[FFmpeg-devel] [PATCH] VP8 MMX optimizations (MC and IDCT dc_add)

[Jason Garrett-Glaser]

On Tue, Jun 22, 2010 at 6:50 PM, Jason Garrett-Glaser
<darkshikari at gmail.com> wrote:
> On Tue, Jun 22, 2010 at 4:31 PM, Jason Garrett-Glaser
> <darkshikari at gmail.com> wrote:
>> On Tue, Jun 22, 2010 at 4:05 PM, Jason Garrett-Glaser
>> <darkshikari at gmail.com> wrote:
>>> On Tue, Jun 22, 2010 at 12:35 PM, Ronald S. Bultje <rsbultje at gmail.com> wrote:
>>>> Hi,
>>>>
>>>> as per $subj.
>>>>
>>>> Speed gain:
>>>> - dc_add goes from 1800 to 1350 cycles (where 1150 is overhead,
>>>> measured as empty asm func), so about 3-3.5x faster.
>>>> - The MC functions are each about 4-5x faster (I only measured the 4x4
>>>> ones, the rest I assume are similarly faster but not measured).
>>>> - Total time spent on a shell-script that decodes the whole testsuite
>>>> (vp8-test-vectors-r1, file 001-017) including shell overhead and
>>>> everything goers from 2.3 to 2.1 seconds with these applied.
>>>>
>>>> Results are bit-identical, and this is my first MMX/etc. ever! Thanks
>>>> to Jason for teaching me. ;-).
>>>>
>>>> Ronald
>>>
>>> New patch attached.
>>>
>>> Jason
>>>
>>
>> Now with SSE2 v-filter motion compensation.
>>
>> Jason
>>
>
> Now with full SSE2 MC. ?I also went and updated the x264asm headers
> (and associated asm) to the latest versions. ?This will be split in
> the real commit.
>
> Jason
>

Now with SSSE3 h-filter.  I'm pretty sure SSSE3 is something like 2-3
times faster in this case, though I haven't benched any of it, I'm
just going by the number of instructions.

Jason
-------------- next part --------------
A non-text attachment was scrubbed...
Name: vp8_asm.diff
Type: application/octet-stream
Size: 48675 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100622/7b04292f/attachment.obj>