1 From 998aa66a10546be207d5dfc1a4a76bc2ce9ea07c Mon Sep 17 00:00:00 2001
2 From: Lynne <dev@lynne.ee>
3 Date: Sun, 17 Mar 2024 19:31:17 +0100
4 Subject: [PATCH] av1dec: add AV1_REF_FRAME_NONE
8 libavcodec/cbs_av1_syntax_template.c | 12 ++++++------
9 2 files changed, 7 insertions(+), 6 deletions(-)
11 diff --git a/libavcodec/av1.h b/libavcodec/av1.h
12 index 8704bc41c1..94e88f8484 100644
13 --- a/libavcodec/av1.h
14 +++ b/libavcodec/av1.h
15 @@ -58,6 +58,7 @@ enum {
17 // Reference frames (section 6.10.24).
19 + AV1_REF_FRAME_NONE = -1,
20 AV1_REF_FRAME_INTRA = 0,
21 AV1_REF_FRAME_LAST = 1,
22 AV1_REF_FRAME_LAST2 = 2,
23 diff --git a/libavcodec/cbs_av1_syntax_template.c b/libavcodec/cbs_av1_syntax_template.c
24 index 2979c5d98f..3f4b13a177 100644
25 --- a/libavcodec/cbs_av1_syntax_template.c
26 +++ b/libavcodec/cbs_av1_syntax_template.c
27 @@ -360,7 +360,7 @@ static int FUNC(set_frame_refs)(CodedBitstreamContext *ctx, RWContext *rw,
30 for (i = 0; i < AV1_REFS_PER_FRAME; i++)
31 - ref_frame_idx[i] = -1;
32 + ref_frame_idx[i] = AV1_REF_FRAME_NONE;
33 ref_frame_idx[AV1_REF_FRAME_LAST - AV1_REF_FRAME_LAST] = current->last_frame_idx;
34 ref_frame_idx[AV1_REF_FRAME_GOLDEN - AV1_REF_FRAME_LAST] = current->golden_frame_idx;
36 @@ -378,7 +378,7 @@ static int FUNC(set_frame_refs)(CodedBitstreamContext *ctx, RWContext *rw,
37 latest_order_hint = shifted_order_hints[current->last_frame_idx];
38 earliest_order_hint = shifted_order_hints[current->golden_frame_idx];
41 + ref = AV1_REF_FRAME_NONE;
42 for (i = 0; i < AV1_NUM_REF_FRAMES; i++) {
43 int hint = shifted_order_hints[i];
44 if (!used_frame[i] && hint >= cur_frame_hint &&
45 @@ -392,7 +392,7 @@ static int FUNC(set_frame_refs)(CodedBitstreamContext *ctx, RWContext *rw,
50 + ref = AV1_REF_FRAME_NONE;
51 for (i = 0; i < AV1_NUM_REF_FRAMES; i++) {
52 int hint = shifted_order_hints[i];
53 if (!used_frame[i] && hint >= cur_frame_hint &&
54 @@ -406,7 +406,7 @@ static int FUNC(set_frame_refs)(CodedBitstreamContext *ctx, RWContext *rw,
59 + ref = AV1_REF_FRAME_NONE;
60 for (i = 0; i < AV1_NUM_REF_FRAMES; i++) {
61 int hint = shifted_order_hints[i];
62 if (!used_frame[i] && hint >= cur_frame_hint &&
63 @@ -423,7 +423,7 @@ static int FUNC(set_frame_refs)(CodedBitstreamContext *ctx, RWContext *rw,
64 for (i = 0; i < AV1_REFS_PER_FRAME - 2; i++) {
65 int ref_frame = ref_frame_list[i];
66 if (ref_frame_idx[ref_frame - AV1_REF_FRAME_LAST] < 0 ) {
68 + ref = AV1_REF_FRAME_NONE;
69 for (j = 0; j < AV1_NUM_REF_FRAMES; j++) {
70 int hint = shifted_order_hints[j];
71 if (!used_frame[j] && hint < cur_frame_hint &&
72 @@ -439,7 +439,7 @@ static int FUNC(set_frame_refs)(CodedBitstreamContext *ctx, RWContext *rw,
77 + ref = AV1_REF_FRAME_NONE;
78 for (i = 0; i < AV1_NUM_REF_FRAMES; i++) {
79 int hint = shifted_order_hints[i];
80 if (ref < 0 || hint < earliest_order_hint) {
84 From cafb4c554845332eeb33284cf6498049997dc67e Mon Sep 17 00:00:00 2001
85 From: Mark Thompson <sw@jkqxz.net>
86 Date: Wed, 20 Mar 2024 20:35:28 +0000
87 Subject: [PATCH] lavc/cbs_av1: Save more frame ordering information
89 This is wanted by the Vulkan decoder.
91 libavcodec/cbs_av1.h | 5 +++++
92 libavcodec/cbs_av1_syntax_template.c | 25 +++++++++++++++++++++----
93 2 files changed, 26 insertions(+), 4 deletions(-)
95 diff --git a/libavcodec/cbs_av1.h b/libavcodec/cbs_av1.h
96 index a5402f069d..a027013bc7 100644
97 --- a/libavcodec/cbs_av1.h
98 +++ b/libavcodec/cbs_av1.h
99 @@ -427,6 +427,8 @@ typedef struct AV1ReferenceFrameState {
100 int bit_depth; // RefBitDepth
101 int order_hint; // RefOrderHint
103 + int saved_order_hints[AV1_TOTAL_REFS_PER_FRAME]; // SavedOrderHints[ref]
105 int8_t loop_filter_ref_deltas[AV1_TOTAL_REFS_PER_FRAME];
106 int8_t loop_filter_mode_deltas[2];
107 uint8_t feature_enabled[AV1_MAX_SEGMENTS][AV1_SEG_LVL_MAX];
108 @@ -464,6 +466,9 @@ typedef struct CodedBitstreamAV1Context {
112 + int order_hints[AV1_TOTAL_REFS_PER_FRAME]; // OrderHints
113 + int ref_frame_sign_bias[AV1_TOTAL_REFS_PER_FRAME]; // RefFrameSignBias
115 AV1ReferenceFrameState ref[AV1_NUM_REF_FRAMES];
118 diff --git a/libavcodec/cbs_av1_syntax_template.c b/libavcodec/cbs_av1_syntax_template.c
119 index 3be1f2d30f..2979c5d98f 100644
120 --- a/libavcodec/cbs_av1_syntax_template.c
121 +++ b/libavcodec/cbs_av1_syntax_template.c
122 @@ -1414,6 +1414,8 @@ static int FUNC(uncompressed_header)(CodedBitstreamContext *ctx, RWContext *rw,
123 priv->ref[i].valid = 0;
124 priv->ref[i].order_hint = 0;
126 + for (i = 0; i < AV1_REFS_PER_FRAME; i++)
127 + priv->order_hints[i + AV1_REF_FRAME_LAST] = 0;
130 flag(disable_cdf_update);
131 @@ -1568,11 +1570,20 @@ static int FUNC(uncompressed_header)(CodedBitstreamContext *ctx, RWContext *rw,
133 flag(use_ref_frame_mvs);
135 - infer(allow_intrabc, 0);
137 + for (i = 0; i < AV1_REFS_PER_FRAME; i++) {
138 + int ref_frame = AV1_REF_FRAME_LAST + i;
139 + int hint = priv->ref[current->ref_frame_idx[i]].order_hint;
140 + priv->order_hints[ref_frame] = hint;
141 + if (!seq->enable_order_hint) {
142 + priv->ref_frame_sign_bias[ref_frame] = 0;
144 + priv->ref_frame_sign_bias[ref_frame] =
145 + cbs_av1_get_relative_dist(seq, hint,
146 + current->order_hint) > 0;
150 - if (!frame_is_intra) {
151 - // Derive reference frame sign biases.
