Message ID | 20220516131558.4163-1-mbonda@nvidia.com |
---|---|
State | New |
Headers | show |
Series | [FFmpeg-devel] AV1 VDPAU hwaccel Decode support | expand |
Context | Check | Description |
---|---|---|
andriy/commit_msg_x86 | warning | The first line of the commit message must start with a context terminated by a colon and a space, for example "lavu/opt: " or "doc: ". |
andriy/make_x86 | success | Make finished |
andriy/make_fate_x86 | success | Make fate finished |
On Mon, 16 May 2022 18:45:58 +0530 ManojGuptaBonda <mbonda-at-nvidia.com@ffmpeg.org> wrote: > Support for VDPAU accelerated AV1 decoding was added with > libvdpau-1.5. Support for the same in ffmpeg is added with this > patch. Profiles related to VDPAU AV1 can be found in latest vdpau.h > present in libvdpau-1.5. > > Add AV1 VDPAU to list of hwaccels and supported formats > Added file vdpau_av1.c and Modified configure to add VDPAU AV1 > support. Mapped AV1 profiles to VDPAU AV1 profiles. Populated the > codec specific params that need to be passed to VDPAU. > --- > Changelog | 1 + > configure | 3 + > libavcodec/Makefile | 1 + > libavcodec/av1dec.c | 13 +- > libavcodec/hwaccels.h | 1 + > libavcodec/vdpau_av1.c | 370 > ++++++++++++++++++++++++++++++++++++ libavcodec/vdpau_internal.h | > 3 + libavcodec/version.h | 2 +- > 8 files changed, 392 insertions(+), 2 deletions(-) > create mode 100644 libavcodec/vdpau_av1.c > <Snip> > diff --git a/libavcodec/vdpau_av1.c b/libavcodec/vdpau_av1.c > new file mode 100644 > index 0000000000..95c1e58cf7 > --- /dev/null > +++ b/libavcodec/vdpau_av1.c > @@ -0,0 +1,370 @@ > +/* > + * AV1 HW decode acceleration through VDPAU > + * > + * Copyright (c) 2022 Manoj Gupta Bonda > + * > + * 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 <vdpau/vdpau.h> > +#include "libavutil/pixdesc.h" > +#include "avcodec.h" > +#include "internal.h" > +#include "av1dec.h" > +#include "hwconfig.h" > +#include "vdpau.h" > +#include "vdpau_internal.h" > + > +static int get_bit_depth_from_seq(const AV1RawSequenceHeader *seq) > +{ > + if (seq->seq_profile == 2 && seq->color_config.high_bitdepth) { > + return seq->color_config.twelve_bit ? 12 : 10; > + } else if (seq->seq_profile <= 2 && > seq->color_config.high_bitdepth) { > + return 10; > + } else { > + return 8; > + } > +} > + > +static int vdpau_av1_start_frame(AVCodecContext *avctx, > + const uint8_t *buffer, uint32_t > size) +{ > + AV1DecContext *s = avctx->priv_data; > + const AV1RawSequenceHeader *seq = s->raw_seq; > + const AV1RawFrameHeader *frame_header = s->raw_frame_header; > + const AV1RawFilmGrainParams *film_grain = > &s->cur_frame.film_grain; + > + struct vdpau_picture_context *pic_ctx = > s->cur_frame.hwaccel_picture_private; > + int i,j; > + > + unsigned char remap_lr_type[4] = { AV1_RESTORE_NONE, > AV1_RESTORE_SWITCHABLE, AV1_RESTORE_WIENER, AV1_RESTORE_SGRPROJ }; + > + > + VdpPictureInfoAV1 *info = &pic_ctx->info.av1; > + const AVPixFmtDescriptor *pixdesc = > av_pix_fmt_desc_get(avctx->sw_pix_fmt); > + if (!pixdesc) { > + return AV_PIX_FMT_NONE; > + } > + > + info->width = avctx->width; > + info->height = avctx->height; > + > + > + info->frame_offset = frame_header->order_hint; > + > + /* Sequence Header */ > + info->profile = seq->seq_profile; > + info->use_128x128_superblock = seq->use_128x128_superblock; > + info->subsampling_x = > seq->color_config.subsampling_x; > + info->subsampling_y = > seq->color_config.subsampling_y; > + info->mono_chrome = seq->color_config.mono_chrome; > + info->bit_depth_minus8 = get_bit_depth_from_seq(seq) - > 8; > + info->enable_filter_intra = seq->enable_filter_intra; > + info->enable_intra_edge_filter = seq->enable_intra_edge_filter; > + info->enable_interintra_compound = > seq->enable_interintra_compound; > + info->enable_masked_compound = seq->enable_masked_compound; > + info->enable_dual_filter = seq->enable_dual_filter; > + info->enable_order_hint = seq->enable_order_hint; > + info->order_hint_bits_minus1 = seq->order_hint_bits_minus_1; > + info->enable_jnt_comp = seq->enable_jnt_comp; > + info->enable_superres = seq->enable_superres; > + info->enable_cdef = seq->enable_cdef; > + info->enable_restoration = seq->enable_restoration; > + info->enable_fgs = > seq->film_grain_params_present; + > + /* Frame Header */ > + info->frame_type = frame_header->frame_type; > + info->show_frame = frame_header->show_frame; > + info->disable_cdf_update = > frame_header->disable_cdf_update; > + info->allow_screen_content_tools = > frame_header->allow_screen_content_tools; > + info->force_integer_mv = > frame_header->force_integer_mv || > + frame_header->frame_type == > AV1_FRAME_INTRA_ONLY || > + frame_header->frame_type == > AV1_FRAME_KEY; > + info->coded_denom = frame_header->coded_denom; > + info->allow_intrabc = frame_header->allow_intrabc; > + info->allow_high_precision_mv = > frame_header->allow_high_precision_mv; > + info->interp_filter = > frame_header->interpolation_filter; > + info->switchable_motion_mode = > frame_header->is_motion_mode_switchable; > + info->use_ref_frame_mvs = > frame_header->use_ref_frame_mvs; > + info->disable_frame_end_update_cdf = > frame_header->disable_frame_end_update_cdf; > + info->delta_q_present = > frame_header->delta_q_present; > + info->delta_q_res = frame_header->delta_q_res; > + info->using_qmatrix = frame_header->using_qmatrix; > + info->coded_lossless = s->cur_frame.coded_lossless; > + info->use_superres = frame_header->use_superres; > + info->tx_mode = frame_header->tx_mode; > + info->reference_mode = > frame_header->reference_select; > + info->allow_warped_motion = > frame_header->allow_warped_motion; > + info->reduced_tx_set = > frame_header->reduced_tx_set; > + info->skip_mode = > frame_header->skip_mode_present; + > + /* Tiling Info */ > + info->num_tile_cols = frame_header->tile_cols; > + info->num_tile_rows = frame_header->tile_rows; > + info->context_update_tile_id = > frame_header->context_update_tile_id; + > + /* CDEF */ > + info->cdef_damping_minus_3 = frame_header->cdef_damping_minus_3; > + info->cdef_bits = frame_header->cdef_bits; > + > + /* SkipModeFrames */ > + info->SkipModeFrame0 = frame_header->skip_mode_present ? > + s->cur_frame.skip_mode_frame_idx[0] : 0; > + info->SkipModeFrame1 = frame_header->skip_mode_present ? > + s->cur_frame.skip_mode_frame_idx[1] : 0; > + > + /* QP Information */ > + info->base_qindex = frame_header->base_q_idx; > + info->qp_y_dc_delta_q = frame_header->delta_q_y_dc; > + info->qp_u_dc_delta_q = frame_header->delta_q_u_dc; > + info->qp_v_dc_delta_q = frame_header->delta_q_v_dc; > + info->qp_u_ac_delta_q = frame_header->delta_q_u_ac; > + info->qp_v_ac_delta_q = frame_header->delta_q_v_ac; > + info->qm_y = frame_header->qm_y; > + info->qm_u = frame_header->qm_u; > + info->qm_v = frame_header->qm_v; > + > + /* Segmentation */ > + info->segmentation_enabled = > frame_header->segmentation_enabled; > + info->segmentation_update_map = > frame_header->segmentation_update_map; > + info->segmentation_update_data = > frame_header->segmentation_update_data; > + info->segmentation_temporal_update = > frame_header->segmentation_temporal_update; + > + /* Loopfilter */ > + info->loop_filter_level[0] = > frame_header->loop_filter_level[0]; > + info->loop_filter_level[1] = > frame_header->loop_filter_level[1]; > + info->loop_filter_level_u = > frame_header->loop_filter_level[2]; > + info->loop_filter_level_v = > frame_header->loop_filter_level[3]; > + info->loop_filter_sharpness = > frame_header->loop_filter_sharpness; > + info->loop_filter_delta_enabled = > frame_header->loop_filter_delta_enabled; > + info->loop_filter_delta_update = > frame_header->loop_filter_delta_update; > + info->loop_filter_mode_deltas[0] = > frame_header->loop_filter_mode_deltas[0]; > + info->loop_filter_mode_deltas[1] = > frame_header->loop_filter_mode_deltas[1]; > + info->delta_lf_present = > frame_header->delta_lf_present; > + info->delta_lf_res = frame_header->delta_lf_res; > + info->delta_lf_multi = frame_header->delta_lf_multi; > + > + /* Restoration */ > + info->lr_type[0] = remap_lr_type[frame_header->lr_type[0]]; > + info->lr_type[1] = remap_lr_type[frame_header->lr_type[1]]; > + info->lr_type[2] = remap_lr_type[frame_header->lr_type[2]]; > + info->lr_unit_size[0] = 1 + frame_header->lr_unit_shift; > + info->lr_unit_size[1] = 1 + frame_header->lr_unit_shift - > frame_header->lr_uv_shift; > + info->lr_unit_size[2] = 1 + frame_header->lr_unit_shift - > frame_header->lr_uv_shift; + > + /* Reference Frames */ > + info->temporal_layer_id = s->cur_frame.temporal_id; > + info->spatial_layer_id = s->cur_frame.spatial_id; > + > + /* Film Grain Params */ > + info->apply_grain = film_grain->apply_grain; > + info->overlap_flag = film_grain->overlap_flag; > + info->scaling_shift_minus8 = > film_grain->grain_scaling_minus_8; > + info->chroma_scaling_from_luma = > film_grain->chroma_scaling_from_luma; > + info->ar_coeff_lag = film_grain->ar_coeff_lag; > + info->ar_coeff_shift_minus6 = > film_grain->ar_coeff_shift_minus_6; > + info->grain_scale_shift = film_grain->grain_scale_shift; > + info->clip_to_restricted_range = > film_grain->clip_to_restricted_range; > + info->num_y_points = film_grain->num_y_points; > + info->num_cb_points = film_grain->num_cb_points; > + info->num_cr_points = film_grain->num_cr_points; > + info->random_seed = film_grain->grain_seed; > + info->cb_mult = film_grain->cb_mult; > + info->cb_luma_mult = film_grain->cb_luma_mult; > + info->cb_offset = film_grain->cb_offset; > + info->cr_mult = film_grain->cr_mult; > + info->cr_luma_mult = film_grain->cr_luma_mult; > + info->cr_offset = film_grain->cr_offset; > + > + /* Tiling Info */ > + for (i = 0; i < frame_header->tile_cols; ++i) { > + info->tile_widths[i] = frame_header->width_in_sbs_minus_1[i] > + 1; > + } > + for (i = 0; i < frame_header->tile_rows; ++i) { > + info->tile_heights[i] = > frame_header->height_in_sbs_minus_1[i] + 1; > + } > + > + /* CDEF */ > + for (i = 0; i < (1 << frame_header->cdef_bits); ++i) { > + info->cdef_y_strength[i] = > (frame_header->cdef_y_pri_strength[i] & 0x0F) | > (frame_header->cdef_y_sec_strength[i] << 4); > + info->cdef_uv_strength[i] = > (frame_header->cdef_uv_pri_strength[i] & 0x0F) | > (frame_header->cdef_uv_sec_strength[i] << 4); > + } > + > + > + /* Segmentation */ > + for (i = 0; i < AV1_MAX_SEGMENTS; ++i) { > + info->segmentation_feature_mask[i] = 0; > + for (j = 0; j < AV1_SEG_LVL_MAX; ++j) { > + info->segmentation_feature_mask[i] |= > frame_header->feature_enabled[i][j] << j; > + info->segmentation_feature_data[i][j] = > frame_header->feature_value[i][j]; > + } > + } > + > + for (i = 0; i < AV1_NUM_REF_FRAMES; ++i) { > + /* Loopfilter */ > + info->loop_filter_ref_deltas[i] = > frame_header->loop_filter_ref_deltas[i]; + > + /* Reference Frames */ > + info->ref_frame_map[i] = > ff_vdpau_get_surface_id(s->ref[i].tf.f) ? > ff_vdpau_get_surface_id(s->ref[i].tf.f) : VDP_INVALID_HANDLE; These usages of tf.f do not compile. I replaced them with .f and it does compile, but then when I tested in mpv (doesn't require any code changes, just run against the patched ffmpeg), there are visual glitches and something is clearly wrong. Doesn't matter whether I use the GL interop or copy-back. This is with 510.68.02 drivers. --phil
