[FFmpeg-devel] [PATCH] avcodec/bitpacked_dec: optimize bitpacked_decode_yuv422p10

[Marton Balint]

On Wed, 13 Dec 2023, Marton Balint wrote:

> From: Devin Heitmueller <devin.heitmueller at ltnglobal.com>
>
> Rework the code a bit to speed up the 10-bit bitpacked decoding
> routine.  This is probably about as fast as I can get it without
> switching to assembly language.
>
> Demonstratable with:
>
> ./ffmpeg -f lavfi -i "smptehdbars=size=3840x2160" -c bitpacked -f image2 -frames:v 1 source.yuv
> ./ffmpeg -f bitpacked -pix_fmt yuv422p10le -s 3840x2160 -c:v bitpacked -i source.yuv -pix_fmt yuv422p10le out.yuv
>
> On my development system, it went from 80ms for a 2160p frame
> down to 20ms (i.e. a 4X speedup).  Good enough for now, I hope...
>
> Comments from Marton:
>
> Originally on my system better performance could be achieved by simply
> switching to the cached bitstream reader, but for Devin it was slower than
> his direct byte operations.
>
> I changed the order of writing output from u/y/v/y to u/v/y/y, and that made
> the code faster than the cached bitstream reader on my system as well.
>
> TIMER measurement of the decode loop on Ryzen 5 3600 with command line:
>
> ./ffmpeg -stream_loop 256 -threads 1 -f bitpacked -pix_fmt yuv422p10le -s 3840x2160 -c:v bitpacked -i source.yuv -pix_fmt yuv422p10le -f null none -loglevel error
>
> Before: 823204127 decicycles in YUV,     256 runs,      0 skips
> After:  315070524 decicycles in YUV,     256 runs,      0 skips
>
> Signed-off-by: Devin Heitmueller <dheitmueller at ltnglobal.com>
> Signed-off-by: Marton Balint <cus at passwd.hu>
> ---
> libavcodec/bitpacked_dec.c | 17 +++++++----------
> 1 file changed, 7 insertions(+), 10 deletions(-)

Will apply.

Regards,
Marton

>
> diff --git a/libavcodec/bitpacked_dec.c b/libavcodec/bitpacked_dec.c
> index c88f861993..54c008bd86 100644
> --- a/libavcodec/bitpacked_dec.c
> +++ b/libavcodec/bitpacked_dec.c
> @@ -28,7 +28,6 @@
>
> #include "avcodec.h"
> #include "codec_internal.h"
> -#include "get_bits.h"
> #include "libavutil/imgutils.h"
> #include "thread.h"
>
> @@ -65,7 +64,7 @@ static int bitpacked_decode_yuv422p10(AVCodecContext *avctx, AVFrame *frame,
> {
>     uint64_t frame_size = (uint64_t)avctx->width * (uint64_t)avctx->height * 20;
>     uint64_t packet_size = (uint64_t)avpkt->size * 8;
> -    GetBitContext bc;
> +    uint8_t *src;
>     uint16_t *y, *u, *v;
>     int ret, i, j;
>
> @@ -79,20 +78,18 @@ static int bitpacked_decode_yuv422p10(AVCodecContext *avctx, AVFrame *frame,
>     if (avctx->width % 2)
>         return AVERROR_PATCHWELCOME;
>
> -    ret = init_get_bits(&bc, avpkt->data, avctx->width * avctx->height * 20);
> -    if (ret)
> -        return ret;
> -
> +    src = avpkt->data;
>     for (i = 0; i < avctx->height; i++) {
>         y = (uint16_t*)(frame->data[0] + i * frame->linesize[0]);
>         u = (uint16_t*)(frame->data[1] + i * frame->linesize[1]);
>         v = (uint16_t*)(frame->data[2] + i * frame->linesize[2]);
>
>         for (j = 0; j < avctx->width; j += 2) {
> -            *u++ = get_bits(&bc, 10);
> -            *y++ = get_bits(&bc, 10);
> -            *v++ = get_bits(&bc, 10);
> -            *y++ = get_bits(&bc, 10);
> +            *u++ = (src[0] << 2) | (src[1] >> 6);
> +            *v++ = ((src[2] << 6) | (src[3] >> 2)) & 0x3ff;
> +            *y++ = ((src[1] << 4) | (src[2] >> 4)) & 0x3ff;
> +            *y++ = ((src[3] << 8) | (src[4]))      & 0x3ff;
> +            src += 5;
>         }
>     }
>
> -- 
> 2.35.3
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
>