[FFmpeg-devel] [RFC] DXVA2 decoding and FFmpeg

[Hendrik Leppkes]

On Thu, May 28, 2015 at 7:39 PM, Stefano Sabatini <stefasab at gmail.com> wrote:
> On date Monday 2015-05-18 13:26:56 +0200, Stefano Sabatini encoded:
>> On Mon, May 18, 2015 at 1:17 PM, Hendrik Leppkes <h.leppkes at gmail.com>
>> wrote:
>>
>> > On Mon, May 18, 2015 at 12:37 PM, Stefano Sabatini <stefasab at gmail.com>
>> > wrote:
>> >
>> [...]
>>
>> > >
>> > > I have a first hackish patch, performed some tests and I got some
>> > > significant performance gains, on my iCore5 with Intel Graphics HD4000 I
>> > > have now the same performance as the software decoder using DXVA2 for
>> > > decoding a H.264 1920x1080 video, but using only a single thread. The
>> > patch
>> > > as is is a hack, since I had to modify the compilation flags to enable
>> > > assembly compilation in the ffmpeg_dxva2.c file. I should probably create
>> > > an optimized copy function in libavutil, comments are welcome.
>> >
>> > FWIW, I never saw any benefits from using a small cache over simply
>> > copying directly to the destination memory, that could potentially
>> > simplify this a bit.
>> >
>>
>>
>> > And yeah, its a huge hack, we don't want new inline assembly.
>> >
>>
>> The sanest approach is probably to add a function to libavutil. The
>> optimized copy would then be accessible to third-party library users, with
>> no assembly hacks involved.
>
> New patch attached, it's still somehow hackish, please advice if you
> consider this approach acceptable.
>

The general concept is fine, but it should not use inline asm, and
someone will want to argue about the name and placement etc... :)