Hi,

On Tue, Mar 29, 2011 at 10:52 PM, Oskar Arvidsson <[email protected]> wrote:
> On Tue, Mar 29, 2011 at 09:25:13PM -0700, Ronald S. Bultje wrote:
>> Hi,
>>
>> On Tue, Mar 29, 2011 at 3:23 PM, Oskar Arvidsson <[email protected]> wrote:
>> > On Tue, Mar 29, 2011 at 11:06:50PM +0100, Måns Rullgård wrote:
>> >> Oskar Arvidsson <[email protected]> writes:
>> >>
>> >> > Hi,
>> >> >
>> >> > On Tue, Mar 29, 2011 at 05:48:45PM +0200, Oskar Arvidsson wrote:
>> >> >> Also, I've done some basic benchmarking on how this patchset affects
>> >> >> 8-bit decoding. I'll post the data from those runs separately.
>> >> >
>> >> > Upon request by mru I extended the benchmark somewhat. I didn't bother
>> >> > doing this with oprofile though.
>> >> >
>> >> > 10 runs decoding 8-bit h264:
>> >> >
>> >> >         |  Before           |  After
>> >> > -------------------------------------------
>> >> >         |  8.629            |  8.726
>> >> >         |  8.606            |  8.709
>> >> >         |  8.619            |  8.716
>> >> >         |  8.609            |  8.683
>> >> >         |  8.619            |  8.706
>> >> >         |  8.606            |  8.786
>> >> >         |  8.643            |  8.659
>> >> >         |  8.626            |  8.653
>> >> >         |  8.613            |  8.673
>> >> >         |  8.629            |  8.733
>> >> > -------------------------------------------
>> >> > Avg:    |  8.62             |  8.70
>> >> > Stddev: |  0.012            |  0.040
>> >> > -------------------------------------------
>> >>
>> >> Are those user time or elapsed time?
>> >
>> > user
>> >
>> >>
>> >> > From the table follows that the patchset gives a speed penalty to 8-bit
>> >> > h264 decoding of about 1% (for this clip).
>> >>
>> >> Note that the change is only slightly greater than one stddev of the
>> >> after numbers.  It might be worthwhile testing another clip with
>> >> different characteristics.
>> >
>> > good point.
>> >
>> >> Do you have any idea why it got slower?
>> >
>> > For one, chroma dequant can't be inlined anymore.
>>
>> The effects of that can be undone by SIMD'ifying it. I don't expect it
>> to make it a lot faster than the original, but at least it wouldn't
>> make it slower anymore. I wouldn't worry about this. Jason, Daniel, me
>> or anyone else can quickly undo this speed bump.
>>
>> > Also, h264.c handles pixel size dynamically.
>>
>> Is there anything we can do to decrease the speed effect of that?
>
> We could template some of the most important functions. There only need
> to be two different versions of each of these functions (one for 8-bit
> and one for n-bit), so the size hit wouldn't be too big.

That sounds like a good solution. Can you test if that helps any in
speeding it back up?

Ronald
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to