avcodec/nvdec: add av1 hwaccel

Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>
Co-authored-by: James Almer <jamrial@gmail.com>

7 files changed, 367 insertions, 2 deletions

diff --git a/libavcodec/Makefile b/libavcodec/Makefile
index 9d75dd68af..0c8aee39b3 100644
--- a/libavcodec/Makefile
+++ b/libavcodec/Makefile
@@ -912,6 +912,7 @@ OBJS-$(CONFIG_VAAPI)                      += vaapi_decode.o
 OBJS-$(CONFIG_VIDEOTOOLBOX)               += videotoolbox.o
 OBJS-$(CONFIG_VDPAU)                      += vdpau.o
 
+OBJS-$(CONFIG_AV1_NVDEC_HWACCEL)          += nvdec_av1.o
 OBJS-$(CONFIG_AV1_VAAPI_HWACCEL)          += vaapi_av1.o
 OBJS-$(CONFIG_H263_VAAPI_HWACCEL)         += vaapi_mpeg4.o
 OBJS-$(CONFIG_H263_VIDEOTOOLBOX_HWACCEL)  += videotoolbox.o
diff --git a/libavcodec/av1dec.c b/libavcodec/av1dec.c
index 7185ad827c..1078f708f7 100644
--- a/libavcodec/av1dec.c
+++ b/libavcodec/av1dec.c
@@ -317,7 +317,7 @@ static int get_pixel_format(AVCodecContext *avctx)
     uint8_t bit_depth;
     int ret;
     enum AVPixelFormat pix_fmt = AV_PIX_FMT_NONE;
-#define HWACCEL_MAX (CONFIG_AV1_VAAPI_HWACCEL)
+#define HWACCEL_MAX (CONFIG_AV1_NVDEC_HWACCEL + CONFIG_AV1_VAAPI_HWACCEL)
     enum AVPixelFormat pix_fmts[HWACCEL_MAX + 2], *fmtp = pix_fmts;
 
     if (seq->seq_profile == 2 && seq->color_config.high_bitdepth)
@@ -380,11 +380,17 @@ static int get_pixel_format(AVCodecContext *avctx)
 
     switch (s->pix_fmt) {
     case AV_PIX_FMT_YUV420P:
+#if CONFIG_AV1_NVDEC_HWACCEL
+        *fmtp++ = AV_PIX_FMT_CUDA;
+#endif
 #if CONFIG_AV1_VAAPI_HWACCEL
         *fmtp++ = AV_PIX_FMT_VAAPI;
 #endif
         break;
     case AV_PIX_FMT_YUV420P10:
+#if CONFIG_AV1_NVDEC_HWACCEL
+        *fmtp++ = AV_PIX_FMT_CUDA;
+#endif
 #if CONFIG_AV1_VAAPI_HWACCEL
         *fmtp++ = AV_PIX_FMT_VAAPI;
 #endif
@@ -968,6 +974,9 @@ AVCodec ff_av1_decoder = {
     .flush                 = av1_decode_flush,
     .profiles              = NULL_IF_CONFIG_SMALL(ff_av1_profiles),
     .hw_configs            = (const AVCodecHWConfigInternal * []) {
+#if CONFIG_AV1_NVDEC_HWACCEL
+        HWACCEL_NVDEC(av1),
+#endif
 #if CONFIG_AV1_VAAPI_HWACCEL
         HWACCEL_VAAPI(av1),
 #endif
diff --git a/libavcodec/hwaccels.h b/libavcodec/hwaccels.h
index 18e9079c55..6c698f6764 100644
--- a/libavcodec/hwaccels.h
+++ b/libavcodec/hwaccels.h
@@ -21,6 +21,7 @@
 
 #include "avcodec.h"
 
+extern const AVHWAccel ff_av1_nvdec_hwaccel;
 extern const AVHWAccel ff_av1_vaapi_hwaccel;
 extern const AVHWAccel ff_h263_vaapi_hwaccel;
 extern const AVHWAccel ff_h263_videotoolbox_hwaccel;
diff --git a/libavcodec/nvdec.c b/libavcodec/nvdec.c
index 52a431c54d..48281293ce 100644
--- a/libavcodec/nvdec.c
+++ b/libavcodec/nvdec.c
@@ -62,6 +62,9 @@ typedef struct NVDECFramePool {
 static int map_avcodec_id(enum AVCodecID id)
 {
     switch (id) {
+#if CONFIG_AV1_NVDEC_HWACCEL
+    case AV_CODEC_ID_AV1:        return cudaVideoCodec_AV1;
+#endif
     case AV_CODEC_ID_H264:       return cudaVideoCodec_H264;
     case AV_CODEC_ID_HEVC:       return cudaVideoCodec_HEVC;
     case AV_CODEC_ID_MJPEG:      return cudaVideoCodec_JPEG;
@@ -258,6 +261,7 @@ int ff_nvdec_decode_uninit(AVCodecContext *avctx)
     NVDECContext *ctx = avctx->internal->hwaccel_priv_data;
 