152 + infer(allow_intrabc, 0);
155 if (seq->reduced_still_picture_header || current->disable_cdf_update)
156 @@ -1674,6 +1685,12 @@ update_refs:
157 .bit_depth = priv->bit_depth,
158 .order_hint = priv->order_hint,
161 + for (int j = 0; j < AV1_REFS_PER_FRAME; j++) {
162 + priv->ref[i].saved_order_hints[j + AV1_REF_FRAME_LAST] =
163 + priv->order_hints[j + AV1_REF_FRAME_LAST];
166 memcpy(priv->ref[i].loop_filter_ref_deltas, current->loop_filter_ref_deltas,
167 sizeof(current->loop_filter_ref_deltas));
168 memcpy(priv->ref[i].loop_filter_mode_deltas, current->loop_filter_mode_deltas,
172 From: Lynne <dev@lynne.ee>
173 Date: Fri, 19 Jan 2024 00:49:02 +0000 (+1000)
174 Subject: vulkan_av1: port to the new stable API
176 X-Git-Url: http://git.ffmpeg.org/gitweb/ffmpeg.git/commitdiff_plain/ecdc94b97f809d5f2b88640842fd0541951ad295
178 vulkan_av1: port to the new stable API
180 Co-Authored-by: Dave Airlie <airlied@redhat.com>
183 --- ffmpeg-6.1.1/configure.orig 2024-04-15 21:35:22.596075876 +0200
184 +++ ffmpeg-6.1.1/configure 2024-04-15 21:41:06.030881997 +0200
185 @@ -7150,8 +7150,8 @@ enabled crystalhd && check_lib crystalhd
188 if enabled vulkan; then
189 - check_pkg_config_header_only vulkan "vulkan >= 1.3.255" "vulkan/vulkan.h" "defined VK_VERSION_1_3" ||
190 - check_cpp_condition vulkan "vulkan/vulkan.h" "defined(VK_VERSION_1_4) || (defined(VK_VERSION_1_3) && VK_HEADER_VERSION >= 255)"
191 + check_pkg_config_header_only vulkan "vulkan >= 1.3.277" "vulkan/vulkan.h" "defined VK_VERSION_1_3" ||
192 + check_cpp_condition vulkan "vulkan/vulkan.h" "defined(VK_VERSION_1_4) || (defined(VK_VERSION_1_3) && VK_HEADER_VERSION >= 277)"
195 if disabled vulkan; then
196 --- ffmpeg-6.1.1/libavcodec/Makefile.orig 2024-04-15 21:35:22.596075876 +0200
197 +++ ffmpeg-6.1.1/libavcodec/Makefile 2024-04-15 21:43:51.719984382 +0200
198 @@ -1284,7 +1284,6 @@ SKIPHEADERS +
199 aacenc_quantization.h \
200 aacenc_quantization_misc.h \
201 bitstream_template.h \
202 - vulkan_video_codec_av1std.h \
203 $(ARCH)/vpx_arith.h \
205 SKIPHEADERS-$(CONFIG_AMF) += amfenc.h
206 @@ -1306,7 +1305,7 @@ SKIPHEADERS-$(CONFIG_XVMC) +
207 SKIPHEADERS-$(CONFIG_VAAPI) += vaapi_decode.h vaapi_hevc.h vaapi_encode.h
208 SKIPHEADERS-$(CONFIG_VDPAU) += vdpau.h vdpau_internal.h
209 SKIPHEADERS-$(CONFIG_VIDEOTOOLBOX) += videotoolbox.h vt_internal.h
210 -SKIPHEADERS-$(CONFIG_VULKAN) += vulkan.h vulkan_video.h vulkan_decode.h vulkan_video_codec_av1std_decode.h
211 +SKIPHEADERS-$(CONFIG_VULKAN) += vulkan.h vulkan_video.h vulkan_decode.h
212 SKIPHEADERS-$(CONFIG_V4L2_M2M) += v4l2_buffers.h v4l2_context.h v4l2_m2m.h
213 SKIPHEADERS-$(CONFIG_ZLIB) += zlib_wrapper.h
215 diff --git a/libavcodec/vulkan_av1.c b/libavcodec/vulkan_av1.c
216 index 5afd5353cc..c9e398eaec 100644
217 --- a/libavcodec/vulkan_av1.c
218 +++ b/libavcodec/vulkan_av1.c
219 @@ -36,33 +36,47 @@ const FFVulkanDecodeDescriptor ff_vk_dec_av1_desc = {
220 typedef struct AV1VulkanDecodePicture {
221 FFVulkanDecodePicture vp;
223 - /* Workaround for a spec issue.
224 - *Can be removed once no longer needed, and threading can be enabled. */
225 + /* TODO: investigate if this can be removed to make decoding completely
227 FFVulkanDecodeContext *dec;
229 - StdVideoAV1MESATile tiles[MAX_TILES];
230 - StdVideoAV1MESATileList tile_list;
231 - const uint32_t *tile_offsets;
232 + uint32_t tile_sizes[MAX_TILES];
234 /* Current picture */
235 - VkVideoDecodeAV1DpbSlotInfoMESA vkav1_ref;
236 - StdVideoAV1MESAFrameHeader av1_frame_header;
237 - VkVideoDecodeAV1PictureInfoMESA av1_pic_info;
238 + StdVideoDecodeAV1ReferenceInfo std_ref;
239 + VkVideoDecodeAV1DpbSlotInfoKHR vkav1_ref;
240 + uint16_t width_in_sbs_minus1[64];
241 + uint16_t height_in_sbs_minus1[64];
242 + uint16_t mi_col_starts[64];
243 + uint16_t mi_row_starts[64];
244 + StdVideoAV1TileInfo tile_info;
245 + StdVideoAV1Quantization quantization;
246 + StdVideoAV1Segmentation segmentation;
247 + StdVideoAV1LoopFilter loop_filter;
248 + StdVideoAV1CDEF cdef;
249 + StdVideoAV1LoopRestoration loop_restoration;
250 + StdVideoAV1GlobalMotion global_motion;
251 + StdVideoAV1FilmGrain film_grain;
252 + StdVideoDecodeAV1PictureInfo std_pic_info;
253 + VkVideoDecodeAV1PictureInfoKHR av1_pic_info;
256 const AV1Frame *ref_src [AV1_NUM_REF_FRAMES];
257 - VkVideoDecodeAV1DpbSlotInfoMESA vkav1_refs[AV1_NUM_REF_FRAMES];
258 + StdVideoDecodeAV1ReferenceInfo std_refs [AV1_NUM_REF_FRAMES];
259 + VkVideoDecodeAV1DpbSlotInfoKHR vkav1_refs[AV1_NUM_REF_FRAMES];
261 uint8_t frame_id_set;
263 + uint8_t ref_frame_sign_bias_mask;
264 } AV1VulkanDecodePicture;
266 static int vk_av1_fill_pict(AVCodecContext *avctx, const AV1Frame **ref_src,
267 VkVideoReferenceSlotInfoKHR *ref_slot, /* Main structure */
268 VkVideoPictureResourceInfoKHR *ref, /* Goes in ^ */
269 - VkVideoDecodeAV1DpbSlotInfoMESA *vkav1_ref, /* Goes in ^ */
270 + StdVideoDecodeAV1ReferenceInfo *vkav1_std_ref,
271 + VkVideoDecodeAV1DpbSlotInfoKHR *vkav1_ref, /* Goes in ^ */
272 const AV1Frame *pic, int is_current, int has_grain,
273 - int dpb_slot_index)
274 + int *saved_order_hints)
276 FFVulkanDecodeContext *dec = avctx->internal->hwaccel_priv_data;
277 AV1VulkanDecodePicture *hp = pic->hwaccel_picture_private;
278 @@ -73,31 +87,42 @@ static int vk_av1_fill_pict(AVCodecContext *avctx, const AV1Frame **ref_src,
282 - *vkav1_ref = (VkVideoDecodeAV1DpbSlotInfoMESA) {
283 - .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_DPB_SLOT_INFO_MESA,
284 - .frameIdx = hp->frame_id,
285 + *vkav1_std_ref = (StdVideoDecodeAV1ReferenceInfo) {
286 + .flags = (StdVideoDecodeAV1ReferenceInfoFlags) {
287 + .disable_frame_end_update_cdf = pic->raw_frame_header->disable_frame_end_update_cdf,
288 + .segmentation_enabled = pic->raw_frame_header->segmentation_enabled,
290 + .frame_type = pic->raw_frame_header->frame_type,
291 + .OrderHint = pic->raw_frame_header->order_hint,
292 + .RefFrameSignBias = hp->ref_frame_sign_bias_mask,
295 - for (unsigned i = 0; i < 7; i++) {
296 - const int idx = pic->raw_frame_header->ref_frame_idx[i];
297 - vkav1_ref->ref_order_hint[i] = pic->raw_frame_header->ref_order_hint[idx];
299 + if (saved_order_hints)
300 + for (int i = 0; i < AV1_TOTAL_REFS_PER_FRAME; i++)
301 + vkav1_std_ref->SavedOrderHints[i] = saved_order_hints[i];
303 + *vkav1_ref = (VkVideoDecodeAV1DpbSlotInfoKHR) {
304 + .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_DPB_SLOT_INFO_KHR,
305 + .pStdReferenceInfo = vkav1_std_ref,
308 - vkav1_ref->disable_frame_end_update_cdf = pic->raw_frame_header->disable_frame_end_update_cdf;
309 + vkav1_std_ref->flags.disable_frame_end_update_cdf = pic->raw_frame_header->disable_frame_end_update_cdf;
310 + vkav1_std_ref->flags.segmentation_enabled = pic->raw_frame_header->segmentation_enabled;
311 + vkav1_std_ref->frame_type = pic->raw_frame_header->frame_type;
313 *ref = (VkVideoPictureResourceInfoKHR) {
314 .sType = VK_STRUCTURE_TYPE_VIDEO_PICTURE_RESOURCE_INFO_KHR,
315 .codedOffset = (VkOffset2D){ 0, 0 },
316 .codedExtent = (VkExtent2D){ pic->f->width, pic->f->height },
317 .baseArrayLayer = ((has_grain || dec->dedicated_dpb) && dec->layered_dpb) ?
