[FFmpeg-devel] GPU Hardware Acceleration [was Re: openCL support]

[Patrick Shirkey]

> On Wed, Mar 14, 2012 at 11:56:15PM +0100, Patrick Shirkey wrote:
>>
>> > On Wed, Mar 14, 2012 at 03:23:39AM +0100, Patrick Shirkey wrote:
>> >> Hi,
>> >>
>> >> Maybe it would help if I rephrased the question.
>> >>
>> >> "Are there any parts of FFMPEG that would benefit from Hardware
>> >> Acceleration that do not already have development in process?"
>> >
>> > libavfilter, high quality deinterlace, noise reduce, temporal
>> > interpolation, postprocess (spp in the highest quality setting comes
>> > to mind as something thats relatively simple, very computationally
>> > intensive and highly parallizeable)
>> >
>>
>> Thanks Michael for your valuable feedback.
>>
>> Of these items can we prioritise them for development? They are all
>> useful
>> but which would have the most benefit?
>
> maybe spp as its too slow in its highest quality setting
>

But you're not sure ;-)  Do other devs want to add their 2c?

> also other open source filters should be investigated to see if
> theres something that could be ported to libavfilter that then
> could be a great candidate for hw accel
>

Do you have anything in mind? What should we be looking for?


>
>>
>>
>> > also as a toy project the mandelbrot source but thats less easy i
>> > suspect.
>> >
>>
>> In which way? The algorithms have to be completely reworked?
>
> maybe its not so bad but the parallizeable computations change both
> in number (if the existing optimiztaions are kept) and also
> significantly in how many iterations each needs from frame to frame.
> I suspect this makes things more tricky than for example spp where
> the number and iterations are constant and things are pretty much
> idependant
>


--
Patrick Shirkey
Boost Hardware Ltd