[FFmpeg-devel] [PATCH 02/11] x86: dcadsp: implement int8x8_fmul_int32
jamrial at gmail.com
Wed Feb 19 05:36:46 CET 2014
On 18/02/14 7:03 AM, Christophe Gisquet wrote:
> 2014-02-18 8:49 GMT+01:00 Christophe Gisquet <christophe.gisquet at gmail.com>:
>> Only for x86_64 (thus win64 is affected). There, SSE2 is guaranteed,
>> so having an inline SSE2 version is actually equivalent to having the
>> non-inline SSE4 called.
> I meant: the 2 have equivalent speed (ie no benefit with the later function).
Ah, I see. I asked because i was writing an avx2 version of the function and
noticed it was not being called at all. Thanks for confirming it was not a bug.
I went ahead and ported decode_hf to avx2 instead, since int8x8_fmul_int32 is
being phased out.
Will send a patch in a moment.
More information about the ffmpeg-devel