[FFmpeg-devel] [PATCH] ffvorbis, better L1 cache use and simplification of code

[Benoit Fouet]

Hi,

Siarhei Siamashka wrote:
> On Tuesday 30 September 2008, Benoit Fouet wrote:
>   
>> Siarhei Siamashka wrote:
>>     
>>> Interleaved forward/backward channels processing in order to increase
>>> chances of stepping on already cached data for the cores with extremely
>>> small data cache. Ensure that IMDCT per-rotation does not introduce cache
>>> write misses (write misses on random memory accesses are bad for ARM
>>> cores with no write-allocate cache as they prevent combining data in the
>>> write buffer).
>>>
>>> ARM11 with 32K of L1 data cache (no L2) shows performance improvement in
>>> the range 0.5-1% which is not so bad considering that IMDCT/IFFT and also
>>> many other important dsputil functions are not assembly optimized for it
>>> yet.
>>>
>>> According to cachegrind simulation, there might be also some improvement
>>> for x86 cores with 16K of L1 data cache (decrease of the number of cache
>>> misses is most visible in this configuration).
>>>       
>> does this patch depend on another one I missed ?
>>     
>
> It's a single selfcontained patch.
>
>   

ok, good

>> it doesn't apply as is...
>>     
>
> It doesn't apply to what? Patch was generated against FFmpeg revision 15451,
> but also applies fine to revision 15492 which is current SVN HEAD.
>
>   

it seems the problem was the end of lines style...

anyway, patch applied

-- 
Benoit Fouet
Purple Labs S.A.
www.purplelabs.com