318 - dpb_slot_index : 0,
320 .imageViewBinding = vkpic->img_view_ref,
323 *ref_slot = (VkVideoReferenceSlotInfoKHR) {
324 .sType = VK_STRUCTURE_TYPE_VIDEO_REFERENCE_SLOT_INFO_KHR,
326 - .slotIndex = dpb_slot_index,
327 + .slotIndex = hp->frame_id,
328 .pPictureResource = ref,
331 @@ -115,15 +140,40 @@ static int vk_av1_create_params(AVCodecContext *avctx, AVBufferRef **buf)
333 const AV1RawSequenceHeader *seq = s->raw_seq;
335 - StdVideoAV1MESASequenceHeader av1_sequence_header;
336 - VkVideoDecodeAV1SessionParametersAddInfoMESA av1_params_info;
337 - VkVideoDecodeAV1SessionParametersCreateInfoMESA av1_params;
338 + StdVideoAV1SequenceHeader av1_sequence_header;
339 + StdVideoAV1TimingInfo av1_timing_info;
340 + StdVideoAV1ColorConfig av1_color_config;
341 + VkVideoDecodeAV1SessionParametersCreateInfoKHR av1_params;
342 VkVideoSessionParametersCreateInfoKHR session_params_create;
346 - av1_sequence_header = (StdVideoAV1MESASequenceHeader) {
347 - .flags = (StdVideoAV1MESASequenceHeaderFlags) {
348 + av1_timing_info = (StdVideoAV1TimingInfo) {
349 + .flags = (StdVideoAV1TimingInfoFlags) {
350 + .equal_picture_interval = seq->timing_info.equal_picture_interval,
352 + .num_units_in_display_tick = seq->timing_info.num_units_in_display_tick,
353 + .time_scale = seq->timing_info.time_scale,
354 + .num_ticks_per_picture_minus_1 = seq->timing_info.num_ticks_per_picture_minus_1,
357 + av1_color_config = (StdVideoAV1ColorConfig) {
358 + .flags = (StdVideoAV1ColorConfigFlags) {
359 + .mono_chrome = seq->color_config.mono_chrome,
360 + .color_range = seq->color_config.color_range,
361 + .separate_uv_delta_q = seq->color_config.separate_uv_delta_q,
363 + .BitDepth = seq->color_config.twelve_bit ? 12 :
364 + seq->color_config.high_bitdepth ? 10 : 8,
365 + .subsampling_x = seq->color_config.subsampling_x,
366 + .subsampling_y = seq->color_config.subsampling_y,
367 + .color_primaries = seq->color_config.color_primaries,
368 + .transfer_characteristics = seq->color_config.transfer_characteristics,
369 + .matrix_coefficients = seq->color_config.matrix_coefficients,
372 + av1_sequence_header = (StdVideoAV1SequenceHeader) {
373 + .flags = (StdVideoAV1SequenceHeaderFlags) {
374 .still_picture = seq->still_picture,
375 .reduced_still_picture_header = seq->reduced_still_picture_header,
376 .use_128x128_superblock = seq->use_128x128_superblock,
377 @@ -152,34 +202,15 @@ static int vk_av1_create_params(AVCodecContext *avctx, AVBufferRef **buf)
378 .delta_frame_id_length_minus_2 = seq->delta_frame_id_length_minus_2,
379 .additional_frame_id_length_minus_1 = seq->additional_frame_id_length_minus_1,
380 .order_hint_bits_minus_1 = seq->order_hint_bits_minus_1,
381 - .timing_info = (StdVideoAV1MESATimingInfo) {
382 - .flags = (StdVideoAV1MESATimingInfoFlags) {
383 - .equal_picture_interval = seq->timing_info.equal_picture_interval,
385 - .num_units_in_display_tick = seq->timing_info.num_units_in_display_tick,
386 - .time_scale = seq->timing_info.time_scale,
387 - .num_ticks_per_picture_minus_1 = seq->timing_info.num_ticks_per_picture_minus_1,
389 - .color_config = (StdVideoAV1MESAColorConfig) {
390 - .flags = (StdVideoAV1MESAColorConfigFlags) {
391 - .mono_chrome = seq->color_config.mono_chrome,
392 - .color_range = seq->color_config.color_range,
393 - .separate_uv_delta_q = seq->color_config.separate_uv_delta_q,
395 - .bit_depth = seq->color_config.twelve_bit ? 12 :
396 - seq->color_config.high_bitdepth ? 10 : 8,
397 - .subsampling_x = seq->color_config.subsampling_x,
398 - .subsampling_y = seq->color_config.subsampling_y,
400 + .seq_force_integer_mv = seq->seq_force_integer_mv,
401 + .seq_force_screen_content_tools = seq->seq_force_screen_content_tools,
402 + .pTimingInfo = &av1_timing_info,
403 + .pColorConfig = &av1_color_config,
406 - av1_params_info = (VkVideoDecodeAV1SessionParametersAddInfoMESA) {
407 - .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_SESSION_PARAMETERS_ADD_INFO_MESA,
408 - .sequence_header = &av1_sequence_header,
410 - av1_params = (VkVideoDecodeAV1SessionParametersCreateInfoMESA) {
411 - .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_SESSION_PARAMETERS_CREATE_INFO_MESA,
412 - .pParametersAddInfo = &av1_params_info,
413 + av1_params = (VkVideoDecodeAV1SessionParametersCreateInfoKHR) {
414 + .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_SESSION_PARAMETERS_CREATE_INFO_KHR,
415 + .pStdSequenceHeader = &av1_sequence_header,
417 session_params_create = (VkVideoSessionParametersCreateInfoKHR) {
418 .sType = VK_STRUCTURE_TYPE_VIDEO_SESSION_PARAMETERS_CREATE_INFO_KHR,
419 @@ -211,8 +242,14 @@ static int vk_av1_start_frame(AVCodecContext *avctx,
421 const AV1RawFrameHeader *frame_header = s->raw_frame_header;
422 const AV1RawFilmGrainParams *film_grain = &s->cur_frame.film_grain;
423 + CodedBitstreamAV1Context *cbs_ctx = (CodedBitstreamAV1Context *)(s->cbc->priv_data);
425 const int apply_grain = !(avctx->export_side_data & AV_CODEC_EXPORT_DATA_FILM_GRAIN) &&
426 film_grain->apply_grain;
427 + StdVideoAV1FrameRestorationType remap_lr_type[4] = { STD_VIDEO_AV1_FRAME_RESTORATION_TYPE_NONE,
428 + STD_VIDEO_AV1_FRAME_RESTORATION_TYPE_SWITCHABLE,
429 + STD_VIDEO_AV1_FRAME_RESTORATION_TYPE_WIENER,
430 + STD_VIDEO_AV1_FRAME_RESTORATION_TYPE_SGRPROJ };
432 if (!dec->session_params) {
433 err = vk_av1_create_params(avctx, &dec->session_params);
434 @@ -233,15 +270,31 @@ static int vk_av1_start_frame(AVCodecContext *avctx,
435 dec->frame_id_alloc_mask |= (1 << slot_idx);
438 - /* Fill in references */
439 - for (int i = 0; i < AV1_NUM_REF_FRAMES; i++) {
440 - const AV1Frame *ref_frame = &s->ref[i];
441 - if (s->ref[i].f->pict_type == AV_PICTURE_TYPE_NONE)
442 + ap->ref_frame_sign_bias_mask = 0x0;
443 + for (int i = 0; i < STD_VIDEO_AV1_TOTAL_REFS_PER_FRAME; i++)
444 + ap->ref_frame_sign_bias_mask |= cbs_ctx->ref_frame_sign_bias[i] << i;
446 + for (int i = 0; i < STD_VIDEO_AV1_REFS_PER_FRAME; i++) {
447 + const int idx = pic->raw_frame_header->ref_frame_idx[i];
448 + const AV1Frame *ref_frame = &s->ref[idx];
449 + AV1VulkanDecodePicture *hp = ref_frame->hwaccel_picture_private;
452 + if (ref_frame->f->pict_type == AV_PICTURE_TYPE_NONE)
455 + for (int j = 0; j < ref_count; j++) {
456 + if (vp->ref_slots[j].slotIndex == hp->frame_id) {
464 - err = vk_av1_fill_pict(avctx, &ap->ref_src[i], &vp->ref_slots[i],
465 - &vp->refs[i], &ap->vkav1_refs[i],
466 - ref_frame, 0, 0, i);
