Changed pack alignment of GPU command buffer commands.

Changed pack alignment of GPU command buffer commands.

Previously all command structures were declared as #pragma pack(1)
which was forcing byte-by-byte access to the command buffer; less
than optimal.

The command buffer and all commands are 4 byte aligned, so
changing to #pragma pack(4) should be safe and optimal.

BUG=339151
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=252777

[vmiura, 2014-02-02 04:53:16]

Please take a look.

[piman, 2014-02-03 22:30:49]

out-of-curiosity, do you have a micro benchmark that shows the benefit? I think
we want this no matter what, but it would be even better with numbers.

https://codereview.chromium.org/152263002/diff/1/gpu/command_buffer/common/cm...
File gpu/command_buffer/common/cmd_buffer_common.h (right):

https://codereview.chromium.org/152263002/diff/1/gpu/command_buffer/common/cm...
gpu/command_buffer/common/cmd_buffer_common.h:96: #pragma pack(push, 4)
Do we even want anything at all, instead of just letting the compiler just use
whatever packing it likes best?

[vmiura, 2014-02-03 22:44:26]

On 2014/02/03 22:30:49, piman wrote:
> out-of-curiosity, do you have a micro benchmark that shows the benefit? I
think
> we want this no matter what, but it would be even better with numbers.

I can make one.  Are there any existing examples?

[vmiura, 2014-02-03 22:44:38]

https://codereview.chromium.org/152263002/diff/1/gpu/command_buffer/common/cm...
File gpu/command_buffer/common/cmd_buffer_common.h (right):

https://codereview.chromium.org/152263002/diff/1/gpu/command_buffer/common/cm...
gpu/command_buffer/common/cmd_buffer_common.h:96: #pragma pack(push, 4)
On 2014/02/03 22:30:49, piman wrote:
> Do we even want anything at all, instead of just letting the compiler just use
> whatever packing it likes best?

The command buffer is 4 byte aligned, so this pack(4) is mainly here to make
sure any 8 byte aligned types in commands won't fault due to bad alignment on
any architecture.

Currently I think just QuerySync has a "uint64 result".

All other commands use just uint32/int32/float, so in theory pack(4) will be a
noop.

[no sievers, 2014-02-21 21:49:00]

lgtm

[vmiura, 2014-02-22 00:14:13]

The CQ bit was checked by vmiura@chromium.org

[commit-bot: I haz the power, 2014-02-22 00:15:56]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/vmiura@chromium.org/152263002/390001

[commit-bot: I haz the power, 2014-02-22 01:31:57]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/vmiura@chromium.org/152263002/390001

[commit-bot: I haz the power, 2014-02-22 08:23:13]

Change committed as 252777