From: Mark Thompson <sw@jkqxz.net>
To: ffmpeg-devel@ffmpeg.org
Subject: Re: [FFmpeg-devel] [PATCH v1 1/7] lavc/vaapi_dec: Create VA parameters dynamically
Date: Mon, 1 Apr 2024 20:52:08 +0100
Message-ID: <e92c1621-6c3b-4c27-85d5-e796cdf21795@jkqxz.net> (raw)
In-Reply-To: <20240328012631.777476-1-fei.w.wang@intel.com>
On 28/03/2024 01:26, fei.w.wang-at-intel.com@ffmpeg.org wrote:
> From: Fei Wang <fei.w.wang@intel.com>
>
> Signed-off-by: Fei Wang <fei.w.wang@intel.com>
> ---
> libavcodec/vaapi_decode.c | 29 ++++++++++++++++++++++-------
> libavcodec/vaapi_decode.h | 7 ++-----
> 2 files changed, 24 insertions(+), 12 deletions(-)
This is because the VVC code is going to want to make a lot more of these param buffers - can we just set a slightly larger fixed limit?
If you always need 20 buffers (say), then this has turned 1 allocation per picture into 3 and used more memory in the non-VVC case as well because of the overhead of that (but if you might variably need up to 200 then this is completely fair).
> diff --git a/libavcodec/vaapi_decode.c b/libavcodec/vaapi_decode.c
> index cca94b5336..1b1972a2a9 100644
> --- a/libavcodec/vaapi_decode.c
> +++ b/libavcodec/vaapi_decode.c
> @@ -38,12 +38,23 @@ int ff_vaapi_decode_make_param_buffer(AVCodecContext *avctx,
> {
> VAAPIDecodeContext *ctx = avctx->internal->hwaccel_priv_data;
> VAStatus vas;
> - VABufferID buffer;
>
> - av_assert0(pic->nb_param_buffers + 1 <= MAX_PARAM_BUFFERS);
> + av_assert0(pic->nb_param_buffers <= pic->param_allocated);
> + if (pic->nb_param_buffers == pic->param_allocated) {
> + pic->param_buffers =
> + av_realloc_array(pic->param_buffers,
> + pic->param_allocated + 16,
> + sizeof(*pic->param_buffers));
Use av_reallocp_array() to avoid leaking the pointer on failure.
> + if (!pic->param_buffers)
> + return AVERROR(ENOMEM);
> +
> + pic->param_allocated += 16;
> + }
> + av_assert0(pic->nb_param_buffers + 1 <= pic->param_allocated);
>
> vas = vaCreateBuffer(ctx->hwctx->display, ctx->va_context,
> - type, size, 1, (void*)data, &buffer);
> + type, size, 1, (void*)data,
> + &pic->param_buffers[pic->nb_param_buffers]);
> if (vas != VA_STATUS_SUCCESS) {
> av_log(avctx, AV_LOG_ERROR, "Failed to create parameter "
> "buffer (type %d): %d (%s).\n",
> @@ -51,14 +62,14 @@ int ff_vaapi_decode_make_param_buffer(AVCodecContext *avctx,
> return AVERROR(EIO);
> }
>
> - pic->param_buffers[pic->nb_param_buffers++] = buffer;
> -
> av_log(avctx, AV_LOG_DEBUG, "Param buffer (type %d, %zu bytes) "
> - "is %#x.\n", type, size, buffer);
> + "is %#x.\n", type, size, pic->param_buffers[pic->nb_param_buffers]);
> +
> + ++pic->nb_param_buffers;
> +
> return 0;
> }
>
> -
> int ff_vaapi_decode_make_slice_buffer(AVCodecContext *avctx,
> VAAPIDecodePicture *pic,
> const void *params_data,
> @@ -215,6 +226,8 @@ fail:
> fail_at_end:
> exit:
> pic->nb_param_buffers = 0;
> + pic->param_allocated = 0;
> + av_freep(&pic->param_buffers);
> pic->nb_slices = 0;
> pic->slices_allocated = 0;
> av_freep(&pic->slice_buffers);
> @@ -228,6 +241,8 @@ int ff_vaapi_decode_cancel(AVCodecContext *avctx,
> ff_vaapi_decode_destroy_buffers(avctx, pic);
>
> pic->nb_param_buffers = 0;
> + pic->param_allocated = 0;
> + av_freep(&pic->param_buffers);
> pic->nb_slices = 0;
> pic->slices_allocated = 0;
> av_freep(&pic->slice_buffers);
> diff --git a/libavcodec/vaapi_decode.h b/libavcodec/vaapi_decode.h
> index 6beda14e52..a41d7ff2ff 100644
> --- a/libavcodec/vaapi_decode.h
> +++ b/libavcodec/vaapi_decode.h
> @@ -32,15 +32,12 @@ static inline VASurfaceID ff_vaapi_get_surface_id(AVFrame *pic)
> return (uintptr_t)pic->data[3];
> }
>
> -enum {
> - MAX_PARAM_BUFFERS = 16,
> -};
> -
> typedef struct VAAPIDecodePicture {
> VASurfaceID output_surface;
>
> int nb_param_buffers;
> - VABufferID param_buffers[MAX_PARAM_BUFFERS];
> + VABufferID *param_buffers;
Previously the array was zeroed at allocation but now it isn't. Can you confirm that that isn't a problem?
> + int param_allocated;
Maybe "nb_param_buffers_allocated" would be clearer.
>
> int nb_slices;
> VABufferID *slice_buffers;
Thanks,
- Mark
_______________________________________________
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:[~2024-04-01 19:51 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-03-28 1:26 fei.w.wang-at-intel.com
2024-03-28 1:26 ` [FFmpeg-devel] [PATCH v1 2/7] lavc/vvc_refs: Move definition of VVC_FRAME_FLAG* to h header fei.w.wang-at-intel.com
2024-03-28 1:26 ` [FFmpeg-devel] [PATCH v1 3/7] lavc/cbs_h266: Add SliceTopLeftTileIdx to H266RawPPS fei.w.wang-at-intel.com
2024-03-28 1:26 ` [FFmpeg-devel] [PATCH v1 4/7] lavc/cbs_h266: Add NumSlicesInTile " fei.w.wang-at-intel.com
2024-03-28 1:26 ` [FFmpeg-devel] [PATCH v1 5/7] lavc/vvc_ps: Add alf raw syntax into VVCALF fei.w.wang-at-intel.com
2024-03-28 1:26 ` [FFmpeg-devel] [PATCH v1 6/7] lavc/vvc_dec: Add hardware decode API fei.w.wang-at-intel.com
2024-03-28 2:04 ` Andreas Rheinhardt
2024-04-02 6:24 ` Wang, Fei W
2024-03-28 1:26 ` [FFmpeg-devel] [PATCH v1 7/7] lavc/vaapi_dec: Add VVC decoder fei.w.wang-at-intel.com
2024-04-02 12:48 ` Nuo Mi
2024-04-03 3:31 ` Wang, Fei W
2024-04-06 5:03 ` Nuo Mi
2024-04-01 19:52 ` Mark Thompson [this message]
2024-04-02 6:16 ` [FFmpeg-devel] [PATCH v1 1/7] lavc/vaapi_dec: Create VA parameters dynamically Wang, Fei W
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=e92c1621-6c3b-4c27-85d5-e796cdf21795@jkqxz.net \
--to=sw@jkqxz.net \
--cc=ffmpeg-devel@ffmpeg.org \
/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