c3a9797d8b
Cr-Commit-Position: refs/heads/master@{#313854}

[Zhenyao Mo, 2015-01-28 23:04:35]

piman: please review.
Others: FYI (feel free to review).

[piman, 2015-01-28 23:29:42]

https://codereview.chromium.org/887513002/diff/1/gpu/command_buffer/service/p...
File gpu/command_buffer/service/program_manager.cc (right):

https://codereview.chromium.org/887513002/diff/1/gpu/command_buffer/service/p...
gpu/command_buffer/service/program_manager.cc:1306: //      max size of an
idenfifier: 1024.
We haven't checked all drivers though, including future ones. Drivers are
allowed to have those limits beyond the minimum required. They could be
UINT_MAX.

We could enforce these in the command buffer, by failing compiles if some of the
values are over limits that we define and are safe for this code, but I don't
think we do that anywhere (beyond hashing uniform names).

https://codereview.chromium.org/887513002/diff/1/gpu/command_buffer/service/p...
gpu/command_buffer/service/program_manager.cc:1349: total_size +=
blocks[ii].name_length;
Do we need to deal with remapped uniform names, when the translator is on and
hashes uniform names?

[Zhenyao Mo, 2015-01-29 01:45:36]

https://codereview.chromium.org/887513002/diff/1/gpu/command_buffer/service/p...
File gpu/command_buffer/service/program_manager.cc (right):

https://codereview.chromium.org/887513002/diff/1/gpu/command_buffer/service/p...
gpu/command_buffer/service/program_manager.cc:1306: //      max size of an
idenfifier: 1024.
On 2015/01/28 23:29:42, piman (Very slow to review) wrote:
> We haven't checked all drivers though, including future ones. Drivers are
> allowed to have those limits beyond the minimum required. They could be
> UINT_MAX.
> 
> We could enforce these in the command buffer, by failing compiles if some of
the
> values are over limits that we define and are safe for this code, but I don't
> think we do that anywhere (beyond hashing uniform names).

OK, I'll use CheckedNumeric instead then.

https://codereview.chromium.org/887513002/diff/1/gpu/command_buffer/service/p...
gpu/command_buffer/service/program_manager.cc:1349: total_size +=
blocks[ii].name_length;
On 2015/01/28 23:29:42, piman (Very slow to review) wrote:
> Do we need to deal with remapped uniform names, when the translator is on and
> hashes uniform names?

Done.

[Zhenyao Mo, 2015-01-29 01:51:06]

Please take another look.

[piman, 2015-01-29 02:05:16]

https://codereview.chromium.org/887513002/diff/20001/gpu/command_buffer/servi...
File gpu/command_buffer/service/program_manager.cc (right):

https://codereview.chromium.org/887513002/diff/20001/gpu/command_buffer/servi...
gpu/command_buffer/service/program_manager.cc:1327: base::CheckedNumeric<size_t>
size = sizeof(UniformBlockInfo);
CheckedNumeric<uint32_t> (here and other places) since we'll cast to uint32_t
later? It'd be unfortunate to have all the checking logic pass, but then
overflow on the cast.

[Zhenyao Mo, 2015-01-29 17:52:30]

https://codereview.chromium.org/887513002/diff/20001/gpu/command_buffer/servi...
File gpu/command_buffer/service/program_manager.cc (right):

https://codereview.chromium.org/887513002/diff/20001/gpu/command_buffer/servi...
gpu/command_buffer/service/program_manager.cc:1327: base::CheckedNumeric<size_t>
size = sizeof(UniformBlockInfo);
On 2015/01/29 02:05:16, piman (Very slow to review) wrote:
> CheckedNumeric<uint32_t> (here and other places) since we'll cast to uint32_t
> later? It'd be unfortunate to have all the checking logic pass, but then
> overflow on the cast.

Done.

[piman, 2015-01-29 19:25:01]

lgtm

[Zhenyao Mo, 2015-01-30 00:38:47]

"signal 11 segv_maperr"

Need to update the GL version from the default 2.0 to 3.1 to avoid crash due to
NULL GL APIs.

This is caused by https://codereview.chromium.org/871763002/

[Zhenyao Mo, 2015-01-30 00:38:53]

The CQ bit was checked by zmo@chromium.org

[commit-bot: I haz the power, 2015-01-30 00:40:18]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/887513002/60001

[commit-bot: I haz the power, 2015-01-30 01:33:25]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2015-01-30 01:34:11]

Patchset 4 (id:??) landed as
https://crrev.com/b2285c12f91b4ded610cde29763417c3a9797d8b
Cr-Commit-Position: refs/heads/master@{#313854}