467 + err = vk_av1_fill_pict(avctx, &ap->ref_src[ref_count], &vp->ref_slots[ref_count],
468 + &vp->refs[ref_count], &ap->std_refs[ref_count], &ap->vkav1_refs[ref_count],
469 + ref_frame, 0, 0, cbs_ctx->ref[idx].saved_order_hints);
473 @@ -249,20 +302,32 @@ static int vk_av1_start_frame(AVCodecContext *avctx,
476 err = vk_av1_fill_pict(avctx, NULL, &vp->ref_slot, &vp->ref,
479 - pic, 1, apply_grain, 8);
480 + pic, 1, apply_grain, NULL);
484 - ap->tile_list.nb_tiles = 0;
485 - ap->tile_list.tile_list = ap->tiles;
487 - ap->av1_pic_info = (VkVideoDecodeAV1PictureInfoMESA) {
488 - .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_PICTURE_INFO_MESA,
489 - .frame_header = &ap->av1_frame_header,
490 - .tile_list = &ap->tile_list,
491 + ap->av1_pic_info = (VkVideoDecodeAV1PictureInfoKHR) {
492 + .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_PICTURE_INFO_KHR,
493 + .pStdPictureInfo = &ap->std_pic_info,
494 + .frameHeaderOffset = 0,
496 + .pTileOffsets = NULL,
497 + .pTileSizes = ap->tile_sizes,
500 + for (int i = 0; i < STD_VIDEO_AV1_REFS_PER_FRAME; i++) {
501 + const int idx = pic->raw_frame_header->ref_frame_idx[i];
502 + const AV1Frame *ref_frame = &s->ref[idx];
503 + AV1VulkanDecodePicture *hp = ref_frame->hwaccel_picture_private;
505 + if (ref_frame->f->pict_type == AV_PICTURE_TYPE_NONE)
506 + ap->av1_pic_info.referenceNameSlotIndices[i] = AV1_REF_FRAME_NONE;
508 + ap->av1_pic_info.referenceNameSlotIndices[i] = hp->frame_id;
511 vp->decode_info = (VkVideoDecodeInfoKHR) {
512 .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_INFO_KHR,
513 .pNext = &ap->av1_pic_info,
514 @@ -279,9 +344,87 @@ static int vk_av1_start_frame(AVCodecContext *avctx,
518 + ap->tile_info = (StdVideoAV1TileInfo) {
519 + .flags = (StdVideoAV1TileInfoFlags) {
520 + .uniform_tile_spacing_flag = frame_header->uniform_tile_spacing_flag,
522 + .TileCols = frame_header->tile_cols,
523 + .TileRows = frame_header->tile_rows,
524 + .context_update_tile_id = frame_header->context_update_tile_id,
525 + .tile_size_bytes_minus_1 = frame_header->tile_size_bytes_minus1,
526 + .pWidthInSbsMinus1 = ap->width_in_sbs_minus1,
527 + .pHeightInSbsMinus1 = ap->height_in_sbs_minus1,
528 + .pMiColStarts = ap->mi_col_starts,
529 + .pMiRowStarts = ap->mi_row_starts,
532 + ap->quantization = (StdVideoAV1Quantization) {
533 + .flags.using_qmatrix = frame_header->using_qmatrix,
534 + .flags.diff_uv_delta = frame_header->diff_uv_delta,
535 + .base_q_idx = frame_header->base_q_idx,
536 + .DeltaQYDc = frame_header->delta_q_y_dc,
537 + .DeltaQUDc = frame_header->delta_q_u_dc,
538 + .DeltaQUAc = frame_header->delta_q_u_ac,
539 + .DeltaQVDc = frame_header->delta_q_v_dc,
540 + .DeltaQVAc = frame_header->delta_q_v_ac,
541 + .qm_y = frame_header->qm_y,
542 + .qm_u = frame_header->qm_u,
543 + .qm_v = frame_header->qm_v,
546 + ap->loop_filter = (StdVideoAV1LoopFilter) {
547 + .flags = (StdVideoAV1LoopFilterFlags) {
548 + .loop_filter_delta_enabled = frame_header->loop_filter_delta_enabled,
549 + .loop_filter_delta_update = frame_header->loop_filter_delta_update,
551 + .loop_filter_sharpness = frame_header->loop_filter_sharpness,
554 + for (int i = 0; i < STD_VIDEO_AV1_MAX_LOOP_FILTER_STRENGTHS; i++)
555 + ap->loop_filter.loop_filter_level[i] = frame_header->loop_filter_level[i];
556 + for (int i = 0; i < STD_VIDEO_AV1_LOOP_FILTER_ADJUSTMENTS; i++)
557 + ap->loop_filter.loop_filter_mode_deltas[i] = frame_header->loop_filter_mode_deltas[i];
559 + ap->cdef = (StdVideoAV1CDEF) {
560 + .cdef_damping_minus_3 = frame_header->cdef_damping_minus_3,
561 + .cdef_bits = frame_header->cdef_bits,
564 + ap->loop_restoration = (StdVideoAV1LoopRestoration) {
565 + .FrameRestorationType[0] = remap_lr_type[frame_header->lr_type[0]],
566 + .FrameRestorationType[1] = remap_lr_type[frame_header->lr_type[1]],
567 + .FrameRestorationType[2] = remap_lr_type[frame_header->lr_type[2]],
568 + .LoopRestorationSize[0] = 1 + frame_header->lr_unit_shift,
569 + .LoopRestorationSize[1] = 1 + frame_header->lr_unit_shift - frame_header->lr_uv_shift,
570 + .LoopRestorationSize[2] = 1 + frame_header->lr_unit_shift - frame_header->lr_uv_shift,
573 + ap->film_grain = (StdVideoAV1FilmGrain) {
574 + .flags = (StdVideoAV1FilmGrainFlags) {
575 + .chroma_scaling_from_luma = film_grain->chroma_scaling_from_luma,
576 + .overlap_flag = film_grain->overlap_flag,
577 + .clip_to_restricted_range = film_grain->clip_to_restricted_range,
579 + .grain_scaling_minus_8 = film_grain->grain_scaling_minus_8,
580 + .ar_coeff_lag = film_grain->ar_coeff_lag,
581 + .ar_coeff_shift_minus_6 = film_grain->ar_coeff_shift_minus_6,
582 + .grain_scale_shift = film_grain->grain_scale_shift,
583 + .grain_seed = film_grain->grain_seed,
584 + .film_grain_params_ref_idx = film_grain->film_grain_params_ref_idx,
585 + .num_y_points = film_grain->num_y_points,
586 + .num_cb_points = film_grain->num_cb_points,
587 + .num_cr_points = film_grain->num_cr_points,
588 + .cb_mult = film_grain->cb_mult,
589 + .cb_luma_mult = film_grain->cb_luma_mult,
590 + .cb_offset = film_grain->cb_offset,
591 + .cr_mult = film_grain->cr_mult,
592 + .cr_luma_mult = film_grain->cr_luma_mult,
593 + .cr_offset = film_grain->cr_offset,
596 /* Setup frame header */
597 - ap->av1_frame_header = (StdVideoAV1MESAFrameHeader) {
598 - .flags = (StdVideoAV1MESAFrameHeaderFlags) {
599 + ap->std_pic_info = (StdVideoDecodeAV1PictureInfo) {
600 + .flags = (StdVideoDecodeAV1PictureInfoFlags) {
601 .error_resilient_mode = frame_header->error_resilient_mode,
602 .disable_cdf_update = frame_header->disable_cdf_update,
603 .use_superres = frame_header->use_superres,
604 @@ -302,174 +445,92 @@ static int vk_av1_start_frame(AVCodecContext *avctx,
605 .reference_select = frame_header->reference_select,
606 .skip_mode_present = frame_header->skip_mode_present,
607 .delta_q_present = frame_header->delta_q_present,
608 + .delta_lf_present = frame_header->delta_lf_present,
609 + .delta_lf_multi = frame_header->delta_lf_multi,
610 + .segmentation_enabled = frame_header->segmentation_enabled,
611 + .segmentation_update_map = frame_header->segmentation_update_map,
612 + .segmentation_temporal_update = frame_header->segmentation_temporal_update,
613 + .segmentation_update_data = frame_header->segmentation_update_data,
614 + .UsesLr = frame_header->lr_type[0] || frame_header->lr_type[1] || frame_header->lr_type[2],
615 + .apply_grain = apply_grain,
