theora - Oct 2009 - Hardware Acelerating Encoding

Hi all,

What does everyone think about using OpenCL to improve the performance
of the encoder? This is particularly important for Live HD scenarios.
Clearly off loading some of the work to the GPU would have a
significant effect on the performance. Anyone got any suggestions of
which part of the encoding process would benefit most form off loading
to the GPU?

Out of interest is the current encoder parallelized to take advantage
of multicore CPUs?

Thanks

Charlie M

On Oct 18, 2009, at 11:29 AM, Charles Mason <charlie.mas at gmail.com>  
wrote:
> Clearly off loading some of the work to the GPU would have a
> significant effect on the performance.
I suppose, but I don't find HD Theora videos using up that much of my  
CPU in the first place.

On Oct 18, 2009, at 12:41 PM, Charles Mason <charlie.mas at gmail.com>  
wrote:
>>> Clearly off loading some of the work to the GPU would have a
>>> significant effect on the performance.
>>
>> I suppose, but I don't find HD Theora videos using up that much of
my
>> CPU in the first place.
>
> Do you mean for encoding or decoding. I find decoding even 1080 HD
> Theora fine on a moderatly powerful machine.
>
> Encoding is quite another thing, I know its got a lot better with the
> new encoder. Are you finding it quick enough for Live 1080 encoding?
Decoding. My experience is that it's not very CPU intensive,  
especially when compared against High Profile H.264 at 1080p. That  
makes my fans speed up but not Theora.

I do alot of encoding but none of it live. The newer version does seem  
slower, although I haven't done any formal testing. Maybe there's  
something to be said for hardware acceleration. If you're up for it, I  
say go ahead. :)