Revert "Exynos decoder: workaround for wrong MFC output format."

Revert "Exynos decoder: workaround for wrong MFC output format."

This reverts r226912 (commit 1d07393296ba6e33eb398cc0e02077aa61fb2ccd).

The corruption is now fixed with a new version of the MFC firmware.

BUG=chrome-os-partner:23078
TEST=Verify that crbug.com/302870 does not reproduce anymore

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=231412

[Ami GONE FROM CHROMIUM, 2013-10-28 15:03:41]

In the CL description please name the (reverted) chromium commit by subversion
revision number, not git sha1.

https://codereview.chromium.org/47553003/diff/1/content/common/gpu/media/exyn...
File content/common/gpu/media/exynos_video_decode_accelerator.cc (right):

https://codereview.chromium.org/47553003/diff/1/content/common/gpu/media/exyn...
content/common/gpu/media/exynos_video_decode_accelerator.cc:404:
format.fmt.pix_mp.pixelformat = V4L2_PIX_FMT_NV12MT_16X16;
Does this improve perf? 
(why is NV12MT_16x16 better than NV12M?  IDK)

[sheu, 2013-10-28 17:24:48]

https://codereview.chromium.org/47553003/diff/1/content/common/gpu/media/exyn...
File content/common/gpu/media/exynos_video_decode_accelerator.cc (right):

https://codereview.chromium.org/47553003/diff/1/content/common/gpu/media/exyn...
content/common/gpu/media/exynos_video_decode_accelerator.cc:404:
format.fmt.pix_mp.pixelformat = V4L2_PIX_FMT_NV12MT_16X16;
On 2013/10/28 15:03:41, Ami Fischman wrote:
> Does this improve perf? 
> (why is NV12MT_16x16 better than NV12M?  IDK)

MT16x16 is a tiled format, and NV12M is plain linear.  We get about a 0.8% perf
improvement from this.  (Yes, I tested, and it's outside of run-to-run noise.)

[sheu, 2013-10-28 17:32:08]

Subversion revision seems to be the primary key for our purposes.  But git SHA
is nice too.

lgtm

[hshi1, 2013-10-28 17:36:23]

On 2013/10/28 15:03:41, Ami Fischman wrote:
> In the CL description please name the (reverted) chromium commit by subversion
> revision number, not git sha1.
> 
>
https://codereview.chromium.org/47553003/diff/1/content/common/gpu/media/exyn...
> File content/common/gpu/media/exynos_video_decode_accelerator.cc (right):
> 
>
https://codereview.chromium.org/47553003/diff/1/content/common/gpu/media/exyn...
> content/common/gpu/media/exynos_video_decode_accelerator.cc:404:
> format.fmt.pix_mp.pixelformat = V4L2_PIX_FMT_NV12MT_16X16;
> Does this improve perf? 
> (why is NV12MT_16x16 better than NV12M?  IDK)

Re: why is NV12MT_16x16 better than NV12M: NV12MT_16x16 is a tiled format with
16x16 tiles coinciding with the 16x16 macroblock boundaries. Decoder's memory
access pattern is localized around the current macroblock (for fetching pixels
from reference frame) thus it is believed that a tiled format is better than a
plain linear format, especially when frame width is larger.

[hshi1, 2013-10-28 17:36:24]

On 2013/10/28 15:03:41, Ami Fischman wrote:
> In the CL description please name the (reverted) chromium commit by subversion
> revision number, not git sha1.
> 
>
https://codereview.chromium.org/47553003/diff/1/content/common/gpu/media/exyn...
> File content/common/gpu/media/exynos_video_decode_accelerator.cc (right):
> 
>
https://codereview.chromium.org/47553003/diff/1/content/common/gpu/media/exyn...
> content/common/gpu/media/exynos_video_decode_accelerator.cc:404:
> format.fmt.pix_mp.pixelformat = V4L2_PIX_FMT_NV12MT_16X16;
> Does this improve perf? 
> (why is NV12MT_16x16 better than NV12M?  IDK)

Re: why is NV12MT_16x16 better than NV12M: NV12MT_16x16 is a tiled format with
16x16 tiles coinciding with the 16x16 macroblock boundaries. Decoder's memory
access pattern is localized around the current macroblock (for fetching pixels
from reference frame) thus it is believed that a tiled format is better than a
plain linear format, especially when frame width is larger.

[Ami GONE FROM CHROMIUM, 2013-10-28 17:38:53]

LGTM
thx for the explanations

[sheu, 2013-10-28 17:41:04]

Updated description and hit 'commit' for ya :-)

[commit-bot: I haz the power, 2013-10-28 17:42:14]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/posciak@chromium.org/47553003/1

[commit-bot: I haz the power, 2013-10-28 21:43:03]

Change committed as 231412