From: Neal Gompa <ngompa13@gmail.com>
To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
Cc: David Rosca <nowrep@gmail.com>
Subject: Re: [FFmpeg-devel] [PATCH v2 2/2] lavc/vaapi_av1: Avoid sending the same slice buffer multiple times
Date: Mon, 27 May 2024 16:00:40 -0400
Message-ID: <CAEg-Je-KrOJXih_RMNDLp2g5vj+8X03_4OdR5v=kPpsPbXNeGQ@mail.gmail.com> (raw)
In-Reply-To: <20240508071247.133615-4-nowrep@gmail.com>
On Wed, May 8, 2024 at 3:13 AM David Rosca <nowrep@gmail.com> wrote:
>
> When there are multiple tiles in one slice buffer, use multiple slice
> params to avoid sending the same slice buffer multiple times and thus
> increasing the bitstream size the driver will need to upload to hw.
> ---
> v2: Avoid allocations every slice.
>
> libavcodec/vaapi_av1.c | 47 +++++++++++++++++++++++++++++-------------
> 1 file changed, 33 insertions(+), 14 deletions(-)
>
> diff --git a/libavcodec/vaapi_av1.c b/libavcodec/vaapi_av1.c
> index 4a90db1e09..4ee33a3ae3 100644
> --- a/libavcodec/vaapi_av1.c
> +++ b/libavcodec/vaapi_av1.c
> @@ -19,6 +19,7 @@
> */
>
> #include "libavutil/frame.h"
> +#include "libavutil/mem.h"
> #include "hwaccel_internal.h"
> #include "vaapi_decode.h"
> #include "internal.h"
> @@ -42,6 +43,9 @@ typedef struct VAAPIAV1DecContext {
> */
> VAAPIAV1FrameRef ref_tab[AV1_NUM_REF_FRAMES];
> AVFrame *tmp_frame;
> +
> + int nb_slice_params;
> + VASliceParameterBufferAV1 *slice_params;
> } VAAPIAV1DecContext;
>
> static VASurfaceID vaapi_av1_surface_id(AV1Frame *vf)
> @@ -97,6 +101,8 @@ static int vaapi_av1_decode_uninit(AVCodecContext *avctx)
> for (int i = 0; i < FF_ARRAY_ELEMS(ctx->ref_tab); i++)
> av_frame_free(&ctx->ref_tab[i].frame);
>
> + av_freep(&ctx->slice_params);
> +
> return ff_vaapi_decode_uninit(avctx);
> }
>
> @@ -393,13 +399,24 @@ static int vaapi_av1_decode_slice(AVCodecContext *avctx,
> {
> const AV1DecContext *s = avctx->priv_data;
> VAAPIDecodePicture *pic = s->cur_frame.hwaccel_picture_private;
> - VASliceParameterBufferAV1 slice_param;
> - int err = 0;
> + VAAPIAV1DecContext *ctx = avctx->internal->hwaccel_priv_data;
> + int err, nb_params;
> +
> + nb_params = s->tg_end - s->tg_start + 1;
> + if (ctx->nb_slice_params < nb_params) {
> + ctx->slice_params = av_realloc_array(ctx->slice_params,
> + nb_params,
> + sizeof(*ctx->slice_params));
> + if (!ctx->slice_params) {
> + ctx->nb_slice_params = 0;
> + err = AVERROR(ENOMEM);
> + goto fail;
> + }
> + ctx->nb_slice_params = nb_params;
> + }
>
> for (int i = s->tg_start; i <= s->tg_end; i++) {
> - memset(&slice_param, 0, sizeof(VASliceParameterBufferAV1));
> -
> - slice_param = (VASliceParameterBufferAV1) {
> + ctx->slice_params[i - s->tg_start] = (VASliceParameterBufferAV1) {
> .slice_data_size = s->tile_group_info[i].tile_size,
> .slice_data_offset = s->tile_group_info[i].tile_offset,
> .slice_data_flag = VA_SLICE_DATA_FLAG_ALL,
> @@ -408,18 +425,20 @@ static int vaapi_av1_decode_slice(AVCodecContext *avctx,
> .tg_start = s->tg_start,
> .tg_end = s->tg_end,
> };
> -
> - err = ff_vaapi_decode_make_slice_buffer(avctx, pic, &slice_param, 1,
> - sizeof(VASliceParameterBufferAV1),
> - buffer,
> - size);
> - if (err) {
> - ff_vaapi_decode_cancel(avctx, pic);
> - return err;
> - }
> }
>
> + err = ff_vaapi_decode_make_slice_buffer(avctx, pic, ctx->slice_params, nb_params,
> + sizeof(VASliceParameterBufferAV1),
> + buffer,
> + size);
> + if (err)
> + goto fail;
> +
> return 0;
> +
> +fail:
> + ff_vaapi_decode_cancel(avctx, pic);
> + return err;
> }
>
> const FFHWAccel ff_av1_vaapi_hwaccel = {
> --
> 2.45.0
>
LGTM and works fine with my AMD GPU.
Reviewed-by: Neal Gompa <ngompa13@gmail.com>
--
真実はいつも一つ!/ Always, there's only one truth!
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next prev parent reply other threads:[~2024-05-27 20:01 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-08 7:11 [FFmpeg-devel] [PATCH v2 1/2] lavc/vaapi_decode: Make it possible to send multiple slice params buffers David Rosca
2024-05-08 7:11 ` [FFmpeg-devel] [PATCH v2 2/2] lavc/vaapi_av1: Avoid sending the same slice buffer multiple times David Rosca
2024-05-27 20:00 ` Neal Gompa [this message]
2024-05-27 7:33 ` [FFmpeg-devel] [PATCH v2 1/2] lavc/vaapi_decode: Make it possible to send multiple slice params buffers Xiang, Haihao
2024-05-27 19:59 ` Neal Gompa
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAEg-Je-KrOJXih_RMNDLp2g5vj+8X03_4OdR5v=kPpsPbXNeGQ@mail.gmail.com' \
--to=ngompa13@gmail.com \
--cc=ffmpeg-devel@ffmpeg.org \
--cc=nowrep@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git