ffmpeg/libavcodec/vaapi_vp9.c
Andreas Rheinhardt 7bd3b73716 avcodec/vp9: Switch to ProgressFrames
This already fixes a race in the vp9-encparams test. In this test,
side data is added to the current frame after having been decoded
(and therefore after ff_thread_finish_setup() has been called).
Yet the update_thread_context callback called ff_thread_ref_frame()
and therefore av_frame_ref() with this frame as source frame and
the ensuing read was unsynchronised with adding the side data,
i.e. there was a data race.

By switching to the ProgressFrame API the implicit av_frame_ref()
is removed and the race fixed except if this frame is later reused by
a show-existing-frame which uses an explicit av_frame_ref().
The vp9-encparams test does not cover this, so this commit
already fixes all the races in this test.

This decoder kept multiple references to the same ThreadFrames
in the same context and therefore had lots of implicit av_frame_ref()
even when decoding single-threaded. This incurred lots of small
allocations: When decoding an ordinary 10s video in single-threaded
mode the number of allocations reported by Valgrind went down
from 57,814 to 20,908; for 10 threads it went down from 84,223 to
21,901.

Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-04-19 13:18:04 +02:00

186 lines
7.3 KiB
C

/*
* VP9 HW decode acceleration through VA API
*
* Copyright (C) 2015 Timo Rothenpieler <timo@rothenpieler.org>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/pixdesc.h"
#include "hwaccel_internal.h"
#include "vaapi_decode.h"
#include "vp9shared.h"
static VASurfaceID vaapi_vp9_surface_id(const VP9Frame *vf)
{
if (vf)
return ff_vaapi_get_surface_id(vf->tf.f);
else
return VA_INVALID_SURFACE;
}
static int vaapi_vp9_start_frame(AVCodecContext *avctx,
av_unused const uint8_t *buffer,
av_unused uint32_t size)
{
const VP9SharedContext *h = avctx->priv_data;
VAAPIDecodePicture *pic = h->frames[CUR_FRAME].hwaccel_picture_private;
VADecPictureParameterBufferVP9 pic_param;
const AVPixFmtDescriptor *pixdesc = av_pix_fmt_desc_get(avctx->sw_pix_fmt);
int err, i;
pic->output_surface = vaapi_vp9_surface_id(&h->frames[CUR_FRAME]);
pic_param = (VADecPictureParameterBufferVP9) {
.frame_width = avctx->width,
.frame_height = avctx->height,
.pic_fields.bits = {
.subsampling_x = pixdesc->log2_chroma_w,
.subsampling_y = pixdesc->log2_chroma_h,
.frame_type = !h->h.keyframe,
.show_frame = !h->h.invisible,
.error_resilient_mode = h->h.errorres,
.intra_only = h->h.intraonly,
.allow_high_precision_mv = h->h.keyframe ? 0 : h->h.highprecisionmvs,
.mcomp_filter_type = h->h.filtermode ^ (h->h.filtermode <= 1),
.frame_parallel_decoding_mode = h->h.parallelmode,
.reset_frame_context = h->h.resetctx,
.refresh_frame_context = h->h.refreshctx,
.frame_context_idx = h->h.framectxid,
.segmentation_enabled = h->h.segmentation.enabled,
.segmentation_temporal_update = h->h.segmentation.temporal,
.segmentation_update_map = h->h.segmentation.update_map,
.last_ref_frame = h->h.refidx[0],
.last_ref_frame_sign_bias = h->h.signbias[0],
.golden_ref_frame = h->h.refidx[1],
.golden_ref_frame_sign_bias = h->h.signbias[1],
