[FFmpeg-devel] [PATCH] libavcodec/vp9: ipred_dl_32x32_16 avx2 implementation

Ronald S. Bultje rsbultje at gmail.com
Mon Jun 5 15:00:41 EEST 2017


Hi,

On Sun, Jun 4, 2017 at 1:52 PM, Ilia Valiakhmetov <zakne0ne at gmail.com>
wrote:

> vp9_diag_downleft_32x32_8bpp_c: 580.2
> vp9_diag_downleft_32x32_8bpp_sse2: 75.6
> vp9_diag_downleft_32x32_8bpp_ssse3: 73.7
> vp9_diag_downleft_32x32_8bpp_avx: 72.7
> vp9_diag_downleft_32x32_10bpp_c: 1101.2
> vp9_diag_downleft_32x32_10bpp_sse2: 145.4
> vp9_diag_downleft_32x32_10bpp_ssse3: 137.5
> vp9_diag_downleft_32x32_10bpp_avx: 134.8
> vp9_diag_downleft_32x32_10bpp_avx2: 94.0
> vp9_diag_downleft_32x32_12bpp_c: 1108.5
> vp9_diag_downleft_32x32_12bpp_sse2: 145.5
> vp9_diag_downleft_32x32_12bpp_ssse3: 137.3
> vp9_diag_downleft_32x32_12bpp_avx: 135.2
> vp9_diag_downleft_32x32_12bpp_avx2: 94.0
>
> ~30% faster than avx implementation
>
> ---
>  libavcodec/x86/vp9dsp_init_16bpp.c    |  2 ++
>  libavcodec/x86/vp9intrapred_16bpp.asm | 63 ++++++++++++++++++++++++++++++
> +++++
>  2 files changed, 65 insertions(+)


LGTM. I'll keep for comments for another few hours before I push.

Ronald


More information about the ffmpeg-devel mailing list