617 - .frame_to_show_map_idx = frame_header->frame_to_show_map_idx,
618 - .frame_presentation_time = frame_header->frame_presentation_time,
619 - .display_frame_id = frame_header->display_frame_id,
620 .frame_type = frame_header->frame_type,
621 .current_frame_id = frame_header->current_frame_id,
622 - .order_hint = frame_header->order_hint,
623 + .OrderHint = frame_header->order_hint,
624 .primary_ref_frame = frame_header->primary_ref_frame,
625 - .frame_width_minus_1 = frame_header->frame_width_minus_1,
626 - .frame_height_minus_1 = frame_header->frame_height_minus_1,
627 - .coded_denom = frame_header->coded_denom,
628 - .render_width_minus_1 = frame_header->render_width_minus_1,
629 - .render_height_minus_1 = frame_header->render_height_minus_1,
630 .refresh_frame_flags = frame_header->refresh_frame_flags,
631 .interpolation_filter = frame_header->interpolation_filter,
632 - .tx_mode = frame_header->tx_mode,
633 - .tiling = (StdVideoAV1MESATileInfo) {
634 - .flags = (StdVideoAV1MESATileInfoFlags) {
635 - .uniform_tile_spacing_flag = frame_header->uniform_tile_spacing_flag,
637 - .tile_cols = frame_header->tile_cols,
638 - .tile_rows = frame_header->tile_rows,
639 - .context_update_tile_id = frame_header->context_update_tile_id,
640 - .tile_size_bytes_minus1 = frame_header->tile_size_bytes_minus1,
642 - .quantization = (StdVideoAV1MESAQuantization) {
643 - .flags.using_qmatrix = frame_header->using_qmatrix,
644 - .base_q_idx = frame_header->base_q_idx,
645 - .delta_q_y_dc = frame_header->delta_q_y_dc,
646 - .diff_uv_delta = frame_header->diff_uv_delta,
647 - .delta_q_u_dc = frame_header->delta_q_u_dc,
648 - .delta_q_u_ac = frame_header->delta_q_u_ac,
649 - .delta_q_v_dc = frame_header->delta_q_v_dc,
650 - .delta_q_v_ac = frame_header->delta_q_v_ac,
651 - .qm_y = frame_header->qm_y,
652 - .qm_u = frame_header->qm_u,
653 - .qm_v = frame_header->qm_v,
655 - .delta_q = (StdVideoAV1MESADeltaQ) {
656 - .flags = (StdVideoAV1MESADeltaQFlags) {
657 - .delta_lf_present = frame_header->delta_lf_present,
658 - .delta_lf_multi = frame_header->delta_lf_multi,
660 - .delta_q_res = frame_header->delta_q_res,
661 - .delta_lf_res = frame_header->delta_lf_res,
663 - .loop_filter = (StdVideoAV1MESALoopFilter) {
664 - .flags = (StdVideoAV1MESALoopFilterFlags) {
665 - .delta_enabled = frame_header->loop_filter_delta_enabled,
666 - .delta_update = frame_header->loop_filter_delta_update,
669 - frame_header->loop_filter_level[0], frame_header->loop_filter_level[1],
670 - frame_header->loop_filter_level[2], frame_header->loop_filter_level[3],
672 - .sharpness = frame_header->loop_filter_sharpness,
674 - frame_header->loop_filter_mode_deltas[0], frame_header->loop_filter_mode_deltas[1],
677 - .cdef = (StdVideoAV1MESACDEF) {
678 - .damping_minus_3 = frame_header->cdef_damping_minus_3,
679 - .bits = frame_header->cdef_bits,
681 - .lr = (StdVideoAV1MESALoopRestoration) {
682 - .lr_unit_shift = frame_header->lr_unit_shift,
683 - .lr_uv_shift = frame_header->lr_uv_shift,
684 - .lr_type = { frame_header->lr_type[0], frame_header->lr_type[1], frame_header->lr_type[2] },
686 - .segmentation = (StdVideoAV1MESASegmentation) {
687 - .flags = (StdVideoAV1MESASegmentationFlags) {
688 - .enabled = frame_header->segmentation_enabled,
689 - .update_map = frame_header->segmentation_update_map,
690 - .temporal_update = frame_header->segmentation_temporal_update,
691 - .update_data = frame_header->segmentation_update_data,
694 - .film_grain = (StdVideoAV1MESAFilmGrainParameters) {
695 - .flags = (StdVideoAV1MESAFilmGrainFlags) {
696 - .apply_grain = apply_grain,
697 - .chroma_scaling_from_luma = film_grain->chroma_scaling_from_luma,
698 - .overlap_flag = film_grain->overlap_flag,
699 - .clip_to_restricted_range = film_grain->clip_to_restricted_range,
701 - .grain_scaling_minus_8 = film_grain->grain_scaling_minus_8,
702 - .ar_coeff_lag = film_grain->ar_coeff_lag,
703 - .ar_coeff_shift_minus_6 = film_grain->ar_coeff_shift_minus_6,
704 - .grain_scale_shift = film_grain->grain_scale_shift,
705 - .grain_seed = film_grain->grain_seed,
706 - .num_y_points = film_grain->num_y_points,
707 - .num_cb_points = film_grain->num_cb_points,
708 - .num_cr_points = film_grain->num_cr_points,
709 - .cb_mult = film_grain->cb_mult,
710 - .cb_luma_mult = film_grain->cb_luma_mult,
711 - .cb_offset = film_grain->cb_offset,
712 - .cr_mult = film_grain->cr_mult,
713 - .cr_luma_mult = film_grain->cr_luma_mult,
714 - .cr_offset = film_grain->cr_offset,
716 + .TxMode = frame_header->tx_mode,
717 + .delta_q_res = frame_header->delta_q_res,
718 + .delta_lf_res = frame_header->delta_lf_res,
719 + .SkipModeFrame[0] = s->cur_frame.skip_mode_frame_idx[0],
720 + .SkipModeFrame[1] = s->cur_frame.skip_mode_frame_idx[1],
721 + .coded_denom = frame_header->coded_denom,
722 + .pTileInfo = &ap->tile_info,
723 + .pQuantization = &ap->quantization,
724 + .pSegmentation = &ap->segmentation,
725 + .pLoopFilter = &ap->loop_filter,
726 + .pCDEF = &ap->cdef,
727 + .pLoopRestoration = &ap->loop_restoration,
728 + .pGlobalMotion = &ap->global_motion,
729 + .pFilmGrain = apply_grain ? &ap->film_grain : NULL,
732 for (int i = 0; i < 64; i++) {
733 - ap->av1_frame_header.tiling.width_in_sbs_minus_1[i] = frame_header->width_in_sbs_minus_1[i];
734 - ap->av1_frame_header.tiling.height_in_sbs_minus_1[i] = frame_header->height_in_sbs_minus_1[i];
735 - ap->av1_frame_header.tiling.tile_start_col_sb[i] = frame_header->tile_start_col_sb[i];
736 - ap->av1_frame_header.tiling.tile_start_row_sb[i] = frame_header->tile_start_row_sb[i];
737 + ap->width_in_sbs_minus1[i] = frame_header->width_in_sbs_minus_1[i];
738 + ap->height_in_sbs_minus1[i] = frame_header->height_in_sbs_minus_1[i];
739 + ap->mi_col_starts[i] = frame_header->tile_start_col_sb[i];
740 + ap->mi_row_starts[i] = frame_header->tile_start_row_sb[i];
743 - for (int i = 0; i < 8; i++) {
744 - ap->av1_frame_header.segmentation.feature_enabled_bits[i] = 0;
745 - for (int j = 0; j < 8; j++) {
746 - ap->av1_frame_header.segmentation.feature_enabled_bits[i] |= (frame_header->feature_enabled[i][j] << j);
747 - ap->av1_frame_header.segmentation.feature_data[i][j] = frame_header->feature_value[i][j];
748 + for (int i = 0; i < STD_VIDEO_AV1_MAX_SEGMENTS; i++) {
749 + ap->segmentation.FeatureEnabled[i] = 0x0;
750 + for (int j = 0; j < STD_VIDEO_AV1_SEG_LVL_MAX; j++) {
751 + ap->segmentation.FeatureEnabled[i] |= (frame_header->feature_enabled[i][j] << j);