Hi Philip, I am currently relying on using ffmpeg with nvcuvid to cross verify my vdpau patch I see the clips with me are matching in output for both VDPAU and NVCUVID/NVDEC. Attached are the script I used to verify and its output. Was unable to use AV1 VDPAU with MPV hence not able to repro the issue you are seeing. looking into it once resolved will check and resend the patch. Please bear with me. Thanks, ManojGupta. > -----Original Message----- > From: Philip Langdale <philipl@overt.org> > Sent: Tuesday, May 17, 2022 10:17 AM > To: FFmpeg development discussions and patches <ffmpeg- > devel@ffmpeg.org> > Cc: ManojGuptaBonda <mbonda-at-nvidia.com@ffmpeg.org>; > vdpau@lists.freedesktop.org; Aaron Plattner <aplattner@nvidia.com>; > Manoj Bonda <mbonda@nvidia.com>; Andy Ritger <ARitger@nvidia.com> > Subject: Re: [FFmpeg-devel] [PATCH] AV1 VDPAU hwaccel Decode support > > External email: Use caution opening links or attachments > > > On Mon, 16 May 2022 18:45:58 +0530 > ManojGuptaBonda <mbonda-at-nvidia.com@ffmpeg.org> wrote: > > > Support for VDPAU accelerated AV1 decoding was added with > > libvdpau-1.5. Support for the same in ffmpeg is added with this patch. > > Profiles related to VDPAU AV1 can be found in latest vdpau.h present > > in libvdpau-1.5. > > > > Add AV1 VDPAU to list of hwaccels and supported formats Added file > > vdpau_av1.c and Modified configure to add VDPAU AV1 support. Mapped > > AV1 profiles to VDPAU AV1 profiles. Populated the codec specific > > params that need to be passed to VDPAU. > > --- > > Changelog | 1 + > > configure | 3 + > > libavcodec/Makefile | 1 + > > libavcodec/av1dec.c | 13 +- > > libavcodec/hwaccels.h | 1 + > > libavcodec/vdpau_av1.c | 370 > > ++++++++++++++++++++++++++++++++++++ > libavcodec/vdpau_internal.h | > > 3 + libavcodec/version.h | 2 +- > > 8 files changed, 392 insertions(+), 2 deletions(-) create mode > > 100644 libavcodec/vdpau_av1.c > > > > <Snip> > > > diff --git a/libavcodec/vdpau_av1.c b/libavcodec/vdpau_av1.c new file > > mode 100644 index 0000000000..95c1e58cf7 > > --- /dev/null > > +++ b/libavcodec/vdpau_av1.c > > @@ -0,0 +1,370 @@ > > +/* > > + * AV1 HW decode acceleration through VDPAU > > + * > > + * Copyright (c) 2022 Manoj Gupta Bonda > > + * > > + * 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 <vdpau/vdpau.h> > > +#include "libavutil/pixdesc.h" > > +#include "avcodec.h" > > +#include "internal.h" > > +#include "av1dec.h" > > +#include "hwconfig.h" > > +#include "vdpau.h" > > +#include "vdpau_internal.h" > > + > > +static int get_bit_depth_from_seq(const AV1RawSequenceHeader *seq) > { > > + if (seq->seq_profile == 2 && seq->color_config.high_bitdepth) { > > + return seq->color_config.twelve_bit ? 12 : 10; > > + } else if (seq->seq_profile <= 2 && > > seq->color_config.high_bitdepth) { > > + return 10; > > + } else { > > + return 8; > > + } > > +} > > + > > +static int vdpau_av1_start_frame(AVCodecContext *avctx, > > + const uint8_t *buffer, uint32_t > > size) +{ > > + AV1DecContext *s = avctx->priv_data; > > + const AV1RawSequenceHeader *seq = s->raw_seq; > > + const AV1RawFrameHeader *frame_header = s->raw_frame_header; > > + const AV1RawFilmGrainParams *film_grain = > > &s->cur_frame.film_grain; + > > + struct vdpau_picture_context *pic_ctx = > > s->cur_frame.hwaccel_picture_private; > > + int i,j; > > + > > + unsigned char remap_lr_type[4] = { AV1_RESTORE_NONE, > > AV1_RESTORE_SWITCHABLE, AV1_RESTORE_WIENER, > AV1_RESTORE_SGRPROJ }; + > > + > > + VdpPictureInfoAV1 *info = &pic_ctx->info.av1; > > + const AVPixFmtDescriptor *pixdesc = > > av_pix_fmt_desc_get(avctx->sw_pix_fmt); > > + if (!pixdesc) { > > + return AV_PIX_FMT_NONE; > > + } > > + > > + info->width = avctx->width; > > + info->height = avctx->height; > > + > > + > > + info->frame_offset = frame_header->order_hint; > > + > > + /* Sequence Header */ > > + info->profile = seq->seq_profile; > > + info->use_128x128_superblock = seq->use_128x128_superblock; > > + info->subsampling_x = > > seq->color_config.subsampling_x; > > + info->subsampling_y = > > seq->color_config.subsampling_y; > > + info->mono_chrome = seq->color_config.mono_chrome; > > + info->bit_depth_minus8 = get_bit_depth_from_seq(seq) - > > 8; > > + info->enable_filter_intra = seq->enable_filter_intra; > > + info->enable_intra_edge_filter = seq->enable_intra_edge_filter; > > + info->enable_interintra_compound = > > seq->enable_interintra_compound; > > + info->enable_masked_compound = seq- > >enable_masked_compound; > > + info->enable_dual_filter = seq->enable_dual_filter; > > + info->enable_order_hint = seq->enable_order_hint; > > + info->order_hint_bits_minus1 = seq->order_hint_bits_minus_1; > > + info->enable_jnt_comp = seq->enable_jnt_comp; > > + info->enable_superres = seq->enable_superres; > > + info->enable_cdef = seq->enable_cdef; > > + info->enable_restoration = seq->enable_restoration; > > + info->enable_fgs = > > seq->film_grain_params_present; + > > + /* Frame Header */ > > + info->frame_type = frame_header->frame_type; > > + info->show_frame = frame_header->show_frame; > > + info->disable_cdf_update = > > frame_header->disable_cdf_update; > > + info->allow_screen_content_tools = > > frame_header->allow_screen_content_tools; > > + info->force_integer_mv = > > frame_header->force_integer_mv || > > + frame_header->frame_type == > > AV1_FRAME_INTRA_ONLY || > > + frame_header->frame_type == > > AV1_FRAME_KEY; > > + info->coded_denom = frame_header->coded_denom; > > + info->allow_intrabc = frame_header->allow_intrabc; > > + info->allow_high_precision_mv = > > frame_header->allow_high_precision_mv; > > + info->interp_filter = > > frame_header->interpolation_filter; > > + info->switchable_motion_mode = > > frame_header->is_motion_mode_switchable; > > + info->use_ref_frame_mvs = > > frame_header->use_ref_frame_mvs; > > + info->disable_frame_end_update_cdf = > > frame_header->disable_frame_end_update_cdf; > > + info->delta_q_present = > > frame_header->delta_q_present; > > + info->delta_q_res = frame_header->delta_q_res; > > + info->using_qmatrix = frame_header->using_qmatrix; > > + info->coded_lossless = s->cur_frame.coded_lossless; > > + info->use_superres = frame_header->use_superres; > > + info->tx_mode = frame_header->tx_mode; > > + info->reference_mode = > > frame_header->reference_select; > > + info->allow_warped_motion = > > frame_header->allow_warped_motion; > > + info->reduced_tx_set = > > frame_header->reduced_tx_set; > > + info->skip_mode = > > frame_header->skip_mode_present; + > > + /* Tiling Info */ > > + info->num_tile_cols = frame_header->tile_cols; > > + info->num_tile_rows = frame_header->tile_rows; > > + info->context_update_tile_id = > > frame_header->context_update_tile_id; + > > + /* CDEF */ > > + info->cdef_damping_minus_3 = frame_header- > >cdef_damping_minus_3; > > + info->cdef_bits = frame_header->cdef_bits; > > + > > + /* SkipModeFrames */ > > + info->SkipModeFrame0 = frame_header->skip_mode_present ? > > + s->cur_frame.skip_mode_frame_idx[0] : 0; > > + info->SkipModeFrame1 = frame_header->skip_mode_present ? > > + s->cur_frame.skip_mode_frame_idx[1] : 0; > > + > > + /* QP Information */ > > + info->base_qindex = frame_header->base_q_idx; > > + info->qp_y_dc_delta_q = frame_header->delta_q_y_dc; > > + info->qp_u_dc_delta_q = frame_header->delta_q_u_dc; > > + info->qp_v_dc_delta_q = frame_header->delta_q_v_dc; > > + info->qp_u_ac_delta_q = frame_header->delta_q_u_ac; > > + info->qp_v_ac_delta_q = frame_header->delta_q_v_ac; > > + info->qm_y = frame_header->qm_y; > > + info->qm_u = frame_header->qm_u; > > + info->qm_v = frame_header->qm_v; > > + > > + /* Segmentation */ > > + info->segmentation_enabled = > > frame_header->segmentation_enabled; > > + info->segmentation_update_map = > > frame_header->segmentation_update_map; > > + info->segmentation_update_data = > > frame_header->segmentation_update_data; > > + info->segmentation_temporal_update = > > frame_header->segmentation_temporal_update; + > > + /* Loopfilter */ > > + info->loop_filter_level[0] = > > frame_header->loop_filter_level[0]; > > + info->loop_filter_level[1] = > > frame_header->loop_filter_level[1]; > > + info->loop_filter_level_u = > > frame_header->loop_filter_level[2]; > > + info->loop_filter_level_v = > > frame_header->loop_filter_level[3]; > > + info->loop_filter_sharpness = > > frame_header->loop_filter_sharpness; > > + info->loop_filter_delta_enabled = > > frame_header->loop_filter_delta_enabled; > > + info->loop_filter_delta_update = > > frame_header->loop_filter_delta_update; > > + info->loop_filter_mode_deltas[0] = > > frame_header->loop_filter_mode_deltas[0]; > > + info->loop_filter_mode_deltas[1] = > > frame_header->loop_filter_mode_deltas[1]; > > + info->delta_lf_present = > > frame_header->delta_lf_present; > > + info->delta_lf_res = frame_header->delta_lf_res; > > + info->delta_lf_multi = frame_header->delta_lf_multi; > > + > > + /* Restoration */ > > + info->lr_type[0] = remap_lr_type[frame_header->lr_type[0]]; > > + info->lr_type[1] = remap_lr_type[frame_header->lr_type[1]]; > > + info->lr_type[2] = remap_lr_type[frame_header->lr_type[2]]; > > + info->lr_unit_size[0] = 1 + frame_header->lr_unit_shift; > > + info->lr_unit_size[1] = 1 + frame_header->lr_unit_shift - > > frame_header->lr_uv_shift; > > + info->lr_unit_size[2] = 1 + frame_header->lr_unit_shift - > > frame_header->lr_uv_shift; + > > + /* Reference Frames */ > > + info->temporal_layer_id = s->cur_frame.temporal_id; > > + info->spatial_layer_id = s->cur_frame.spatial_id; > > + > > + /* Film Grain Params */ > > + info->apply_grain = film_grain->apply_grain; > > + info->overlap_flag = film_grain->overlap_flag; > > + info->scaling_shift_minus8 = > > film_grain->grain_scaling_minus_8; > > + info->chroma_scaling_from_luma = > > film_grain->chroma_scaling_from_luma; > > + info->ar_coeff_lag = film_grain->ar_coeff_lag; > > + info->ar_coeff_shift_minus6 = > > film_grain->ar_coeff_shift_minus_6; > > + info->grain_scale_shift = film_grain->grain_scale_shift; > > + info->clip_to_restricted_range = > > film_grain->clip_to_restricted_range; > > + info->num_y_points = film_grain->num_y_points; > > + info->num_cb_points = film_grain->num_cb_points; > > + info->num_cr_points = film_grain->num_cr_points; > > + info->random_seed = film_grain->grain_seed; > > + info->cb_mult = film_grain->cb_mult; > > + info->cb_luma_mult = film_grain->cb_luma_mult; > > + info->cb_offset = film_grain->cb_offset; > > + info->cr_mult = film_grain->cr_mult; > > + info->cr_luma_mult = film_grain->cr_luma_mult; > > + info->cr_offset = film_grain->cr_offset; > > + > > + /* Tiling Info */ > > + for (i = 0; i < frame_header->tile_cols; ++i) { > > + info->tile_widths[i] = frame_header->width_in_sbs_minus_1[i] > > + 1; > > + } > > + for (i = 0; i < frame_header->tile_rows; ++i) { > > + info->tile_heights[i] = > > frame_header->height_in_sbs_minus_1[i] + 1; > > + } > > + > > + /* CDEF */ > > + for (i = 0; i < (1 << frame_header->cdef_bits); ++i) { > > + info->cdef_y_strength[i] = > > (frame_header->cdef_y_pri_strength[i] & 0x0F) | > > (frame_header->cdef_y_sec_strength[i] << 4); > > + info->cdef_uv_strength[i] = > > (frame_header->cdef_uv_pri_strength[i] & 0x0F) | > > (frame_header->cdef_uv_sec_strength[i] << 4); > > + } > > + > > + > > + /* Segmentation */ > > + for (i = 0; i < AV1_MAX_SEGMENTS; ++i) { > > + info->segmentation_feature_mask[i] = 0; > > + for (j = 0; j < AV1_SEG_LVL_MAX; ++j) { > > + info->segmentation_feature_mask[i] |= > > frame_header->feature_enabled[i][j] << j; > > + info->segmentation_feature_data[i][j] = > > frame_header->feature_value[i][j]; > > + } > > + } > > + > > + for (i = 0; i < AV1_NUM_REF_FRAMES; ++i) { > > + /* Loopfilter */ > > + info->loop_filter_ref_deltas[i] = > > frame_header->loop_filter_ref_deltas[i]; + > > + /* Reference Frames */ > > + info->ref_frame_map[i] = > > ff_vdpau_get_surface_id(s->ref[i].tf.f) ? > > ff_vdpau_get_surface_id(s->ref[i].tf.f) : VDP_INVALID_HANDLE; > > These usages of tf.f do not compile. I replaced them with .f and it does > compile, but then when I tested in mpv (doesn't require any code changes, > just run against the patched ffmpeg), there are visual glitches and something > is clearly wrong. Doesn't matter whether I use the GL interop or copy-back. > This is with 510.68.02 drivers. > > --phil Test av1_stream_00000 Test av1_stream_00001 Test av1_stream_00002 Test av1_stream_00003 Test av1_stream_00004 Test av1_stream_00005 Test av1_stream_00007 Test av1_stream_00008 Test av1_stream_00009 Test av1_stream_00010 Test av1_stream_00011 Test av1_stream_00012 Test av1_stream_00014 Test av1_stream_00015 Test av1_stream_00016 Test av1_stream_00022 Test av1_stream_00028 target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size Test av1_stream_00023 Test av1_stream_00030 Test av1_stream_00324 Test av1_stream_00325 Test av1_stream_00326 Test av1_stream_00327 Test av1_stream_00328 Test av1_stream_00329 Test av1_stream_00330 Test av1_stream_00331 target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size target surface doesn't match Frame size Test av1_stream_00410 Test av1_stream_00411 Test av1_stream_00447 Test av1_stream_00448 Test av1_stream_00449 Test av1_stream_00450.ivf Test av1_stream_00451 Test Chimera-AV1-10bit-1280x720-2380kbps Test Chimera-AV1-8bit-1280x720-3363kbps.ivf Test av1_test_mul_tile_groups_tiles.av1 Test av1_test_mul_tiles.av1
Hi Philip, The issue seems to be from MPV side rather than from ffmpeg patch. I was able to decode the clip Halo_AV1 properly bitmatching with nvcuvid output with ffmpeg + VDPAU. Please find the cmds I used to verify the same. ./ffmpeg -hwaccel vdpau -i /home/bondamanoj/VideoClips/AV1Clips/Halo_AV1.mp4 -y decoded_vdpau.yuv ./ffmpeg -hwaccel nvdec -i /home/bondamanoj/VideoClips/AV1Clips/Halo_AV1.mp4 decoded_nvcuvid.yuv cmp decoded_vdpau.yuv ~/Downloads/ffmpeg-git2/FFmpeg/decoded_nvcuvid.yuv The same clip is dumped using mpv via cmd ./mpv --hwdec=vdpau-copy --o=Halo_dump_av1.yuv --of=rawvideo /home/bondamanoj/VideoClips/AV1Clips/Halo_AV1.mp4 Is differing and is having corruption while playing. Thanks, ManojGupta. > -----Original Message----- > From: Philip Langdale <philipl@overt.org> > Sent: Thursday, May 19, 2022 10:25 PM > To: Manoj Bonda <mbonda@nvidia.com> > Cc: Andy Ritger <ARitger@nvidia.com>; Aaron Plattner > <aplattner@nvidia.com>; Mandar Godse <mgodse@nvidia.com>; Ashish Roy > <AROY@nvidia.com>; Hardy Doelfel <hdoelfel@nvidia.com> > Subject: Re: [FFmpeg-devel] [PATCH] AV1 VDPAU hwaccel Decode support > > External email: Use caution opening links or attachments > > > On Thu, 19 May 2022 09:01:26 +0000 > Manoj Bonda <mbonda@nvidia.com> wrote: > > > Hi Philip, > > > > As mentioned earlier I am getting output matching with nvcuvid for > > VDPAU While using ffmpeg. > > With MPV I am not seeing issue with below cmd ./mpv --vo=vdpau > > --hwdec=vdpau > > /home/bondamanoj/VideoClips/AV1Clips/Chimera-AV1-8bit-1280x720- > 3363kbp > > s.ivf > > > > I am getting mpv failures with --vo=opengl or --vo=gpu and not > > decoding the clip. Can you please share us the cmds and the clip used > > by you where you saw corruption. So I can debug and fix the issue. > > Ok, it seems weird. > > I've been using youtube clips, so here's one from their original AV1 playlist > (The Halo Infinite announce trailer): > > https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww. > youtube.com%2Fwatch%3Fv%3DFmdb- > KmlzD8&data=05%7C01%7Cmbonda%40nvidia.com%7Cf69926f451734 > 86cf2f208da39b8430d%7C43083d15727340c1b7db39efd9ccc17a%7C0%7C0 > %7C637885760832069373%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLj > AwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C% > 7C%7C&sdata=Kw2mXnUdaupPhwBq13MCCiciZo6alDqTpM6sh88TMtg% > 3D&reserved=0 > > You then need yt-dlp installed. > > If you tell mpv to play the URL directly, it will use yt-dlp to stream it, and it > appears to play just fine. > > However, if you use yt-dlp to download the stream to either mp4 or mkv, and > then play that back, the playback is glitched. > > With nvdec, it plays fine either way. > > Minimum Command line: > > mpv --no-config --hwdec=vdpau --gpu-context=x11 <file or url> > > And switch vdpau for nvdec to compare the two. > > Hope that helps, > > --phil
On Tue, 21 Jun 2022 02:45:21 +0000 Manoj Bonda <mbonda@nvidia.com> wrote: > Hi Philip, > > The issue seems to be from MPV side rather than from ffmpeg patch. > I was able to decode the clip Halo_AV1 properly bitmatching with > nvcuvid output with ffmpeg + VDPAU. > > Please find the cmds I used to verify the same. > > ./ffmpeg -hwaccel vdpau -i > /home/bondamanoj/VideoClips/AV1Clips/Halo_AV1.mp4 -y > decoded_vdpau.yuv ./ffmpeg -hwaccel nvdec -i > /home/bondamanoj/VideoClips/AV1Clips/Halo_AV1.mp4 > decoded_nvcuvid.yuv cmp decoded_vdpau.yuv > ~/Downloads/ffmpeg-git2/FFmpeg/decoded_nvcuvid.yuv > > The same clip is dumped using mpv via cmd > ./mpv --hwdec=vdpau-copy --o=Halo_dump_av1.yuv --of=rawvideo > /home/bondamanoj/VideoClips/AV1Clips/Halo_AV1.mp4 > > Is differing and is having corruption while playing. > > > Thanks, > ManojGupta. Hi Manoj, Ok. In that case, I guess it's reasonable to merge. I'm puzzled as to the behaviour as other codecs work fine and other hwaccels used by the same av1 decoder also work fine, so I don't see where there is room for things to go wrong specifically in mpv specifically for this hwaccel. Anyway, I'll merge it later this week. Thanks, --phil
