[FFmpeg-devel] [PATCH 4/4] VP8: optimize VP8Context struct ordering

[Jason Garrett-Glaser]

2011/3/11 Jason Garrett-Glaser <jason at x264.com>:
> 2011/3/11 M?ns Rullg?rd <mans at mansr.com>:
>> Jason Garrett-Glaser <jason at x264.com> writes:
>>
>>> Shaves at least 3KB off code size on x86, should improve cache utilization.
>>> This would probably be useful to do for other decoders/encoders as well.
>>
>> How much of a difference does this make to performance?
>
> I don't know, I didn't measure and my computer has too high an stddev
> to do so effectively. ?I welcome anyone else to bench it.
>
> Jason
>

Done on another computer, x86_64, where the benefit code-size is a bit
less.  I ran 20 runs of each binary, before and after, and calculated
standard error.

Time before: 9.972s +/- 0.003s
Time after: 9.955s +/- 0.003s

This is about 0.17% +/- 0.06% faster overall.

Jason