[FFmpeg-devel] [PATCH v2 3/3] avfilter/vf_convolution: Add X86 SIMD optimizations for filter_column()

[Paul B Mahol]

What is status of this?

On 12/22/19, xujunzz at sjtu.edu.cn <xujunzz at sjtu.edu.cn> wrote:
> From: Xu Jun <xujunzz at sjtu.edu.cn>
>
> Performance improves about 10% compared to v1.
>
> Tested using this command:
> ./ffmpeg_g -s 1280*720 -pix_fmt yuv420p -i test.yuv -vf convolution="1 2 3 4
> 5 6 7 8 9:1 2 3 4 5 6 7 8 9:1 2 3 4 5 6 7 8 9:1 2 3 4 5 6 7 8
> 9:1/45:1/45:1/45:1/45:1:2:3:4:column:column:column:column" -an -vframes 5000
> -f null /dev/null -benchmark
>
> after patch:
> frame= 4317 fps=600 q=-0.0 Lsize=N/A time=00:02:52.68 bitrate=N/A speed=
> 24x
> video:2260kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB
> muxing overhead: unknown
> bench: utime=21.540s stime=2.091s rtime=7.197s
>
> before patch:
> frame= 4317 fps=263 q=-0.0 Lsize=N/A time=00:02:52.68 bitrate=N/A
> speed=10.5x
> video:2260kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB
> muxing overhead: unknown
> bench: utime=74.377s stime=1.880s rtime=16.420s
>
> Signed-off-by: Xu Jun <xujunzz at sjtu.edu.cn>
> ---
>  libavfilter/x86/vf_convolution.asm    | 202 ++++++++++++++++++++++++++
>  libavfilter/x86/vf_convolution_init.c |   9 ++
>  2 files changed, 211 insertions(+)