[FFmpeg-devel] [PATCH] SSE dct32()

[Vitor Sessak]

On 07/05/2010 12:55 AM, Michael Niedermayer wrote:
> On Sat, Jul 03, 2010 at 06:47:52PM +0200, Vitor Sessak wrote:
>> On 06/30/2010 10:45 PM, Vitor Sessak wrote:
>>> On 06/30/2010 10:11 PM, Vitor Sessak wrote:
>>>> On 06/30/2010 09:47 PM, Michael Niedermayer wrote:
>>>>> On Wed, Jun 30, 2010 at 06:59:34PM +0200, Vitor Sessak wrote:
>>>>> [...]
>>>>>>>> Anyway, in what does it influences the patch to move dct32() to
>>>>>>>> shared
>>>>>>>> code? New version attached (dct32_common.diff)...
>>>>>>>
>>>>>>> ping?
>>>>>>
>>>>>> No route to host?
>>>>>
>>>>> yes, sorry, patch ok
>>>>> also others (loren/jason) are welcome to help review mp3 optim patches
>>>>> if they want
>>>>
>>>> Applied.
>>>
>>> And now a SSE optimized version.
>>
>> Ping?
>
> its probably possible to improve its speed by interleaving the input from
> several dcts. This way all elements in registers could always be handled equal
> no shuf*
>
> Of course that would be a bit of work to change the surrouding code to
> deal with such blocks efficiently
>
> well, you asked ;)
> iam not seriously suggesting you do that work.
> real review left to jason/loren/ronald/... i have too many patches and other
> issues piled up ATM. in worst case just commit i trust you that your code is
> not that bad to block this.

Applied after discussing it on IRC with Ronald.

-Vitor