Re: [FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add SSE2 shuffle_bytes functions

2024-06-10 Thread Andreas Rheinhardt
James Almer: > On 6/9/2024 12:36 PM, James Almer wrote: >>> So I would not add these SSE2 versions. But the one missing SSSE3 >>> version (shuffle_bytes_2103_ssse3) is of course worth it. >> >> I will look into that. > > I'm not sure why you said it's missing, because it's there. Sorry for having

Re: [FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add SSE2 shuffle_bytes functions

2024-06-10 Thread James Almer
On 6/9/2024 12:36 PM, James Almer wrote: So I would not add these SSE2 versions. But the one missing SSSE3 version (shuffle_bytes_2103_ssse3) is of course worth it. I will look into that. I'm not sure why you said it's missing, because it's there. _

Re: [FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add SSE2 shuffle_bytes functions

2024-06-09 Thread Rémi Denis-Courmont
Le sunnuntaina 9. kesäkuuta 2024, 18.36.35 EEST James Almer a écrit : > I obviously don't have that kind of information. You'd need to look at > things like Steam's, Firefox's or Chrome's hardware surveys. As discussed on IRC yesterday, Steam claims that 106.85% of processors support SSE2 (and as

Re: [FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add SSE2 shuffle_bytes functions

2024-06-09 Thread James Almer
On 6/8/2024 12:55 PM, Andreas Rheinhardt wrote: James Almer: On 6/6/2024 11:48 AM, Andreas Rheinhardt wrote: James Almer: And remove shuffle_bytes_2103_mmxext. shuffle_bytes_0321_c: 28.1 shuffle_bytes_0321_sse2: 13.6 shuffle_bytes_0321_ssse3: 9.6 shuffle_bytes_0321_avx2: 7.1 shuffle_bytes_123

Re: [FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add SSE2 shuffle_bytes functions

2024-06-08 Thread Rémi Denis-Courmont
Le lauantaina 8. kesäkuuta 2024, 18.55.53 EEST Andreas Rheinhardt a écrit : > I think we should not optimize for CPUs that do not even have x86-64 v2. > So I would not add these SSE2 versions. We certainly should consider ditching SSE2 where SSSE3 is available now or in the near future. But in th

Re: [FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add SSE2 shuffle_bytes functions

2024-06-08 Thread Andreas Rheinhardt
James Almer: > On 6/6/2024 11:48 AM, Andreas Rheinhardt wrote: >> James Almer: >>> And remove shuffle_bytes_2103_mmxext. >>> >>> shuffle_bytes_0321_c: 28.1 >>> shuffle_bytes_0321_sse2: 13.6 >>> shuffle_bytes_0321_ssse3: 9.6 >>> shuffle_bytes_0321_avx2: 7.1 >>> shuffle_bytes_1230_c: 52.6 >>> shuffle

Re: [FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add SSE2 shuffle_bytes functions

2024-06-06 Thread James Almer
On 6/6/2024 11:48 AM, Andreas Rheinhardt wrote: James Almer: And remove shuffle_bytes_2103_mmxext. shuffle_bytes_0321_c: 28.1 shuffle_bytes_0321_sse2: 13.6 shuffle_bytes_0321_ssse3: 9.6 shuffle_bytes_0321_avx2: 7.1 shuffle_bytes_1230_c: 52.6 shuffle_bytes_1230_sse2: 12.1 shuffle_bytes_1230_ssse

Re: [FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add SSE2 shuffle_bytes functions

2024-06-06 Thread Andreas Rheinhardt
James Almer: > And remove shuffle_bytes_2103_mmxext. > > shuffle_bytes_0321_c: 28.1 > shuffle_bytes_0321_sse2: 13.6 > shuffle_bytes_0321_ssse3: 9.6 > shuffle_bytes_0321_avx2: 7.1 > shuffle_bytes_1230_c: 52.6 > shuffle_bytes_1230_sse2: 12.1 > shuffle_bytes_1230_ssse3: 8.6 > shuffle_bytes_1230_avx2:

[FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add SSE2 shuffle_bytes functions

2024-06-06 Thread James Almer
And remove shuffle_bytes_2103_mmxext. shuffle_bytes_0321_c: 28.1 shuffle_bytes_0321_sse2: 13.6 shuffle_bytes_0321_ssse3: 9.6 shuffle_bytes_0321_avx2: 7.1 shuffle_bytes_1230_c: 52.6 shuffle_bytes_1230_sse2: 12.1 shuffle_bytes_1230_ssse3: 8.6 shuffle_bytes_1230_avx2: 6.6 shuffle_bytes_2103_c: 29.1 s