752 + ap->segmentation.FeatureData[i][j] = frame_header->feature_value[i][j];
755 - ap->av1_frame_header.loop_filter.ref_deltas[i] = frame_header->loop_filter_ref_deltas[i];
757 - ap->av1_frame_header.cdef.y_pri_strength[i] = frame_header->cdef_y_pri_strength[i];
758 - ap->av1_frame_header.cdef.y_sec_strength[i] = frame_header->cdef_y_sec_strength[i];
759 - ap->av1_frame_header.cdef.uv_pri_strength[i] = frame_header->cdef_uv_pri_strength[i];
760 - ap->av1_frame_header.cdef.uv_sec_strength[i] = frame_header->cdef_uv_sec_strength[i];
762 - ap->av1_frame_header.ref_order_hint[i] = frame_header->ref_order_hint[i];
763 - ap->av1_frame_header.global_motion[i] = (StdVideoAV1MESAGlobalMotion) {
764 - .flags = (StdVideoAV1MESAGlobalMotionFlags) {
765 - .gm_invalid = s->cur_frame.gm_invalid[i],
767 - .gm_type = s->cur_frame.gm_type[i],
769 - s->cur_frame.gm_params[i][0], s->cur_frame.gm_params[i][1],
770 - s->cur_frame.gm_params[i][2], s->cur_frame.gm_params[i][3],
771 - s->cur_frame.gm_params[i][4], s->cur_frame.gm_params[i][5],
776 - for (int i = 0; i < 7; i++) {
777 - ap->av1_frame_header.ref_frame_idx[i] = frame_header->ref_frame_idx[i];
778 - ap->av1_frame_header.delta_frame_id_minus1[i] = frame_header->delta_frame_id_minus1[i];
779 + for (int i = 0; i < STD_VIDEO_AV1_TOTAL_REFS_PER_FRAME; i++)
780 + ap->loop_filter.loop_filter_ref_deltas[i] = frame_header->loop_filter_ref_deltas[i];
782 + for (int i = 0; i < STD_VIDEO_AV1_MAX_CDEF_FILTER_STRENGTHS; i++) {
783 + ap->cdef.cdef_y_pri_strength[i] = frame_header->cdef_y_pri_strength[i];
784 + ap->cdef.cdef_y_sec_strength[i] = frame_header->cdef_y_sec_strength[i];
785 + ap->cdef.cdef_uv_pri_strength[i] = frame_header->cdef_uv_pri_strength[i];
786 + ap->cdef.cdef_uv_sec_strength[i] = frame_header->cdef_uv_sec_strength[i];
789 - ap->av1_pic_info.skip_mode_frame_idx[0] = s->cur_frame.skip_mode_frame_idx[0];
790 - ap->av1_pic_info.skip_mode_frame_idx[1] = s->cur_frame.skip_mode_frame_idx[1];
791 + for (int i = 0; i < STD_VIDEO_AV1_NUM_REF_FRAMES; i++) {
792 + ap->std_pic_info.OrderHints[i] = frame_header->ref_order_hint[i];
793 + ap->global_motion.GmType[i] = s->cur_frame.gm_type[i];
794 + for (int j = 0; j < STD_VIDEO_AV1_GLOBAL_MOTION_PARAMS; j++) {
795 + ap->global_motion.gm_params[i][j] = s->cur_frame.gm_params[i][j];
800 - for (int i = 0; i < 14; i++) {
801 - ap->av1_frame_header.film_grain.point_y_value[i] = film_grain->point_y_value[i];
802 - ap->av1_frame_header.film_grain.point_y_scaling[i] = film_grain->point_y_scaling[i];
803 + for (int i = 0; i < STD_VIDEO_AV1_MAX_NUM_Y_POINTS; i++) {
804 + ap->film_grain.point_y_value[i] = film_grain->point_y_value[i];
805 + ap->film_grain.point_y_scaling[i] = film_grain->point_y_scaling[i];
808 - for (int i = 0; i < 10; i++) {
809 - ap->av1_frame_header.film_grain.point_cb_value[i] = film_grain->point_cb_value[i];
810 - ap->av1_frame_header.film_grain.point_cb_scaling[i] = film_grain->point_cb_scaling[i];
811 - ap->av1_frame_header.film_grain.point_cr_value[i] = film_grain->point_cr_value[i];
812 - ap->av1_frame_header.film_grain.point_cr_scaling[i] = film_grain->point_cr_scaling[i];
813 + for (int i = 0; i < STD_VIDEO_AV1_MAX_NUM_CB_POINTS; i++) {
814 + ap->film_grain.point_cb_value[i] = film_grain->point_cb_value[i];
815 + ap->film_grain.point_cb_scaling[i] = film_grain->point_cb_scaling[i];
816 + ap->film_grain.point_cr_value[i] = film_grain->point_cr_value[i];
817 + ap->film_grain.point_cr_scaling[i] = film_grain->point_cr_scaling[i];
820 - for (int i = 0; i < 24; i++) {
821 - ap->av1_frame_header.film_grain.ar_coeffs_y_plus_128[i] = film_grain->ar_coeffs_y_plus_128[i];
822 - ap->av1_frame_header.film_grain.ar_coeffs_cb_plus_128[i] = film_grain->ar_coeffs_cb_plus_128[i];
823 - ap->av1_frame_header.film_grain.ar_coeffs_cr_plus_128[i] = film_grain->ar_coeffs_cr_plus_128[i];
825 + for (int i = 0; i < STD_VIDEO_AV1_MAX_NUM_POS_LUMA; i++)
826 + ap->film_grain.ar_coeffs_y_plus_128[i] = film_grain->ar_coeffs_y_plus_128[i];
828 - ap->av1_frame_header.film_grain.ar_coeffs_cb_plus_128[24] = film_grain->ar_coeffs_cb_plus_128[24];
829 - ap->av1_frame_header.film_grain.ar_coeffs_cr_plus_128[24] = film_grain->ar_coeffs_cr_plus_128[24];
830 + for (int i = 0; i < STD_VIDEO_AV1_MAX_NUM_POS_CHROMA; i++) {
831 + ap->film_grain.ar_coeffs_cb_plus_128[i] = film_grain->ar_coeffs_cb_plus_128[i];
832 + ap->film_grain.ar_coeffs_cr_plus_128[i] = film_grain->ar_coeffs_cr_plus_128[i];
836 - /* Workaround for a spec issue. */
840 @@ -484,25 +545,20 @@ static int vk_av1_decode_slice(AVCodecContext *avctx,
841 AV1VulkanDecodePicture *ap = s->cur_frame.hwaccel_picture_private;
842 FFVulkanDecodePicture *vp = &ap->vp;
844 + /* Too many tiles, exceeding all defined levels in the AV1 spec */
845 + if (ap->av1_pic_info.tileCount > MAX_TILES)
846 + return AVERROR(ENOSYS);
848 for (int i = s->tg_start; i <= s->tg_end; i++) {
849 - ap->tiles[ap->tile_list.nb_tiles] = (StdVideoAV1MESATile) {
850 - .size = s->tile_group_info[i].tile_size,
851 - .offset = s->tile_group_info[i].tile_offset,
852 - .row = s->tile_group_info[i].tile_row,
853 - .column = s->tile_group_info[i].tile_column,
854 - .tg_start = s->tg_start,
855 - .tg_end = s->tg_end,
857 + ap->tile_sizes[ap->av1_pic_info.tileCount] = s->tile_group_info[i].tile_size;
859 err = ff_vk_decode_add_slice(avctx, vp,
860 data + s->tile_group_info[i].tile_offset,
861 s->tile_group_info[i].tile_size, 0,
862 - &ap->tile_list.nb_tiles,
863 - &ap->tile_offsets);
864 + &ap->av1_pic_info.tileCount,
865 + &ap->av1_pic_info.pTileOffsets);
869 - ap->tiles[ap->tile_list.nb_tiles - 1].offset = ap->tile_offsets[ap->tile_list.nb_tiles - 1];
873 @@ -518,7 +574,7 @@ static int vk_av1_end_frame(AVCodecContext *avctx)
874 FFVulkanDecodePicture *rvp[AV1_NUM_REF_FRAMES] = { 0 };
875 AVFrame *rav[AV1_NUM_REF_FRAMES] = { 0 };
877 - if (!ap->tile_list.nb_tiles)
878 + if (!ap->av1_pic_info.tileCount)
881 if (!dec->session_params) {
882 @@ -536,7 +592,7 @@ static int vk_av1_end_frame(AVCodecContext *avctx)
885 av_log(avctx, AV_LOG_VERBOSE, "Decoding frame, %"SIZE_SPECIFIER" bytes, %i tiles\n",
886 - vp->slices_size, ap->tile_list.nb_tiles);
887 + vp->slices_size, ap->av1_pic_info.tileCount);
889 return ff_vk_decode_frame(avctx, pic->f, vp, rav, rvp);
891 @@ -580,8 +636,6 @@ const FFHWAccel ff_av1_vulkan_hwaccel = {
892 * flexibility, this index cannot be present anywhere.