.alt_ref_frame = h->h.refidx[2],
.alt_ref_frame_sign_bias = h->h.signbias[2],
.lossless_flag = h->h.lossless,
},
.filter_level = h->h.filter.level,
.sharpness_level = h->h.filter.sharpness,
.log2_tile_rows = h->h.tiling.log2_tile_rows,
.log2_tile_columns = h->h.tiling.log2_tile_cols,
.frame_header_length_in_bytes = h->h.uncompressed_header_size,
.first_partition_size = h->h.compressed_header_size,
.profile = h->h.profile,
.bit_depth = h->h.bpp,
};
for (i = 0; i < 7; i++)
pic_param.mb_segment_tree_probs[i] = h->h.segmentation.prob[i];
if (h->h.segmentation.temporal) {
for (i = 0; i < 3; i++)
pic_param.segment_pred_probs[i] = h->h.segmentation.pred_prob[i];
} else {
memset(pic_param.segment_pred_probs, 255, sizeof(pic_param.segment_pred_probs));
}
for (i = 0; i < 8; i++) {
if (h->refs[i].f)
pic_param.reference_frames[i] = ff_vaapi_get_surface_id(h->refs[i].f);
else
pic_param.reference_frames[i] = VA_INVALID_ID;
}
err = ff_vaapi_decode_make_param_buffer(avctx, pic,
VAPictureParameterBufferType,
&pic_param, sizeof(pic_param));
if (err < 0) {
ff_vaapi_decode_cancel(avctx, pic);
return err;
}
return 0;
}
static int vaapi_vp9_end_frame(AVCodecContext *avctx)
{
const VP9SharedContext *h = avctx->priv_data;
VAAPIDecodePicture *pic = h->frames[CUR_FRAME].hwaccel_picture_private;
return ff_vaapi_decode_issue(avctx, pic);
}
static int vaapi_vp9_decode_slice(AVCodecContext *avctx,
const uint8_t *buffer,
uint32_t size)
{
const VP9SharedContext *h = avctx->priv_data;
VAAPIDecodePicture *pic = h->frames[CUR_FRAME].hwaccel_picture_private;
VASliceParameterBufferVP9 slice_param;
int err, i;
slice_param = (VASliceParameterBufferVP9) {
.slice_data_size = size,
.slice_data_offset = 0,
.slice_data_flag = VA_SLICE_DATA_FLAG_ALL,
};
for (i = 0; i < 8; i++) {
slice_param.seg_param[i] = (VASegmentParameterVP9) {
.segment_flags.fields = {
.segment_reference_enabled = h->h.segmentation.feat[i].ref_enabled,
.segment_reference = h->h.segmentation.feat[i].ref_val,
.segment_reference_skipped = h->h.segmentation.feat[i].skip_enabled,
},
.luma_dc_quant_scale = h->h.segmentation.feat[i].qmul[0][0],
.luma_ac_quant_scale = h->h.segmentation.feat[i].qmul[0][1],
.chroma_dc_quant_scale = h->h.segmentation.feat[i].qmul[1][0],
.chroma_ac_quant_scale = h->h.segmentation.feat[i].qmul[1][1],
};
memcpy(slice_param.seg_param[i].filter_level, h->h.segmentation.feat[i].lflvl, sizeof(slice_param.seg_param[i].filter_level));
}
err = ff_vaapi_decode_make_slice_buffer(avctx, pic,
&slice_param, sizeof(slice_param),
buffer, size);
if (err) {
ff_vaapi_decode_cancel(avctx, pic);
return err;
}
return 0;
}
const FFHWAccel ff_vp9_vaapi_hwaccel = {
.p.name = "vp9_vaapi",
.p.type = AVMEDIA_TYPE_VIDEO,
.p.id = AV_CODEC_ID_VP9,
.p.pix_fmt = AV_PIX_FMT_VAAPI,
.start_frame = vaapi_vp9_start_frame,
.end_frame = vaapi_vp9_end_frame,
.decode_slice = vaapi_vp9_decode_slice,
.frame_priv_data_size = sizeof(VAAPIDecodePicture),
.init = ff_vaapi_decode_init,
.uninit = ff_vaapi_decode_uninit,
.frame_params = ff_vaapi_common_frame_params,
.priv_data_size = sizeof(VAAPIDecodeContext),
.caps_internal = HWACCEL_CAP_ASYNC_SAFE,
};