[FFmpeg-devel] [PATCH v6] avcodec/v210dec: add the frame and slice threading support

[Michael Niedermayer]

On Tue, Oct 15, 2019 at 06:37:10PM +0800, lance.lmwang at gmail.com wrote:
> From: Limin Wang <lance.lmwang at gmail.com>
> 
> Threading is to avoid a core cpu being occupied fully with other filters like scale,
> regarding performance, if your cpu frequency is very high, the gain is very small, but
> with more cores and fewer cpu MHz cpus, you will get more improvements.
> 
> The following is my testing results of performance on two different system:
> 1, testing result with my old mac pro
> ./ffmpeg -y -i ./4k_4096_3072.mov -c:v v210 -f rawvideo -frames 10 ./1.v210
> 
> ./ffmpeg -threads 1 -s 4096x3072 -stream_loop 100 -i ./1.v210 -benchmark -f null -
> frame= 1010 fps= 42 q=-0.0 Lsize=N/A time=00:00:40.40 bitrate=N/A speed=1.69x
> 
> patch applied:
> ./ffmpeg -threads 4 -thread_type frame+slice  -s 4096x3072 -stream_loop 100 -i ./1.v210 -benchmark -f null -
> frame= 1010 fps= 55 q=-0.0 Lsize=N/A time=00:00:40.40 bitrate=N/A speed=2.22x
> 
> 2, testing result with x86 server (Intel(R) Xeon(R) CPU E5-2650 v2 @ 2.60GHz):
> ./ffmpeg  -y -i ./4k_3840_2160.ts -c:v v210 -f rawvideo -frames 50 ./2.v210
> 
> ./ffmpeg -threads 1 -s 3840x2160  -stream_loop 20 -i ./2.v210 -benchmark -f null -
> frame= 1050 fps= 80 q=-0.0 Lsize=N/A time=00:00:42.00 bitrate=N/A speed=3.19x
> 
> patch applied:
> ./ffmpeg -threads 2 -thread_type frame+slice  -s 3840x2160  -stream_loop 20 -i ./2.v210 -benchmark -f null -
> frame= 1050 fps=111 q=-0.0 Lsize=N/A time=00:00:42.00 bitrate=N/A speed=4.45x
> 
> ./ffmpeg -threads 4 -thread_type frame+slice  -s 3840x2160  -stream_loop 20 -i ./2.v210 -benchmark -f null -
> frame= 1050 fps=145 q=-0.0 Lsize=N/A time=00:00:42.00 bitrate=N/A speed=5.81x
> 
> Signed-off-by: Limin Wang <lance.lmwang at gmail.com>
> ---
>  libavcodec/v210dec.c | 135 ++++++++++++++++++++++++++++---------------
>  libavcodec/v210dec.h |   1 +
>  2 files changed, 88 insertions(+), 48 deletions(-)
> 
> diff --git a/libavcodec/v210dec.c b/libavcodec/v210dec.c
> index 5a33d8c089..91c2fe0d07 100644
> --- a/libavcodec/v210dec.c
> +++ b/libavcodec/v210dec.c
> @@ -28,6 +28,7 @@
>  #include "libavutil/internal.h"
>  #include "libavutil/mem.h"
>  #include "libavutil/intreadwrite.h"
> +#include "thread.h"
>  
>  #define READ_PIXELS(a, b, c)         \
>      do {                             \
> @@ -37,6 +38,12 @@
>          *c++ = (val >> 20) & 0x3FF;  \
>      } while (0)
>  
> +typedef struct ThreadData {
> +    AVFrame *frame;
> +    uint8_t *buf;
> +    int stride;
> +} ThreadData;
> +
>  static void v210_planar_unpack_c(const uint32_t *src, uint16_t *y, uint16_t *u, uint16_t *v, int width)
>  {
>      uint32_t val;
> @@ -64,21 +71,90 @@ static av_cold int decode_init(AVCodecContext *avctx)
>      avctx->pix_fmt             = AV_PIX_FMT_YUV422P10;
>      avctx->bits_per_raw_sample = 10;
>  
> +    s->thread_count  = av_clip(avctx->thread_count, 1, avctx->height/4);
>      s->aligned_input = 0;
>      ff_v210dec_init(s);
>  
>      return 0;
>  }
>  
> +static int v210_decode_slice(AVCodecContext *avctx, void *arg, int jobnr, int threadnr)
> +{
> +    V210DecContext *s = avctx->priv_data;
> +    int h, w;
> +    ThreadData *td = arg;
> +    AVFrame *frame = td->frame;
> +    int stride = td->stride;
> +    int slice_h = avctx->height / s->thread_count;
> +    int slice_m = avctx->height % s->thread_count;
> +    int slice_start;
> +    int slice_end;
> +    uint8_t *psrc;
> +    uint16_t *y, *u, *v;
> +
> +    slice_start = jobnr * slice_h;
> +    slice_start += FFMIN(jobnr, slice_m);
> +    slice_end = slice_start + slice_h;
> +    if (jobnr < slice_m)
> +        slice_end ++;

I suggest to use code similar to what filters use, for example yadif
    int slice_start = (height *  jobnr   ) / s->thread_count;
    int slice_end   = (height * (jobnr+1)) / s->thread_count;
    
This is simpler

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Whats the most studid thing your enemy could do ? Blow himself up
Whats the most studid thing you could do ? Give up your rights and
freedom because your enemy blew himself up.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20191017/ab147ef8/attachment.sig>