> -----Original Message-----
> From: Wang, Fei W
> Sent: Wednesday, September 30, 2020 1:58 PM
> To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
> Subject: RE: [FFmpeg-devel] [PATCH V3 2/2] avcodec: add av1 VAAPI decoder
>
> > -----Original Message-----
> > From: ffmpeg-devel <ffmpeg-devel-boun...@ffmpeg.org> On Behalf Of Mark
> > Thompson
> > Sent: Wednesday, September 30, 2020 1:02 AM
> > To: ffmpeg-devel@ffmpeg.org
> > Subject: Re: [FFmpeg-devel] [PATCH V3 2/2] avcodec: add av1 VAAPI
> > decoder
> >
> > On 09/09/2020 04:39, Fei Wang wrote:
> > > Example cmdline:
> > > ffmpeg -hwaccel vaapi -hwaccel_device /dev/dri/renderD128 -v verbose \
> > > -c:v av1 -i input.ivf -pix_fmt yuv420p -vsync passthrough -f md5 \
> > > -y out.md5
> > >
> > > Signed-off-by: Fei Wang <fei.w.w...@intel.com>
> > > ---
> > > Changelog | 1 +
> > > configure | 3 +
> > > libavcodec/Makefile | 1 +
> > > libavcodec/av1dec.c | 23 ++-
> > > libavcodec/hwaccels.h | 1 +
> > > libavcodec/vaapi_av1.c | 285
> > ++++++++++++++++++++++++++++++++++++++
> > > libavcodec/vaapi_decode.c | 8 ++
> > > libavcodec/version.h | 2 +-
> > > 8 files changed, 322 insertions(+), 2 deletions(-)
> > > create mode 100644 libavcodec/vaapi_av1.c
> > >
> > > diff --git a/Changelog b/Changelog
> > > index a89010d7e1..fb1fd55571 100644
> > > --- a/Changelog
> > > +++ b/Changelog
> > > @@ -22,6 +22,7 @@ version <next>:
> > > - MODS demuxer
> > > - PhotoCD decoder
> > > - AV1 decoder (Hardware acceleration used only)
> > > +- AV1 VAAPI decoder
> > >
> > >
> > > version 4.3:
> > > diff --git a/configure b/configure
> > > index 5893ae3473..81f3c80e8e 100755
> > > --- a/configure
> > > +++ b/configure
> > > @@ -2917,6 +2917,8 @@ videotoolbox_hwaccel_deps="videotoolbox
> > pthreads"
> > > videotoolbox_hwaccel_extralibs="-framework QuartzCore"
> > > xvmc_deps="X11_extensions_XvMClib_h"
> > >
> > > +av1_vaapi_hwaccel_deps="vaapi_1_8"
> > > +av1_vaapi_hwaccel_select="av1_decoder"
> > > h263_vaapi_hwaccel_deps="vaapi"
> > > h263_vaapi_hwaccel_select="h263_decoder"
> > > h263_videotoolbox_hwaccel_deps="videotoolbox"
> > > @@ -6680,6 +6682,7 @@ if enabled vaapi; then
> > > fi
> > >
> > > check_cpp_condition vaapi_1 "va/va.h" "VA_CHECK_VERSION(1, 0, 0)"
> > > + check_cpp_condition vaapi_1_8 "va/va.h" "VA_CHECK_VERSION(1, 8, 0)"
> >
> > Add a check for the AV1 structure needed like the other codecs below
> > rather than doing this. The VAAPI_1 version macro exists for the API
> > incompatibilities between VAAPI versions 0 and 1.
> >
> > >
> > > check_type "va/va.h va/va_dec_hevc.h"
> "VAPictureParameterBufferHEVC"
> > > check_struct "va/va.h" "VADecPictureParameterBufferVP9"
> > > bit_depth diff --git a/libavcodec/Makefile b/libavcodec/Makefile
> > > index
> > > 5729cdd8e7..4d54ddc684 100644
> > > --- a/libavcodec/Makefile
> > > +++ b/libavcodec/Makefile
> > > @@ -954,6 +954,7 @@ OBJS-$(CONFIG_VP9_NVDEC_HWACCEL) +=
> > nvdec_vp9.o
> > > OBJS-$(CONFIG_VP9_VAAPI_HWACCEL) += vaapi_vp9.o
> > > OBJS-$(CONFIG_VP9_VDPAU_HWACCEL) += vdpau_vp9.o
> > > OBJS-$(CONFIG_VP8_QSV_HWACCEL) += qsvdec_other.o
> > > +OBJS-$(CONFIG_AV1_VAAPI_HWACCEL) += vaapi_av1.o
> >
> > This should be in alphabetical order, so put it at the top of the hwaccel
> > section.