diff --git a/Changelog b/Changelog index 4d42a0f681..a713edf566 100644 --- a/Changelog +++ b/Changelog @@ -16,6 +16,7 @@ version 5.1: - blurdetect filter - multiply video filter - PGS subtitle frame merge bitstream filter +- VDPAU AV1 hwaccel version 5.0: diff --git a/configure b/configure index 0faf2455b7..702fb4d741 100755 --- a/configure +++ b/configure @@ -3025,6 +3025,8 @@ av1_nvdec_hwaccel_deps="nvdec CUVIDAV1PICPARAMS" av1_nvdec_hwaccel_select="av1_decoder" av1_vaapi_hwaccel_deps="vaapi VADecPictureParameterBufferAV1_bit_depth_idx" av1_vaapi_hwaccel_select="av1_decoder" +av1_vdpau_hwaccel_deps="vdpau VdpPictureInfoAV1" +av1_vdpau_hwaccel_select="av1_decoder" h263_vaapi_hwaccel_deps="vaapi" h263_vaapi_hwaccel_select="h263_decoder" h263_videotoolbox_hwaccel_deps="videotoolbox" @@ -6391,6 +6393,7 @@ check_func_headers mfapi.h MFCreateAlignedMemoryBuffer -lmfplat check_type "vdpau/vdpau.h" "VdpPictureInfoHEVC" check_type "vdpau/vdpau.h" "VdpPictureInfoVP9" +check_type "vdpau/vdpau.h" "VdpPictureInfoAV1" if [ -z "$nvccflags" ]; then nvccflags=$nvccflags_default diff --git a/libavcodec/Makefile b/libavcodec/Makefile index 38425d2f22..813989cee9 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -954,6 +954,7 @@ OBJS-$(CONFIG_AV1_D3D11VA_HWACCEL) += dxva2_av1.o OBJS-$(CONFIG_AV1_DXVA2_HWACCEL) += dxva2_av1.o OBJS-$(CONFIG_AV1_NVDEC_HWACCEL) += nvdec_av1.o OBJS-$(CONFIG_AV1_VAAPI_HWACCEL) += vaapi_av1.o +OBJS-$(CONFIG_AV1_VDPAU_HWACCEL) += vdpau_av1.o OBJS-$(CONFIG_H263_VAAPI_HWACCEL) += vaapi_mpeg4.o OBJS-$(CONFIG_H263_VIDEOTOOLBOX_HWACCEL) += videotoolbox.o OBJS-$(CONFIG_H264_D3D11VA_HWACCEL) += dxva2_h264.o diff --git a/libavcodec/av1dec.c b/libavcodec/av1dec.c index 1c09b1d6d6..dd83373dd6 100644 --- a/libavcodec/av1dec.c +++ b/libavcodec/av1dec.c @@ -441,7 +441,8 @@ static int get_pixel_format(AVCodecContext *avctx) #define HWACCEL_MAX (CONFIG_AV1_DXVA2_HWACCEL + \ CONFIG_AV1_D3D11VA_HWACCEL * 2 + \ CONFIG_AV1_NVDEC_HWACCEL + \ - CONFIG_AV1_VAAPI_HWACCEL) + CONFIG_AV1_VAAPI_HWACCEL + \ + CONFIG_AV1_VDPAU_HWACCEL) enum AVPixelFormat pix_fmts[HWACCEL_MAX + 2], *fmtp = pix_fmts; if (seq->seq_profile == 2 && seq->color_config.high_bitdepth) @@ -518,6 +519,9 @@ static int get_pixel_format(AVCodecContext *avctx) #endif #if CONFIG_AV1_VAAPI_HWACCEL *fmtp++ = AV_PIX_FMT_VAAPI; +#endif +#if CONFIG_AV1_VDPAU_HWACCEL + *fmtp++ = AV_PIX_FMT_VDPAU; #endif break; case AV_PIX_FMT_YUV420P10: @@ -533,6 +537,9 @@ static int get_pixel_format(AVCodecContext *avctx) #endif #if CONFIG_AV1_VAAPI_HWACCEL *fmtp++ = AV_PIX_FMT_VAAPI; +#endif +#if CONFIG_AV1_VDPAU_HWACCEL + *fmtp++ = AV_PIX_FMT_VDPAU; #endif break; case AV_PIX_FMT_GRAY8: @@ -1271,6 +1278,10 @@ const FFCodec ff_av1_decoder = { #if CONFIG_AV1_VAAPI_HWACCEL HWACCEL_VAAPI(av1), #endif +#if CONFIG_AV1_VDPAU_HWACCEL + HWACCEL_VDPAU(av1), +#endif + NULL }, }; diff --git a/libavcodec/hwaccels.h b/libavcodec/hwaccels.h index ff4daca0e2..aca55831f3 100644 --- a/libavcodec/hwaccels.h +++ b/libavcodec/hwaccels.h @@ -26,6 +26,7 @@ extern const AVHWAccel ff_av1_d3d11va2_hwaccel; extern const AVHWAccel ff_av1_dxva2_hwaccel; extern const AVHWAccel ff_av1_nvdec_hwaccel; extern const AVHWAccel ff_av1_vaapi_hwaccel; +extern const AVHWAccel ff_av1_vdpau_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/vdpau_av1.c b/libavcodec/vdpau_av1.c new file mode 100644 index 0000000000..95c1e58cf7 --- /dev/null +++ b/libavcodec/vdpau_av1.c @@ -0,0 +1,370 @@ +/* + * AV1 HW decode acceleration through VDPAU + * + * Copyright (c) 2022 Manoj Gupta Bonda + * + * 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 <vdpau/vdpau.h> +#include "libavutil/pixdesc.h" +#include "avcodec.h" +#include "internal.h" +#include "av1dec.h" +#include "hwconfig.h" +#include "vdpau.h" +#include "vdpau_internal.h" + +static int get_bit_depth_from_seq(const AV1RawSequenceHeader *seq) +{ + if (seq->seq_profile == 2 && seq->color_config.high_bitdepth) { + return seq->color_config.twelve_bit ? 12 : 10; + } else if (seq->seq_profile <= 2 && seq->color_config.high_bitdepth) { + return 10; + } else { + return 8; + } +} + +static int vdpau_av1_start_frame(AVCodecContext *avctx, + const uint8_t *buffer, uint32_t size) +{ + AV1DecContext *s = avctx->priv_data; + const AV1RawSequenceHeader *seq = s->raw_seq; + const AV1RawFrameHeader *frame_header = s->raw_frame_header; + const AV1RawFilmGrainParams *film_grain = &s->cur_frame.film_grain; + + struct vdpau_picture_context *pic_ctx = s->cur_frame.hwaccel_picture_private; + int i,j; + + unsigned char remap_lr_type[4] = { AV1_RESTORE_NONE, AV1_RESTORE_SWITCHABLE, AV1_RESTORE_WIENER, AV1_RESTORE_SGRPROJ }; + + + VdpPictureInfoAV1 *info = &pic_ctx->info.av1; + const AVPixFmtDescriptor *pixdesc = av_pix_fmt_desc_get(avctx->sw_pix_fmt); + if (!pixdesc) { + return AV_PIX_FMT_NONE; + } + + info->width = avctx->width; + info->height = avctx->height; + + + info->frame_offset = frame_header->order_hint; + + /* Sequence Header */ + info->profile = seq->seq_profile; + info->use_128x128_superblock = seq->use_128x128_superblock; + info->subsampling_x = seq->color_config.subsampling_x; + info->subsampling_y = seq->color_config.subsampling_y; + info->mono_chrome = seq->color_config.mono_chrome; + info->bit_depth_minus8 = get_bit_depth_from_seq(seq) - 8; + info->enable_filter_intra = seq->enable_filter_intra; + info->enable_intra_edge_filter = seq->enable_intra_edge_filter; + info->enable_interintra_compound = seq->enable_interintra_compound; + info->enable_masked_compound = seq->enable_masked_compound; + info->enable_dual_filter = seq->enable_dual_filter; + info->enable_order_hint = seq->enable_order_hint; + info->order_hint_bits_minus1 = seq->order_hint_bits_minus_1; + info->enable_jnt_comp = seq->enable_jnt_comp; + info->enable_superres = seq->enable_superres; + info->enable_cdef = seq->enable_cdef; + info->enable_restoration = seq->enable_restoration; + info->enable_fgs = seq->film_grain_params_present; + + /* Frame Header */ + info->frame_type = frame_header->frame_type; + info->show_frame = frame_header->show_frame; + info->disable_cdf_update = frame_header->disable_cdf_update; + info->allow_screen_content_tools = frame_header->allow_screen_content_tools; + info->force_integer_mv = frame_header->force_integer_mv || + frame_header->frame_type == AV1_FRAME_INTRA_ONLY || + frame_header->frame_type == AV1_FRAME_KEY; + info->coded_denom = frame_header->coded_denom; + info->allow_intrabc = frame_header->allow_intrabc; + info->allow_high_precision_mv = frame_header->allow_high_precision_mv; + info->interp_filter = frame_header->interpolation_filter; + info->switchable_motion_mode = frame_header->is_motion_mode_switchable; + info->use_ref_frame_mvs = frame_header->use_ref_frame_mvs; + info->disable_frame_end_update_cdf = frame_header->disable_frame_end_update_cdf; + info->delta_q_present = frame_header->delta_q_present; + info->delta_q_res = frame_header->delta_q_res; + info->using_qmatrix = frame_header->using_qmatrix; + info->coded_lossless = s->cur_frame.coded_lossless; + info->use_superres = frame_header->use_superres; + info->tx_mode = frame_header->tx_mode; + info->reference_mode = frame_header->reference_select; + info->allow_warped_motion = frame_header->allow_warped_motion; + info->reduced_tx_set = frame_header->reduced_tx_set; + info->skip_mode = frame_header->skip_mode_present; + + /* Tiling Info */ + info->num_tile_cols = frame_header->tile_cols; + info->num_tile_rows = frame_header->tile_rows; + info->context_update_tile_id = frame_header->context_update_tile_id; + + /* CDEF */ + info->cdef_damping_minus_3 = frame_header->cdef_damping_minus_3; + info->cdef_bits = frame_header->cdef_bits; + + /* SkipModeFrames */ + info->SkipModeFrame0 = frame_header->skip_mode_present ? + s->cur_frame.skip_mode_frame_idx[0] : 0; + info->SkipModeFrame1 = frame_header->skip_mode_present ? + s->cur_frame.skip_mode_frame_idx[1] : 0; + + /* QP Information */ + info->base_qindex = frame_header->base_q_idx; + info->qp_y_dc_delta_q = frame_header->delta_q_y_dc; + info->qp_u_dc_delta_q = frame_header->delta_q_u_dc; + info->qp_v_dc_delta_q = frame_header->delta_q_v_dc; + info->qp_u_ac_delta_q = frame_header->delta_q_u_ac; + info->qp_v_ac_delta_q = frame_header->delta_q_v_ac; + info->qm_y = frame_header->qm_y; + info->qm_u = frame_header->qm_u; + info->qm_v = frame_header->qm_v; + + /* Segmentation */ + info->segmentation_enabled = frame_header->segmentation_enabled; + info->segmentation_update_map = frame_header->segmentation_update_map; + info->segmentation_update_data = frame_header->segmentation_update_data; + info->segmentation_temporal_update = frame_header->segmentation_temporal_update; + + /* Loopfilter */ + info->loop_filter_level[0] = frame_header->loop_filter_level[0]; + info->loop_filter_level[1] = frame_header->loop_filter_level[1]; + info->loop_filter_level_u = frame_header->loop_filter_level[2]; + info->loop_filter_level_v = frame_header->loop_filter_level[3]; + info->loop_filter_sharpness = frame_header->loop_filter_sharpness; + info->loop_filter_delta_enabled = frame_header->loop_filter_delta_enabled; + info->loop_filter_delta_update = frame_header->loop_filter_delta_update; + info->loop_filter_mode_deltas[0] = frame_header->loop_filter_mode_deltas[0]; + info->loop_filter_mode_deltas[1] = frame_header->loop_filter_mode_deltas[1]; + info->delta_lf_present = frame_header->delta_lf_present; + info->delta_lf_res = frame_header->delta_lf_res; + info->delta_lf_multi = frame_header->delta_lf_multi; + + /* Restoration */ + info->lr_type[0] = remap_lr_type[frame_header->lr_type[0]]; + info->lr_type[1] = remap_lr_type[frame_header->lr_type[1]]; + info->lr_type[2] = remap_lr_type[frame_header->lr_type[2]]; + info->lr_unit_size[0] = 1 + frame_header->lr_unit_shift; + info->lr_unit_size[1] = 1 + frame_header->lr_unit_shift - frame_header->lr_uv_shift; + info->lr_unit_size[2] = 1 + frame_header->lr_unit_shift - frame_header->lr_uv_shift; + + /* Reference Frames */ + info->temporal_layer_id = s->cur_frame.temporal_id; + info->spatial_layer_id = s->cur_frame.spatial_id; + + /* Film Grain Params */ + info->apply_grain = film_grain->apply_grain; + info->overlap_flag = film_grain->overlap_flag; + info->scaling_shift_minus8 = film_grain->grain_scaling_minus_8; + info->chroma_scaling_from_luma = film_grain->chroma_scaling_from_luma; + info->ar_coeff_lag = film_grain->ar_coeff_lag; + info->ar_coeff_shift_minus6 = film_grain->ar_coeff_shift_minus_6; + info->grain_scale_shift = film_grain->grain_scale_shift; + info->clip_to_restricted_range = film_grain->clip_to_restricted_range; + info->num_y_points = film_grain->num_y_points; + info->num_cb_points = film_grain->num_cb_points; + info->num_cr_points = film_grain->num_cr_points; + info->random_seed = film_grain->grain_seed; + info->cb_mult = film_grain->cb_mult; + info->cb_luma_mult = film_grain->cb_luma_mult; + info->cb_offset = film_grain->cb_offset; + info->cr_mult = film_grain->cr_mult; + info->cr_luma_mult = film_grain->cr_luma_mult; + info->cr_offset = film_grain->cr_offset; + + /* Tiling Info */ + for (i = 0; i < frame_header->tile_cols; ++i) { + info->tile_widths[i] = frame_header->width_in_sbs_minus_1[i] + 1; + } + for (i = 0; i < frame_header->tile_rows; ++i) { + info->tile_heights[i] = frame_header->height_in_sbs_minus_1[i] + 1; + } + + /* CDEF */ + for (i = 0; i < (1 << frame_header->cdef_bits); ++i) { + info->cdef_y_strength[i] = (frame_header->cdef_y_pri_strength[i] & 0x0F) | (frame_header->cdef_y_sec_strength[i] << 4); + info->cdef_uv_strength[i] = (frame_header->cdef_uv_pri_strength[i] & 0x0F) | (frame_header->cdef_uv_sec_strength[i] << 4); + } + + + /* Segmentation */ + for (i = 0; i < AV1_MAX_SEGMENTS; ++i) { + info->segmentation_feature_mask[i] = 0; + for (j = 0; j < AV1_SEG_LVL_MAX; ++j) { + info->segmentation_feature_mask[i] |= frame_header->feature_enabled[i][j] << j; + info->segmentation_feature_data[i][j] = frame_header->feature_value[i][j]; + } + } + + for (i = 0; i < AV1_NUM_REF_FRAMES; ++i) { + /* Loopfilter */ + info->loop_filter_ref_deltas[i] = frame_header->loop_filter_ref_deltas[i]; + + /* Reference Frames */ + info->ref_frame_map[i] = ff_vdpau_get_surface_id(s->ref[i].tf.f) ? ff_vdpau_get_surface_id(s->ref[i].tf.f) : VDP_INVALID_HANDLE; + } + + if (frame_header->primary_ref_frame == AV1_PRIMARY_REF_NONE) { + info->primary_ref_frame = -1; + } else { + int8_t pri_ref_idx = frame_header->ref_frame_idx[frame_header->primary_ref_frame]; + info->primary_ref_frame = info->ref_frame_map[pri_ref_idx]; + } + + for (i = 0; i < AV1_REFS_PER_FRAME; ++i) { + /* Ref Frame List */ + int8_t ref_idx = frame_header->ref_frame_idx[i]; + AVFrame *ref_frame = s->ref[ref_idx].tf.f; + + info->ref_frame[i].index = info->ref_frame_map[ref_idx]; + info->ref_frame[i].width = ref_frame->width; + info->ref_frame[i].height = ref_frame->height; + + /* Global Motion */ + info->global_motion[i].invalid = !frame_header->is_global[AV1_REF_FRAME_LAST + i]; + info->global_motion[i].wmtype = s->cur_frame.gm_type[AV1_REF_FRAME_LAST + i]; + for (j = 0; j < 6; ++j) { + info->global_motion[i].wmmat[j] = s->cur_frame.gm_params[AV1_REF_FRAME_LAST + i][j]; + } + } + + /* Film Grain Params */ + if (film_grain->apply_grain) { + for (i = 0; i < 14; ++i) { + info->scaling_points_y[i][0] = film_grain->point_y_value[i]; + info->scaling_points_y[i][1] = film_grain->point_y_scaling[i]; + } + for (i = 0; i < 10; ++i) { + info->scaling_points_cb[i][0] = film_grain->point_cb_value[i]; + info->scaling_points_cb[i][1] = film_grain->point_cb_scaling[i]; + info->scaling_points_cr[i][0] = film_grain->point_cr_value[i]; + info->scaling_points_cr[i][1] = film_grain->point_cr_scaling[i]; + } + for (i = 0; i < 24; ++i) { + info->ar_coeffs_y[i] = (short)film_grain->ar_coeffs_y_plus_128[i] - 128; + } + for (i = 0; i < 25; ++i) { + info->ar_coeffs_cb[i] = (short)film_grain->ar_coeffs_cb_plus_128[i] - 128; + info->ar_coeffs_cr[i] = (short)film_grain->ar_coeffs_cr_plus_128[i] - 128; + } + } + + + return ff_vdpau_common_start_frame(pic_ctx, buffer, size); + +} + +static int vdpau_av1_decode_slice(AVCodecContext *avctx, + const uint8_t *buffer, uint32_t size) +{ + const AV1DecContext *s = avctx->priv_data; + const AV1RawFrameHeader *frame_header = s->raw_frame_header; + struct vdpau_picture_context *pic_ctx = s->cur_frame.hwaccel_picture_private; + VdpPictureInfoAV1 *info = &pic_ctx->info.av1; + int val; + int nb_slices; + VdpBitstreamBuffer *buffers = pic_ctx->bitstream_buffers; + int bitstream_len = 0; + + nb_slices = frame_header->tile_cols * frame_header->tile_rows; + /* Shortcut if all tiles are in the same buffer*/ + if (nb_slices == s->tg_end - s->tg_start + 1) { + for (int i = 0; i < nb_slices; ++i) { + info->tile_info[i*2 ] = s->tile_group_info[i].tile_offset; + info->tile_info[i*2 + 1] = info->tile_info[i*2] + s->tile_group_info[i].tile_size; + } + val = ff_vdpau_add_buffer(pic_ctx, buffer, size); + if (val) { + return val; + } + + return 0; + } + + for(int i = 0; i < pic_ctx->bitstream_buffers_used; i++) { + bitstream_len += buffers->bitstream_bytes; + buffers++; + } + + for (uint32_t tile_num = s->tg_start; tile_num <= s->tg_end; ++tile_num) { + info->tile_info[tile_num*2 ] = bitstream_len + s->tile_group_info[tile_num].tile_offset; + info->tile_info[tile_num*2 + 1] = info->tile_info[tile_num*2] + s->tile_group_info[tile_num].tile_size; + } + + val = ff_vdpau_add_buffer(pic_ctx, buffer, size); + if (val) { + return val; + } + + return 0; +} + +static int vdpau_av1_end_frame(AVCodecContext *avctx) +{ + const AV1DecContext *s = avctx->priv_data; + struct vdpau_picture_context *pic_ctx = s->cur_frame.hwaccel_picture_private; + + int val; + + val = ff_vdpau_common_end_frame(avctx, s->cur_frame.tf.f, pic_ctx); + if (val < 0) + return val; + + return 0; +} + +static int vdpau_av1_init(AVCodecContext *avctx) +{ + VdpDecoderProfile profile; + uint32_t level = avctx->level; + + switch (avctx->profile) { + case FF_PROFILE_AV1_MAIN: + profile = VDP_DECODER_PROFILE_AV1_MAIN; + break; + case FF_PROFILE_AV1_HIGH: + profile = VDP_DECODER_PROFILE_AV1_HIGH; + break; + case FF_PROFILE_AV1_PROFESSIONAL: + profile = VDP_DECODER_PROFILE_AV1_PROFESSIONAL; + break; + default: + return AVERROR(ENOTSUP); + } + + return ff_vdpau_common_init(avctx, profile, level); +} + +const AVHWAccel ff_av1_vdpau_hwaccel = { + .name = "av1_vdpau", + .type = AVMEDIA_TYPE_VIDEO, + .id = AV_CODEC_ID_AV1, + .pix_fmt = AV_PIX_FMT_VDPAU, + .start_frame = vdpau_av1_start_frame, + .end_frame = vdpau_av1_end_frame, + .decode_slice = vdpau_av1_decode_slice, + .frame_priv_data_size = sizeof(struct vdpau_picture_context), + .init = vdpau_av1_init, + .uninit = ff_vdpau_common_uninit, + .frame_params = ff_vdpau_common_frame_params, + .priv_data_size = sizeof(VDPAUContext), + .caps_internal = HWACCEL_CAP_ASYNC_SAFE, +}; diff --git a/libavcodec/vdpau_internal.h b/libavcodec/vdpau_internal.h index b6ea078cb2..347576e0c3 100644 --- a/libavcodec/vdpau_internal.h +++ b/libavcodec/vdpau_internal.h @@ -57,6 +57,9 @@ union VDPAUPictureInfo { #ifdef VDP_DECODER_PROFILE_VP9_PROFILE_0 VdpPictureInfoVP9 vp9; #endif +#ifdef VDP_DECODER_PROFILE_AV1_MAIN + VdpPictureInfoAV1 av1; +#endif }; typedef struct VDPAUHWContext { diff --git a/libavcodec/version.h b/libavcodec/version.h index 87b7284a95..5183deb68b 100644 --- a/libavcodec/version.h +++ b/libavcodec/version.h @@ -29,7 +29,7 @@ #include "version_major.h" -#define LIBAVCODEC_VERSION_MINOR 28 +#define LIBAVCODEC_VERSION_MINOR 29 #define LIBAVCODEC_VERSION_MICRO 100 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \