Re: [FFmpeg-devel] [PATCH 2/2] x86/videodsp: add ff_emu_edge_{hfix, hvar}_avx2

2014-09-24 Thread James Almer
On 24/09/14 3:39 PM, Ronald S. Bultje wrote: > Hi, > > On Wed, Sep 24, 2014 at 8:14 AM, Ronald S. Bultje > wrote: > >> On Tue, Sep 23, 2014 at 5:46 PM, James Almer wrote: >> >>> ~15% faster than sse2 >>> >>> Signed-off-by: James Almer >>> --- >>> ssse3's pshufb didn't prove to be faster, but f

Re: [FFmpeg-devel] [PATCH 2/2] x86/videodsp: add ff_emu_edge_{hfix, hvar}_avx2

2014-09-24 Thread Ronald S. Bultje
Hi, On Wed, Sep 24, 2014 at 8:14 AM, Ronald S. Bultje wrote: > On Tue, Sep 23, 2014 at 5:46 PM, James Almer wrote: > >> ~15% faster than sse2 >> >> Signed-off-by: James Almer >> --- >> ssse3's pshufb didn't prove to be faster, but fortunately avx2's broadcast >> instructions did. >> Maybe we c

Re: [FFmpeg-devel] [PATCH 2/2] x86/videodsp: add ff_emu_edge_{hfix, hvar}_avx2

2014-09-24 Thread Ronald S. Bultje
Hi, On Tue, Sep 23, 2014 at 5:46 PM, James Almer wrote: > ~15% faster than sse2 > > Signed-off-by: James Almer > --- > ssse3's pshufb didn't prove to be faster, but fortunately avx2's broadcast > instructions did. > Maybe we could drop that FIXME altogheter. > Yes let's do that (it was a remin

[FFmpeg-devel] [PATCH 2/2] x86/videodsp: add ff_emu_edge_{hfix, hvar}_avx2

2014-09-23 Thread James Almer
~15% faster than sse2 Signed-off-by: James Almer --- ssse3's pshufb didn't prove to be faster, but fortunately avx2's broadcast instructions did. Maybe we could drop that FIXME altogheter. libavcodec/x86/videodsp.asm| 27 +++ libavcodec/x86/videodsp_init.c | 36