> >
> > >
> > > # libavformat dependencies
> > > OBJS-$(CONFIG_ISO_MEDIA) += mpeg4audio.o mpegaudiodata.o
> > > diff --git a/libavcodec/av1dec.c b/libavcodec/av1dec.c index
> > > eeeec2ca32..4098cbb655 100644
> > > --- a/libavcodec/av1dec.c
> > > +++ b/libavcodec/av1dec.c
> > > @@ -262,7 +262,7 @@ static int get_pixel_format(AVCodecContext *avctx)
> > > uint8_t bit_depth;
> > > int ret;
> > > enum AVPixelFormat pix_fmt = AV_PIX_FMT_NONE; -#define
> > > HWACCEL_MAX (0)
> > > +#define HWACCEL_MAX (CONFIG_AV1_VAAPI_HWACCEL)
> > > enum AVPixelFormat pix_fmts[HWACCEL_MAX + 1], *fmtp =
> > > pix_fmts;
> > >
> > > if (seq->seq_profile == 2 && seq->color_config.high_bitdepth)
> > > @@
> > > -323,6 +323,24 @@ static int get_pixel_format(AVCodecContext *avctx)
> > > return -1;
> > > s->pix_fmt = pix_fmt;
> > >
> > > + switch (s->pix_fmt) {
> > > + case AV_PIX_FMT_YUV420P:
> > > +#if CONFIG_AV1_VAAPI_HWACCEL
> > > + *fmtp++ = AV_PIX_FMT_VAAPI; #endif
> > > + break;
> > > + case AV_PIX_FMT_YUV420P10:
> > > +#if CONFIG_AV1_VAAPI_HWACCEL
> > > + *fmtp++ = AV_PIX_FMT_VAAPI; #endif
> > > + break;
> > > + case AV_PIX_FMT_YUV420P12:
> > > +#if CONFIG_AV1_VAAPI_HWACCEL
> > > + *fmtp++ = AV_PIX_FMT_VAAPI; #endif
> > > + break;
> > > + }
> >
> > Are you sure this combination is right? It doesn't match any profile
> > (4:2:0 8- and 10-bit are profile 0, but 12-bit requires profile 2), so
> > checking for support will be confusing.
>
> Actually s->pix_fmt is get from profile and bit depth read from sequence
> header,
> so it doesn't need to involve profile again here, just use s->pix_fmt to set
> hardware pix fmt.
>
> >
> > > +
> > > *fmtp = AV_PIX_FMT_NONE;
> > > avctx->sw_pix_fmt = s->pix_fmt;
> > > ret = ff_thread_get_format(avctx, pix_fmts); @@ -855,6 +873,9
> > > @@ AVCodec ff_av1_decoder = {
> > > .flush = av1_decode_flush,
> > > .profiles = NULL_IF_CONFIG_SMALL(ff_av1_profiles),
> > > .hw_configs = (const AVCodecHWConfigInternal * []) {
> > > +#if CONFIG_AV1_VAAPI_HWACCEL
> > > + HWACCEL_VAAPI(av1),
> > > +#endif
> > > NULL
> > > },
> > > };
> > > diff --git a/libavcodec/hwaccels.h b/libavcodec/hwaccels.h index
> > > 6109c89bd6..18e9079c55 100644
> > > --- a/libavcodec/hwaccels.h
> > > +++ b/libavcodec/hwaccels.h
> > > @@ -21,6 +21,7 @@
> > >
> > > #include "avcodec.h"
> > >
> > > +extern const AVHWAccel ff_av1_vaapi_hwaccel;
> > > extern const AVHWAccel ff_h263_vaapi_hwaccel;
> > > extern const AVHWAccel ff_h263_videotoolbox_hwaccel;
> > > extern const AVHWAccel ff_h264_d3d11va_hwaccel; diff --git
> > > a/libavcodec/vaapi_av1.c b/libavcodec/vaapi_av1.c new file mode
> > > 100644 index 0000000000..c5aeaf8b21
> > > --- /dev/null
> > > +++ b/libavcodec/vaapi_av1.c
> > > @@ -0,0 +1,285 @@
> > > +/*
> > > + * AV1 HW decode acceleration through VA API
> > > + *
> > > + * 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 "hwconfig.h"
> > > +#include "vaapi_decode.h"
> > > +#include "av1dec.h"
> > > +
> > > +static VASurfaceID vaapi_av1_surface_id(AV1Frame *vf) {
> > > + if (vf)
> > > + return ff_vaapi_get_surface_id(vf->tf.f);
> > > + else
> > > + return VA_INVALID_SURFACE;
> > > +}
> > > +
> > > +static VAAV1TransformationType vaapi_av1_get_gm_type(uint8_t type) {
> > > + if (type == AV1_WARP_MODEL_IDENTITY)
> > > + return VAAV1TransformationIdentity;
> > > + else if (type == AV1_WARP_MODEL_TRANSLATION)
> > > + return VAAV1TransformationTranslation;
> > > + else if (type == AV1_WARP_MODEL_ROTZOOM)
> > > + return VAAV1TransformationRotzoom;
> > > + else if (type == AV1_WARP_MODEL_AFFINE)
> > > + return VAAV1TransformationAffine;
> > > + else
> > > + return VAAV1TransformationIdentity; }
> >
> > Does this identity function really need to exist?