     av_freep(&ctx->bitstream);
+    av_freep(&ctx->bitstream_internal);
     ctx->bitstream_len       = 0;
     ctx->bitstream_allocated = 0;
 
diff --git a/libavcodec/nvdec.h b/libavcodec/nvdec.h
index 22b66d0bc8..66f3ca59e7 100644
--- a/libavcodec/nvdec.h
+++ b/libavcodec/nvdec.h
@@ -59,6 +59,7 @@ typedef struct NVDECContext {
     uint8_t      *bitstream;
     int           bitstream_len;
     unsigned int  bitstream_allocated;
+    uint8_t      *bitstream_internal;
 
     unsigned     *slice_offsets;
     int           nb_slices;
diff --git a/libavcodec/nvdec_av1.c b/libavcodec/nvdec_av1.c
new file mode 100644
index 0000000000..9da1c840e3
--- /dev/null
+++ b/libavcodec/nvdec_av1.c
@@ -0,0 +1,349 @@
+/*
+ * AV1 HW decode acceleration through NVDEC
+ *
+ * Copyright (c) 2020 Timo Rothenpieler
+ *
+ * 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 "avcodec.h"
+#include "nvdec.h"
+#include "decode.h"
+#include "internal.h"
+#include "av1dec.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 nvdec_av1_start_frame(AVCodecContext *avctx, const uint8_t *buffer, uint32_t size)
+{
+    const AV1DecContext *s = avctx->priv_data;
+    const AV1RawSequenceHeader *seq = s->raw_seq;
+    const AV1RawFrameHeader *frame_header = s->raw_frame_header;
+
+    NVDECContext      *ctx = avctx->internal->hwaccel_priv_data;
+    CUVIDPICPARAMS     *pp = &ctx->pic_params;
+    CUVIDAV1PICPARAMS *ppc = &pp->CodecSpecific.av1;
+    FrameDecodeData *fdd;
+    NVDECFrame *cf;
+    AVFrame *cur_frame = s->cur_frame.tf.f;
+
+    unsigned char remap_lr_type[4] = { AV1_RESTORE_NONE, AV1_RESTORE_SWITCHABLE, AV1_RESTORE_WIENER, AV1_RESTORE_SGRPROJ };
+
+    int ret, i, j;
+
+    ret = ff_nvdec_start_frame_sep_ref(avctx, cur_frame, frame_header->apply_grain);
+    if (ret < 0)
+        return ret;
+
+    fdd = (FrameDecodeData*)cur_frame->private_ref->data;
+    cf  = (NVDECFrame*)fdd->hwaccel_priv;
+
+    *pp = (CUVIDPICPARAMS) {
+        .PicWidthInMbs     = (cur_frame->width  + 15) / 16,
+        .FrameHeightInMbs  = (cur_frame->height + 15) / 16,
+        .CurrPicIdx        = cf->idx,
+        .ref_pic_flag      = !!frame_header->refresh_frame_flags,
+        .intra_pic_flag    = frame_header->frame_type == AV1_FRAME_INTRA_ONLY ||
+                             frame_header->frame_type == AV1_FRAME_KEY,
+
+        .CodecSpecific.av1 = {
+            .width  = cur_frame->width,
+            .height = cur_frame->height,
+
+            .frame_offset = frame_header->order_hint,
+            .decodePicIdx = cf->ref_idx,
+
+            /* Sequence Header */
+            .profile                    = seq->seq_profile,
+            .use_128x128_superblock     = seq->use_128x128_superblock,
+            .subsampling_x              = seq->color_config.subsampling_x,
+            .subsampling_y              = seq->color_config.subsampling_y,
+            .mono_chrome                = seq->color_config.mono_chrome,
+            .bit_depth_minus8           = get_bit_depth_from_seq(seq) - 8,
+            .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,
+            .order_hint_bits_minus1     = seq->order_hint_bits_minus_1,
+            .enable_jnt_comp            = seq->enable_jnt_comp,
+            .enable_superres            = seq->enable_superres,
+            .enable_cdef                = seq->enable_cdef,
+            .enable_restoration         = seq->enable_restoration,
+            .enable_fgs                 = seq->film_grain_params_present,
+
+            /* Frame Header */
+            .frame_type                   = frame_header->frame_type,
+            .show_frame                   = frame_header->show_frame,
+            .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 ||
+                                            frame_header->frame_type == AV1_FRAME_INTRA_ONLY ||
+                                            frame_header->frame_type == AV1_FRAME_KEY,