> >
> > > +
> > > +static uint8_t vaapi_av1_get_bit_depth_idx(AVCodecContext *avctx) {
> > > + AV1DecContext *s = avctx->priv_data;
> > > + const AV1RawSequenceHeader *seq = s->raw_seq;
> > > + uint8_t bit_depth = 8;
> > > +
> > > + if (seq->seq_profile == 2 && seq->color_config.high_bitdepth)
> > > + bit_depth = seq->color_config.twelve_bit ? 12 : 10;
> > > + else if (seq->seq_profile <= 2)
> > > + bit_depth = seq->color_config.high_bitdepth ? 10 : 8;
> > > + else {
> > > + av_log(avctx, AV_LOG_ERROR,
> > > + "Couldn't get bit depth from profile:%d.\n",
> > > + seq->seq_profile);
> >
> > I think you want to eliminate this case earlier to avoid it. If you
> > do get here then you can't continue usefully, so this would need to return
> > an
> error.
> >
> > > + }
> > > + return bit_depth == 8 ? 0 : bit_depth == 10 ? 1 : 2; }
> > > +
> > > +static int vaapi_av1_start_frame(AVCodecContext *avctx,
> > > + av_unused const uint8_t *buffer,
> > > + av_unused uint32_t size) {
> > > + AV1DecContext *s = avctx->priv_data;
> > > + const AV1RawSequenceHeader *seq = s->raw_seq;
> > > + const AV1RawFrameHeader *frame_header = s->raw_frame_header;
> > > + VAAPIDecodePicture *pic = s->cur_frame.hwaccel_picture_private;
> > > + VADecPictureParameterBufferAV1 pic_param;
> > > + int err = 0;
> > > + uint8_t lr_type[4] = {AV1_RESTORE_NONE, AV1_RESTORE_SWITCHABLE,
> > > +AV1_RESTORE_WIENER, AV1_RESTORE_SGRPROJ};
> >
> > This is the remap_lr_type[] table from the standard? Maybe use that name.
> >
> > > +
> > > + pic->output_surface = vaapi_av1_surface_id(&s->cur_frame);
> > > +
> > > + memset(&pic_param, 0, sizeof(VADecPictureParameterBufferAV1));
> > > + pic_param = (VADecPictureParameterBufferAV1) {
> > > + .profile = seq->seq_profile,
> > > + .order_hint_bits_minus_1 = seq->order_hint_bits_minus_1,
> > > + .bit_depth_idx = vaapi_av1_get_bit_depth_idx(avctx),
> > > + .seq_info_fields.fields = {
> > > + .still_picture = seq->still_picture,
> > > + .use_128x128_superblock = seq->use_128x128_superblock,
> > > + .enable_filter_intra = seq->enable_filter_intra,
> > > + .enable_intra_edge_filter = seq->enable_intra_edge_filter,
> > > + .enable_interintra_compound =
> > > seq->enable_interintra_compound,
> > > + .enable_masked_compound = seq->enable_masked_compound,
> > > + .enable_dual_filter = seq->enable_dual_filter,
> > > + .enable_order_hint = seq->enable_order_hint,
> > > + .enable_jnt_comp = seq->enable_jnt_comp,
> > > + .enable_cdef = seq->enable_cdef,
> > > + .mono_chrome = seq->color_config.mono_chrome,
> > > + .color_range = seq->color_config.color_range,
> > > + .subsampling_x = seq->color_config.subsampling_x,
> > > + .subsampling_y = seq->color_config.subsampling_y,
> > > + .chroma_sample_position = seq-
> > >color_config.chroma_sample_position,
> > > + .film_grain_params_present =
> > > + seq->film_grain_params_present,
> >
> > These tables would be easier to read with some horizontal alignment at the
> > '='
> > (see other vaapi_codec.c files).
> >
> > > + },
> > > + .current_frame = pic->output_surface,
> > > + .current_display_picture = pic->output_surface,
> > > + .frame_width_minus1 = frame_header->frame_width_minus_1,
> > > + .frame_height_minus1 = frame_header->frame_height_minus_1,
> > > + .primary_ref_frame = frame_header->primary_ref_frame,
> > > + .order_hint = frame_header->order_hint,
> > > + .seg_info.segment_info_fields.bits = {
> > > + .enabled = frame_header->segmentation_enabled,
> > > + .update_map = frame_header->segmentation_update_map,
> > > + .temporal_update =
> > > frame_header->segmentation_temporal_update,
> > > + .update_data = frame_header->segmentation_update_data,
> > > + },
> > > + .film_grain_info.film_grain_info_fields.bits = {
> > > + .apply_grain = frame_header->apply_grain,
> > > + .chroma_scaling_from_luma = frame_header-
> > >chroma_scaling_from_luma,
> > > + .grain_scaling_minus_8 = frame_header->grain_scaling_minus_8,
> > > + .ar_coeff_lag = frame_header->ar_coeff_lag,
> > > + .ar_coeff_shift_minus_6 =
> > > frame_header->ar_coeff_shift_minus_6,
> > > + .grain_scale_shift = frame_header->grain_scale_shift,
> > > + .overlap_flag = frame_header->overlap_flag,
> > > + .clip_to_restricted_range =
> > > frame_header->clip_to_restricted_range,
> > > + },
> > > + .tile_cols = frame_header->tile_cols,
> > > + .tile_rows = frame_header->tile_rows,
> > > + .context_update_tile_id = frame_header->context_update_tile_id,
> > > + .pic_info_fields.bits = {
> > > + .frame_type = frame_header->frame_type,
> > > + .show_frame = frame_header->show_frame,
> > > + .showable_frame = frame_header->showable_frame,
> > > + .error_resilient_mode = frame_header->error_resilient_mode,
> > > + .disable_cdf_update = frame_header->disable_cdf_update,
> > > + .allow_screen_content_tools = frame_header-
> > >allow_screen_content_tools,
> > > + .force_integer_mv = frame_header->force_integer_mv,
> > > + .allow_intrabc = frame_header->allow_intrabc,
> > > + .use_superres = frame_header->use_superres,
> > > + .allow_high_precision_mv = frame_header-
> >allow_high_precision_mv,
> > > + .is_motion_mode_switchable = frame_header-
> > >is_motion_mode_switchable,
> > > + .use_ref_frame_mvs = frame_header->use_ref_frame_mvs,
> > > + .disable_frame_end_update_cdf = frame_header-
> > >disable_frame_end_update_cdf,
> > > + .uniform_tile_spacing_flag = frame_header-
> > >uniform_tile_spacing_flag,
> > > + .allow_warped_motion = frame_header->allow_warped_motion,
> > > + },
> > > + .interp_filter = frame_header->interpolation_filter,
> > > + .filter_level[0] = frame_header->loop_filter_level[0],
> > > + .filter_level[1] = frame_header->loop_filter_level[1],
> > > + .filter_level_u = frame_header->loop_filter_level[2],
> > > + .filter_level_v = frame_header->loop_filter_level[3],
> > > + .loop_filter_info_fields.bits = {
> > > + .sharpness_level = frame_header->loop_filter_sharpness,
> > > + .mode_ref_delta_enabled =
> > > s->cur_frame.loop_filter_delta_enabled,
> > > + .mode_ref_delta_update =
> > > frame_header->loop_filter_delta_update,
> > > + },
> > > + .base_qindex = frame_header->base_q_idx,
> > > + .qmatrix_fields.bits.using_qmatrix = frame_header->using_qmatrix,
> > > + .mode_control_fields.bits = {
> > > + .delta_q_present_flag = frame_header->delta_q_present,
> > > + .log2_delta_q_res = frame_header->delta_q_res,
> > > + .tx_mode = frame_header->tx_mode,
> > > + .reference_select = frame_header->reference_select,
> > > + .reduced_tx_set_used = frame_header->reduced_tx_set,
> > > + .skip_mode_present = frame_header->skip_mode_present,
> > > + },
> > > + .cdef_damping_minus_3 = frame_header->cdef_damping_minus_3,
> > > + .cdef_bits = frame_header->cdef_bits,
> > > + .loop_restoration_fields.bits = {
> > > + .yframe_restoration_type = lr_type[frame_header->lr_type[0]],
> > > + .cbframe_restoration_type =
> > > lr_type[frame_header->lr_type[1]],
> > > + .crframe_restoration_type =
> > > lr_type[frame_header->lr_type[2]],
> > > + .lr_unit_shift = frame_header->lr_unit_shift,
> > > + .lr_uv_shift = frame_header->lr_uv_shift,
> > > + },
> > > + };
> > > +
> > > + for (int i = 0; i < 8; i++) {
> >
> > The magic number would be clearer as AV1_NUM_REF_FRAMES.
> >
> > > + if (pic_param.pic_info_fields.bits.frame_type == AV1_FRAME_KEY)
> > > + pic_param.ref_frame_map[i] = VA_INVALID_ID;
> > > + else
> > > + pic_param.ref_frame_map[i] =
> > > vaapi_av1_surface_id(&s->ref[i]);
> > > + av_log(avctx, AV_LOG_DEBUG, "Vaapi set ref surface id: %d -
> > > 0x:%x\n",
> > > + i, pic_param.ref_frame_map[i]);
> > > + }
> > > + for (int i = 0; i < 7; i++) {
> >
> > AV_REFS_PER_FRAME
> >
> > > + pic_param.ref_frame_idx[i] = frame_header->ref_frame_idx[i];
> > > + }
> > > + for (int i = 0; i < 8; i++) {
> >
> > AV_TOTAL_REFS_PER_FRAME
> >
> > > + pic_param.ref_deltas[i] = s->cur_frame.loop_filter_ref_deltas[i];
> > > + }
> > > + for (int i = 0; i < 2; i++) {
> > > + pic_param.mode_deltas[i] =
> > > s->cur_frame.loop_filter_mode_deltas[i];
> > > + }
> > > + for (int i = 0; i < 8; i++) {
> >
> > 1 << cdef_bits? The other values aren't set.
> >
> > > + pic_param.cdef_y_strengths[i] =
> > > + (frame_header->cdef_y_pri_strength[i] << 2) +
> > > + frame_header->cdef_y_sec_strength[i];
> > > + pic_param.cdef_uv_strengths[i] =
> > > + (frame_header->cdef_uv_pri_strength[i] << 2) +
> > > + frame_header->cdef_uv_sec_strength[i];
> > > + }
> > > + for (int i = 0; i < frame_header->tile_cols; i++) {
> > > + pic_param.width_in_sbs_minus_1[i] =
> > > + frame_header->width_in_sbs_minus_1[i];
> > > + }
> > > + for (int i = 0; i < frame_header->tile_rows; i++) {
> > > + pic_param.height_in_sbs_minus_1[i] =
> > > + frame_header->height_in_sbs_minus_1[i];
> > > + }
> > > + for (int i = AV1_REF_FRAME_LAST; i <= AV1_REF_FRAME_ALTREF; i++) {
> > > + pic_param.wm[i - 1].wmtype =
> > > + vaapi_av1_get_gm_type(s->cur_frame.gm_type[i]);
> > > + for (int j=0; j<6; j++)
> > > + pic_param.wm[i - 1].wmmat[j] = s->cur_frame.gm_params[i][j];
> > > + }
> > > + err = ff_vaapi_decode_make_param_buffer(avctx, pic,
> > > + VAPictureParameterBufferType,
> > > + &pic_param,
> > > sizeof(pic_param));
> > > + if (err < 0)
> > > + goto fail;
> > > +
> > > + return 0;
> > > +
> > > +fail:
> > > + ff_vaapi_decode_cancel(avctx, pic);
> > > + return err;
> > > +}
> > > +
> > > +static int vaapi_av1_end_frame(AVCodecContext *avctx) {
> > > + const AV1DecContext *s = avctx->priv_data;
> > > + VAAPIDecodePicture *pic = s->cur_frame.hwaccel_picture_private;
> > > + return ff_vaapi_decode_issue(avctx, pic); }
> > > +
> > > +static int vaapi_av1_decode_slice(AVCodecContext *avctx,
> > > + const uint8_t *buffer,
> > > + uint32_t size) {
> > > + const AV1DecContext *s = avctx->priv_data;
> > > + VAAPIDecodePicture *pic = s->cur_frame.hwaccel_picture_private;
> > > + VASliceParameterBufferAV1 slice_param;
> > > + uint32_t offset;
> > > + int err = 0;
> > > +
> > > + for (int i = s->tg_start; i <= s->tg_end; i++) {
> > > + memset(&slice_param, 0, sizeof(VASliceParameterBufferAV1));
> > > + offset = s->tile_group_info[i].tile_offset;
> > > +
> > > + slice_param = (VASliceParameterBufferAV1) {
> > > + .slice_data_size = s->tile_group_info[i].tile_size,
> > > + .slice_data_offset = 0,
> > > + .slice_data_flag = VA_SLICE_DATA_FLAG_ALL,
> > > + .tile_row = s->tile_group_info[i].tile_row,
> > > + .tile_column = s->tile_group_info[i].tile_column,
> > > + .tg_start = s->tg_start,
> > > + .tg_end = s->tg_end,
> >
> > Horizontal alignment would be nice again.
> >
> > > + };
> > > +
> > > + err = ff_vaapi_decode_make_slice_buffer(avctx, pic, &slice_param,
> > > +
> > > sizeof(VASliceParameterBufferAV1),
> > > + buffer + offset,
> > > +
> > > s->tile_group_info[i].tile_size);
> > > + if (err) {
> > > + ff_vaapi_decode_cancel(avctx, pic);
> > > + return err;
> > > + }
> > > + }
> > > +
> > > + return 0;
> > > +}
> > > +
> > > +const AVHWAccel ff_av1_vaapi_hwaccel = {
> > > + .name = "av1_vaapi",
> > > + .type = AVMEDIA_TYPE_VIDEO,
> > > + .id = AV_CODEC_ID_AV1,
> > > + .pix_fmt = AV_PIX_FMT_VAAPI,
> > > + .start_frame = vaapi_av1_start_frame,
> > > + .end_frame = vaapi_av1_end_frame,
> > > + .decode_slice = vaapi_av1_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,
> > > +};
> > > diff --git a/libavcodec/vaapi_decode.c b/libavcodec/vaapi_decode.c
> > > index 5e4f62baad..2a38910cf8 100644
> > > --- a/libavcodec/vaapi_decode.c
> > > +++ b/libavcodec/vaapi_decode.c
> > > @@ -410,6 +410,11 @@ static const struct {
> > > #if VA_CHECK_VERSION(0, 39, 0)
> > > MAP(VP9, VP9_2, VP9Profile2 ),
> > > #endif
> > > +#if VA_CHECK_VERSION(1, 8, 0)
> > > + MAP(AV1, AV1_MAIN, AV1Profile0),
> > > + MAP(AV1, AV1_HIGH, AV1Profile1),
> >
> > What about the 12-bit YUV 4:2:0 referenced above? that requires profile 2.
>
> Profile 2 is not supported by libva yet, can consider to add it when libva
> support.
>
> >
> > > +#endif
> > > +
> > > #undef MAP
> > > };
> > >
> > > @@ -580,6 +585,9 @@ static int
> > vaapi_decode_make_config(AVCodecContext *avctx,
> > > case AV_CODEC_ID_VP8:
> > > frames->initial_pool_size += 3;
> > > break;
> > > + case AV_CODEC_ID_AV1:
> > > + frames->initial_pool_size += 8;
> > > + break;
> >
> > It can be in the same case as VP9, since the reference structure is
> > the same (like H.26[45]).
> >
> > > default:
> > > frames->initial_pool_size += 2;
> > > }
> > > diff --git a/libavcodec/version.h b/libavcodec/version.h index
> > > 5ce4ba55d9..90e48a2ece 100644
> > > --- a/libavcodec/version.h
> > > +++ b/libavcodec/version.h
> > > @@ -29,7 +29,7 @@
> > >
> > > #define LIBAVCODEC_VERSION_MAJOR 58
> > > #define LIBAVCODEC_VERSION_MINOR 106 -#define
> > > LIBAVCODEC_VERSION_MICRO 100
> > > +#define LIBAVCODEC_VERSION_MICRO 101
> > >
> > > #define LIBAVCODEC_VERSION_INT
> > AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
> > >
> > > LIBAVCODEC_VERSION_MINOR, \
> > >
> >
Hi Mark, I have applied your comments and submitted v4 patch, could you help to
review it? Thanks.
> > 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-requ...@ffmpeg.org with subject "unsubscribe".
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