+            .coded_denom                  = frame_header->coded_denom,
+            .allow_intrabc                = frame_header->allow_intrabc,
+            .allow_high_precision_mv      = frame_header->allow_high_precision_mv,
+            .interp_filter                = frame_header->interpolation_filter,
+            .switchable_motion_mode       = 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,
+            .delta_q_present              = frame_header->delta_q_present,
+            .delta_q_res                  = frame_header->delta_q_res,
+            .using_qmatrix                = frame_header->using_qmatrix,
+            .coded_lossless               = s->cur_frame.coded_lossless,
+            .use_superres                 = frame_header->use_superres,
+            .tx_mode                      = frame_header->tx_mode,
+            .reference_mode               = frame_header->reference_select,
+            .allow_warped_motion          = frame_header->allow_warped_motion,
+            .reduced_tx_set               = frame_header->reduced_tx_set,
+            .skip_mode                    = frame_header->skip_mode_present,
+
+            /* Tiling Info */
+            .num_tile_cols          = frame_header->tile_cols,
+            .num_tile_rows          = frame_header->tile_rows,
+            .context_update_tile_id = frame_header->context_update_tile_id,
+
+            /* CDEF */
+            .cdef_damping_minus_3 = frame_header->cdef_damping_minus_3,
+            .cdef_bits            = frame_header->cdef_bits,
+
+            /* SkipModeFrames */
+            .SkipModeFrame0 = frame_header->skip_mode_present ?
+                              s->cur_frame.skip_mode_frame_idx[0] : 0,
+            .SkipModeFrame1 = frame_header->skip_mode_present ?
+                              s->cur_frame.skip_mode_frame_idx[1] : 0,
+
+            /* QP Information */
+            .base_qindex     = frame_header->base_q_idx,
+            .qp_y_dc_delta_q = frame_header->delta_q_y_dc,
+            .qp_u_dc_delta_q = frame_header->delta_q_u_dc,
+            .qp_v_dc_delta_q = frame_header->delta_q_v_dc,
+            .qp_u_ac_delta_q = frame_header->delta_q_u_ac,
+            .qp_v_ac_delta_q = frame_header->delta_q_v_ac,
+            .qm_y            = frame_header->qm_y,
+            .qm_u            = frame_header->qm_u,
+            .qm_v            = frame_header->qm_v,
+
+            /* Segmentation */
+            .segmentation_enabled         = frame_header->segmentation_enabled,
+            .segmentation_update_map      = frame_header->segmentation_update_map,
+            .segmentation_update_data     = frame_header->segmentation_update_data,
+            .segmentation_temporal_update = frame_header->segmentation_temporal_update,
+
+            /* Loopfilter */
+            .loop_filter_level[0]       = frame_header->loop_filter_level[0],
+            .loop_filter_level[1]       = frame_header->loop_filter_level[1],
+            .loop_filter_level_u        = frame_header->loop_filter_level[2],
+            .loop_filter_level_v        = frame_header->loop_filter_level[3],
+            .loop_filter_sharpness      = frame_header->loop_filter_sharpness,
+            .loop_filter_delta_enabled  = frame_header->loop_filter_delta_enabled,
+            .loop_filter_delta_update   = frame_header->loop_filter_delta_update,
+            .loop_filter_mode_deltas[0] = frame_header->loop_filter_mode_deltas[0],
+            .loop_filter_mode_deltas[1] = frame_header->loop_filter_mode_deltas[1],
+            .delta_lf_present           = frame_header->delta_lf_present,
+            .delta_lf_res               = frame_header->delta_lf_res,
+            .delta_lf_multi             = frame_header->delta_lf_multi,
+
+            /* Restoration */
+            .lr_type[0]      = remap_lr_type[frame_header->lr_type[0]],
+            .lr_type[1]      = remap_lr_type[frame_header->lr_type[1]],
+            .lr_type[2]      = remap_lr_type[frame_header->lr_type[2]],
+            .lr_unit_size[0] = 1 + frame_header->lr_unit_shift,
+            .lr_unit_size[1] = 1 + frame_header->lr_unit_shift - frame_header->lr_uv_shift,
+            .lr_unit_size[2] = 1 + frame_header->lr_unit_shift - frame_header->lr_uv_shift,
+
+            /* Reference Frames */
+            .temporal_layer_id = s->cur_frame.temporal_id,
+            .spatial_layer_id  = s->cur_frame.spatial_id,
+
+            /* Film Grain Params */
+            .apply_grain              = frame_header->apply_grain,
+            .overlap_flag             = frame_header->overlap_flag,
+            .scaling_shift_minus8     = frame_header->grain_scaling_minus_8,
+            .chroma_scaling_from_luma = frame_header->chroma_scaling_from_luma,
+            .ar_coeff_lag             = frame_header->ar_coeff_lag,
+            .ar_coeff_shift_minus6    = frame_header->ar_coeff_shift_minus_6,
+            .grain_scale_shift        = frame_header->grain_scale_shift,
+            .clip_to_restricted_range = frame_header->clip_to_restricted_range,
+            .num_y_points             = frame_header->num_y_points,
+            .num_cb_points            = frame_header->num_cb_points,
+            .num_cr_points            = frame_header->num_cr_points,
+            .random_seed              = frame_header->grain_seed,
+            .cb_mult                  = frame_header->cb_mult,
+            .cb_luma_mult             = frame_header->cb_luma_mult,
+            .cb_offset                = frame_header->cb_offset,
+            .cr_mult                  = frame_header->cr_mult,
+            .cr_luma_mult             = frame_header->cr_luma_mult,
+            .cr_offset                = frame_header->cr_offset
+        }
+    };
+
+    /* Tiling Info */
+    for (i = 0; i < frame_header->tile_cols; ++i) {
+        ppc->tile_widths[i] = frame_header->width_in_sbs_minus_1[i] + 1;
+    }
+    for (i = 0; i < frame_header->tile_rows; ++i) {
+        ppc->tile_heights[i] = frame_header->height_in_sbs_minus_1[i] + 1;
+    }
+
+    /* CDEF */
+    for (i = 0; i < (1 << frame_header->cdef_bits); ++i) {
+        ppc->cdef_y_strength[i] = (frame_header->cdef_y_pri_strength[i] & 0x0F) | (frame_header->cdef_y_sec_strength[i] << 4);
+        ppc->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) {
+        ppc->segmentation_feature_mask[i] = 0;
+        for (j = 0; j < AV1_SEG_LVL_MAX; ++j) {
+            ppc->segmentation_feature_mask[i] |= frame_header->feature_enabled[i][j] << j;
+            ppc->segmentation_feature_data[i][j] = frame_header->feature_value[i][j];
+        }
+    }
+
+    for (i = 0; i < AV1_NUM_REF_FRAMES; ++i) {
+        /* Loopfilter */
+        ppc->loop_filter_ref_deltas[i] = frame_header->loop_filter_ref_deltas[i];
+
+        /* Reference Frames */
+        ppc->ref_frame_map[i] = ff_nvdec_get_ref_idx(s->ref[i].tf.f);
+    }
+
+    if (frame_header->primary_ref_frame == AV1_PRIMARY_REF_NONE) {
+        ppc->primary_ref_frame = -1;
+    } else {
+        int8_t pri_ref_idx = frame_header->ref_frame_idx[frame_header->primary_ref_frame];
+        ppc->primary_ref_frame = ppc->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;
+
+        ppc->ref_frame[i].index = ppc->ref_frame_map[ref_idx];
+        ppc->ref_frame[i].width = ref_frame->width;
+        ppc->ref_frame[i].height = ref_frame->height;
+
+        /* Global Motion */
+        ppc->global_motion[i].invalid = !frame_header->is_global[AV1_REF_FRAME_LAST + i];
+        ppc->global_motion[i].wmtype = s->cur_frame.gm_type[AV1_REF_FRAME_LAST + i];
+        for (j = 0; j < 6; ++j) {
+            ppc->global_motion[i].wmmat[j] = s->cur_frame.gm_params[AV1_REF_FRAME_LAST + i][j];
+        }
+    }
+
+    /* Film Grain Params */
+    if (frame_header->apply_grain) {
+        for (i = 0; i < 14; ++i) {
+            ppc->scaling_points_y[i][0] = frame_header->point_y_value[i];
+            ppc->scaling_points_y[i][1] = frame_header->point_y_scaling[i];
+        }
+        for (i = 0; i < 10; ++i) {
+            ppc->scaling_points_cb[i][0] = frame_header->point_cb_value[i];
+            ppc->scaling_points_cb[i][1] = frame_header->point_cb_scaling[i];
+            ppc->scaling_points_cr[i][0] = frame_header->point_cr_value[i];
+            ppc->scaling_points_cr[i][1] = frame_header->point_cr_scaling[i];
+        }
+        for (i = 0; i < 24; ++i) {
+            ppc->ar_coeffs_y[i] = (short)frame_header->ar_coeffs_y_plus_128[i] - 128;
+        }
+        for (i = 0; i < 25; ++i) {
+            ppc->ar_coeffs_cb[i] = (short)frame_header->ar_coeffs_cb_plus_128[i] - 128;
+            ppc->ar_coeffs_cr[i] = (short)frame_header->ar_coeffs_cr_plus_128[i] - 128;
+        }
+    }
+
+    return 0;
+}
+
+static int nvdec_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;
+    NVDECContext *ctx = avctx->internal->hwaccel_priv_data;
+    void *tmp;
+
+    ctx->nb_slices = frame_header->tile_cols * frame_header->tile_rows;
+
+    tmp = av_fast_realloc(ctx->slice_offsets, &ctx->slice_offsets_allocated,
+                          ctx->nb_slices * 2 * sizeof(*ctx->slice_offsets));
+    if (!tmp) {
+        return AVERROR(ENOMEM);
+    }
+    ctx->slice_offsets = tmp;
+
+    /* Shortcut if all tiles are in the same buffer */
+    if (ctx->nb_slices == s->tg_end - s->tg_start + 1) {
+        ctx->bitstream = (uint8_t*)buffer;
+        ctx->bitstream_len = size;
+
+        for (int i = 0; i < ctx->nb_slices; ++i) {
+            ctx->slice_offsets[i*2    ] = s->tile_group_info[i].tile_offset;
+            ctx->slice_offsets[i*2 + 1] = ctx->slice_offsets[i*2] + s->tile_group_info[i].tile_size;
+        }
+
+        return 0;
+    }
+
+    tmp = av_fast_realloc(ctx->bitstream_internal, &ctx->bitstream_allocated,
+                          ctx->bitstream_len + size);
+    if (!tmp) {
+        return AVERROR(ENOMEM);
+    }
+    ctx->bitstream = ctx->bitstream_internal = tmp;
+
+    memcpy(ctx->bitstream + ctx->bitstream_len, buffer, size);
+
+    for (uint32_t tile_num = s->tg_start; tile_num <= s->tg_end; ++tile_num) {
+        ctx->slice_offsets[tile_num*2    ] = ctx->bitstream_len + s->tile_group_info[tile_num].tile_offset;
+        ctx->slice_offsets[tile_num*2 + 1] = ctx->slice_offsets[tile_num*2] + s->tile_group_info[tile_num].tile_size;
+    }
+    ctx->bitstream_len += size;
+
+    return 0;
+}
+
+static int nvdec_av1_frame_params(AVCodecContext *avctx, AVBufferRef *hw_frames_ctx)
+{
+    /* Maximum of 8 reference frames, but potentially stored twice due to film grain */
+    return ff_nvdec_frame_params(avctx, hw_frames_ctx, 8 * 2, 0);
+}
+
+const AVHWAccel ff_av1_nvdec_hwaccel = {
+    .name                 = "av1_nvdec",
+    .type                 = AVMEDIA_TYPE_VIDEO,
+    .id                   = AV_CODEC_ID_AV1,
+    .pix_fmt              = AV_PIX_FMT_CUDA,
+    .start_frame          = nvdec_av1_start_frame,
+    .end_frame            = ff_nvdec_simple_end_frame,
+    .decode_slice         = nvdec_av1_decode_slice,
+    .frame_params         = nvdec_av1_frame_params,
+    .init                 = ff_nvdec_decode_init,
+    .uninit               = ff_nvdec_decode_uninit,
+    .priv_data_size       = sizeof(NVDECContext),
+};
diff --git a/libavcodec/version.h b/libavcodec/version.h
index 5173d0f090..a595e32832 100644
--- a/libavcodec/version.h
+++ b/libavcodec/version.h
@@ -29,7 +29,7 @@
 
 #define LIBAVCODEC_VERSION_MAJOR  58
 #define LIBAVCODEC_VERSION_MINOR 112
-#define LIBAVCODEC_VERSION_MICRO 101
+#define LIBAVCODEC_VERSION_MICRO 102
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
                                                LIBAVCODEC_VERSION_MINOR, \