893 * The current implementation tracks the index for the driver and submits it
894 * as necessary information. Due to needing to modify the decoding context,
895 - * which is not thread-safe, on frame free, threading is disabled.
896 - * In the future, once this is fixed in the spec, the workarounds may be removed
897 - * and threading enabled. */
898 + * which is not thread-safe, on frame free, threading is disabled. */
899 .caps_internal = HWACCEL_CAP_ASYNC_SAFE,
901 --- ffmpeg-6.1.1/libavcodec/vulkan_decode.c.orig 2023-11-11 01:25:17.000000000 +0100
902 +++ ffmpeg-6.1.1/libavcodec/vulkan_decode.c 2024-04-15 21:49:59.451325544 +0200
903 @@ -50,7 +50,7 @@ static const VkVideoProfileInfoKHR *get_
904 VkStructureType profile_struct_type =
905 codec_id == AV_CODEC_ID_H264 ? VK_STRUCTURE_TYPE_VIDEO_DECODE_H264_PROFILE_INFO_KHR :
906 codec_id == AV_CODEC_ID_HEVC ? VK_STRUCTURE_TYPE_VIDEO_DECODE_H265_PROFILE_INFO_KHR :
907 - codec_id == AV_CODEC_ID_AV1 ? VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_PROFILE_INFO_MESA :
908 + codec_id == AV_CODEC_ID_AV1 ? VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_PROFILE_INFO_KHR :
911 profile_list = ff_vk_find_struct(ctx->s.hwfc->create_pnext,
912 @@ -663,7 +663,7 @@ static VkResult vulkan_setup_profile(AVC
913 const struct FFVkCodecMap *vk_codec,
914 VkVideoDecodeH264CapabilitiesKHR *h264_caps,
915 VkVideoDecodeH265CapabilitiesKHR *h265_caps,
916 - VkVideoDecodeAV1CapabilitiesMESA *av1_caps,
917 + VkVideoDecodeAV1CapabilitiesKHR *av1_caps,
918 VkVideoCapabilitiesKHR *caps,
919 VkVideoDecodeCapabilitiesKHR *dec_caps,
921 @@ -674,7 +674,7 @@ static VkResult vulkan_setup_profile(AVC
923 VkVideoDecodeH264ProfileInfoKHR *h264_profile = &prof->h264_profile;
924 VkVideoDecodeH265ProfileInfoKHR *h265_profile = &prof->h265_profile;
925 - VkVideoDecodeAV1ProfileInfoMESA *av1_profile = &prof->av1_profile;
926 + VkVideoDecodeAV1ProfileInfoKHR *av1_profile = &prof->av1_profile;
928 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(avctx->sw_pix_fmt);
930 @@ -702,8 +702,9 @@ static VkResult vulkan_setup_profile(AVC
931 } else if (avctx->codec_id == AV_CODEC_ID_AV1) {
932 dec_caps->pNext = av1_caps;
933 usage->pNext = av1_profile;
934 - av1_profile->sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_PROFILE_INFO_MESA;
935 - av1_profile->stdProfileIdc = cur_profile;
936 + av1_profile->sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_PROFILE_INFO_KHR;
937 + av1_profile->stdProfile = cur_profile;
938 + av1_profile->filmGrainSupport = !(avctx->export_side_data & AV_CODEC_EXPORT_DATA_FILM_GRAIN);
941 usage->sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_USAGE_INFO_KHR;
942 @@ -758,8 +759,8 @@ static int vulkan_decode_get_profile(AVC
943 VkVideoDecodeH265CapabilitiesKHR h265_caps = {
944 .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_H265_CAPABILITIES_KHR,
946 - VkVideoDecodeAV1CapabilitiesMESA av1_caps = {
947 - .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_CAPABILITIES_MESA,
948 + VkVideoDecodeAV1CapabilitiesKHR av1_caps = {
949 + .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_CAPABILITIES_KHR,
952 VkPhysicalDeviceVideoFormatInfoKHR fmt_info = {
953 @@ -782,7 +783,7 @@ static int vulkan_decode_get_profile(AVC
954 cur_profile = avctx->profile;
955 base_profile = avctx->codec_id == AV_CODEC_ID_H264 ? AV_PROFILE_H264_CONSTRAINED_BASELINE :
956 avctx->codec_id == AV_CODEC_ID_H265 ? AV_PROFILE_HEVC_MAIN :
957 - avctx->codec_id == AV_CODEC_ID_AV1 ? STD_VIDEO_AV1_MESA_PROFILE_MAIN :
958 + avctx->codec_id == AV_CODEC_ID_AV1 ? STD_VIDEO_AV1_PROFILE_MAIN :
961 ret = vulkan_setup_profile(avctx, prof, hwctx, vk, vk_codec,
962 @@ -830,7 +831,7 @@ static int vulkan_decode_get_profile(AVC
964 max_level = avctx->codec_id == AV_CODEC_ID_H264 ? ff_vk_h264_level_to_av(h264_caps.maxLevelIdc) :
965 avctx->codec_id == AV_CODEC_ID_H265 ? ff_vk_h265_level_to_av(h265_caps.maxLevelIdc) :
966 - avctx->codec_id == AV_CODEC_ID_AV1 ? av1_caps.maxLevelIdc :
967 + avctx->codec_id == AV_CODEC_ID_AV1 ? av1_caps.maxLevel :
970 av_log(avctx, AV_LOG_VERBOSE, "Decoder capabilities for %s profile \"%s\":\n",
971 @@ -901,17 +902,11 @@ static int vulkan_decode_get_profile(AVC
972 "VK_VIDEO_DECODE_CAPABILITY_DPB_AND_OUTPUT_COINCIDE_BIT_KHR set "
973 "but VK_VIDEO_CAPABILITY_SEPARATE_REFERENCE_IMAGES_BIT_KHR is unset!\n");
974 return AVERROR_EXTERNAL;
975 - } else if (!(dec_caps->flags & VK_VIDEO_DECODE_CAPABILITY_DPB_AND_OUTPUT_DISTINCT_BIT_KHR) &&
976 - avctx->codec_id == AV_CODEC_ID_AV1) {
977 - av_log(avctx, AV_LOG_ERROR, "Cannot initialize Vulkan decoding session, buggy driver: "
978 - "codec is AV1, but VK_VIDEO_DECODE_CAPABILITY_DPB_AND_OUTPUT_DISTINCT_BIT_KHR isn't set!\n");
979 - return AVERROR_EXTERNAL;
982 /* TODO: make dedicated_dpb tunable */
983 dec->dedicated_dpb = !(dec_caps->flags & VK_VIDEO_DECODE_CAPABILITY_DPB_AND_OUTPUT_COINCIDE_BIT_KHR);
984 dec->layered_dpb = !(caps->flags & VK_VIDEO_CAPABILITY_SEPARATE_REFERENCE_IMAGES_BIT_KHR);
985 - dec->external_fg = av1_caps.flags & VK_VIDEO_DECODE_AV1_CAPABILITY_EXTERNAL_FILM_GRAIN_MESA;
987 if (dec->dedicated_dpb) {
988 fmt_info.imageUsage = VK_IMAGE_USAGE_VIDEO_DECODE_DPB_BIT_KHR;
989 @@ -1118,8 +1113,10 @@ int ff_vk_decode_init(AVCodecContext *av
990 VkVideoDecodeH265SessionParametersCreateInfoKHR h265_params = {
991 .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_H265_SESSION_PARAMETERS_CREATE_INFO_KHR,
993 - VkVideoDecodeAV1SessionParametersCreateInfoMESA av1_params = {
994 - .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_SESSION_PARAMETERS_CREATE_INFO_MESA,
995 + StdVideoAV1SequenceHeader av1_empty_seq = { 0 };
996 + VkVideoDecodeAV1SessionParametersCreateInfoKHR av1_params = {
997 + .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_SESSION_PARAMETERS_CREATE_INFO_KHR,
998 + .pStdSequenceHeader = &av1_empty_seq,
1000 VkVideoSessionParametersCreateInfoKHR session_params_create = {
1001 .sType = VK_STRUCTURE_TYPE_VIDEO_SESSION_PARAMETERS_CREATE_INFO_KHR,
1002 diff --git a/libavcodec/vulkan_decode.h b/libavcodec/vulkan_decode.h
1003 index a43e328d73..7ba8b239cb 100644
1004 --- a/libavcodec/vulkan_decode.h
1005 +++ b/libavcodec/vulkan_decode.h
1006 @@ -37,7 +37,7 @@ typedef struct FFVulkanDecodeDescriptor {
1007 typedef struct FFVulkanDecodeProfileData {
1008 VkVideoDecodeH264ProfileInfoKHR h264_profile;
1009 VkVideoDecodeH265ProfileInfoKHR h265_profile;
1010 - VkVideoDecodeAV1ProfileInfoMESA av1_profile;
1011 + VkVideoDecodeAV1ProfileInfoKHR av1_profile;
1012 VkVideoDecodeUsageInfoKHR usage;
1013 VkVideoProfileInfoKHR profile;
1014 VkVideoProfileListInfoKHR profile_list;
1015 --- ffmpeg-6.1.1/libavcodec/vulkan_video.h.orig 2024-04-15 21:35:22.599409192 +0200
1016 +++ ffmpeg-6.1.1/libavcodec/vulkan_video.h 2024-04-15 21:50:11.117929007 +0200
1020 #include <vk_video/vulkan_video_codecs_common.h>
1021 -#include "vulkan_video_codec_av1std.h"
1022 -#include "vulkan_video_codec_av1std_decode.h"
1024 #define CODEC_VER_MAJ(ver) (ver >> 22)
1025 #define CODEC_VER_MIN(ver) ((ver >> 12) & ((1 << 10) - 1))
1026 diff --git a/libavcodec/vulkan_video_codec_av1std_decode.h b/libavcodec/vulkan_video_codec_av1std_decode.h
1027 deleted file mode 100644
1028 index e2f37b4e6e..0000000000
1029 --- a/libavcodec/vulkan_video_codec_av1std_decode.h
1032 -/* Copyright 2023 Lynne
1033 - * Copyright 2023 Dave Airlie
1035 - * Licensed under the Apache License, Version 2.0 (the "License");
1036 - * you may not use this file except in compliance with the License.
1037 - * You may obtain a copy of the License at
1039 - * http://www.apache.org/licenses/LICENSE-2.0
1041 - * Unless required by applicable law or agreed to in writing, software
1042 - * distributed under the License is distributed on an "AS IS" BASIS,
1043 - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
1044 - * See the License for the specific language governing permissions and
1045 - * limitations under the License.
1048 -#ifndef VULKAN_VIDEO_CODEC_AV1STD_DECODE_H_
1049 -#define VULKAN_VIDEO_CODEC_AV1STD_DECODE_H_ 1
1052 -** This header is NOT YET generated from the Khronos Vulkan XML API Registry.
1059 -#define vulkan_video_codec_av1std_decode 1
1068 diff --git a/libavcodec/vulkan_video_codec_av1std.h b/libavcodec/vulkan_video_codec_av1std.h
1069 deleted file mode 100644
1070 index c91589eee2..0000000000
1071 --- a/libavcodec/vulkan_video_codec_av1std.h
1074 -/* Copyright 2023 Lynne
1075 - * Copyright 2023 Dave Airlie
1077 - * Licensed under the Apache License, Version 2.0 (the "License");
1078 - * you may not use this file except in compliance with the License.
1079 - * You may obtain a copy of the License at
1081 - * http://www.apache.org/licenses/LICENSE-2.0
1083 - * Unless required by applicable law or agreed to in writing, software
1084 - * distributed under the License is distributed on an "AS IS" BASIS,
1085 - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
1086 - * See the License for the specific language governing permissions and
1087 - * limitations under the License.
1090 -#ifndef VULKAN_VIDEO_CODEC_AV1STD_H_
1091 -#define VULKAN_VIDEO_CODEC_AV1STD_H_ 1
1094 -** This header is NOT YET generated from the Khronos Vulkan XML API Registry.
1101 -#define vulkan_video_codec_av1std 1
1103 -#define VK_MAKE_VIDEO_STD_VERSION(major, minor, patch) \
1104 - ((((uint32_t)(major)) << 22) | (((uint32_t)(minor)) << 12) | ((uint32_t)(patch)))
1105 -#define VK_STD_VULKAN_VIDEO_CODEC_AV1_DECODE_API_VERSION_0_1_0 VK_MAKE_VIDEO_STD_VERSION(0, 1, 0)
1106 -#define VK_STD_VULKAN_VIDEO_CODEC_AV1_DECODE_SPEC_VERSION VK_STD_VULKAN_VIDEO_CODEC_AV1_DECODE_API_VERSION_0_1_0
1107 -#define VK_STD_VULKAN_VIDEO_CODEC_AV1_DECODE_EXTENSION_NAME "VK_STD_vulkan_video_codec_av1_decode"
1109 -typedef enum StdVideoAV1MESAProfile {
1110 - STD_VIDEO_AV1_MESA_PROFILE_MAIN = 0,
1111 - STD_VIDEO_AV1_MESA_PROFILE_HIGH = 1,
1112 - STD_VIDEO_AV1_MESA_PROFILE_PROFESSIONAL = 2,
1113 -} StdVideoAV1MESAProfile;
1115 -typedef enum StdVideoAV1MESALevel {
1116 - STD_VIDEO_AV1_MESA_LEVEL_2_0 = 0,
1117 - STD_VIDEO_AV1_MESA_LEVEL_2_1 = 1,
1118 - STD_VIDEO_AV1_MESA_LEVEL_2_2 = 2,
1119 - STD_VIDEO_AV1_MESA_LEVEL_2_3 = 3,
1120 - STD_VIDEO_AV1_MESA_LEVEL_3_0 = 4,
1121 - STD_VIDEO_AV1_MESA_LEVEL_3_1 = 5,
1122 - STD_VIDEO_AV1_MESA_LEVEL_3_2 = 6,
1123 - STD_VIDEO_AV1_MESA_LEVEL_3_3 = 7,
1124 - STD_VIDEO_AV1_MESA_LEVEL_4_0 = 8,
1125 - STD_VIDEO_AV1_MESA_LEVEL_4_1 = 9,
1126 - STD_VIDEO_AV1_MESA_LEVEL_4_2 = 10,
1127 - STD_VIDEO_AV1_MESA_LEVEL_4_3 = 11,
1128 - STD_VIDEO_AV1_MESA_LEVEL_5_0 = 12,
1129 - STD_VIDEO_AV1_MESA_LEVEL_5_1 = 13,
1130 - STD_VIDEO_AV1_MESA_LEVEL_5_2 = 14,
1131 - STD_VIDEO_AV1_MESA_LEVEL_5_3 = 15,
1132 - STD_VIDEO_AV1_MESA_LEVEL_6_0 = 16,
1133 - STD_VIDEO_AV1_MESA_LEVEL_6_1 = 17,
1134 - STD_VIDEO_AV1_MESA_LEVEL_6_2 = 18,
1135 - STD_VIDEO_AV1_MESA_LEVEL_6_3 = 19,
1136 - STD_VIDEO_AV1_MESA_LEVEL_7_0 = 20,
1137 - STD_VIDEO_AV1_MESA_LEVEL_7_1 = 21,
1138 - STD_VIDEO_AV1_MESA_LEVEL_7_2 = 22,
1139 - STD_VIDEO_AV1_MESA_LEVEL_7_3 = 23,
1140 - STD_VIDEO_AV1_MESA_LEVEL_MAX = 31,
1141 -} StdVideoAV1MESALevel;
1143 -typedef struct StdVideoAV1MESAFilmGrainFlags {
1144 - uint8_t apply_grain;
1145 - uint8_t chroma_scaling_from_luma;
1146 - uint8_t overlap_flag;
1147 - uint8_t clip_to_restricted_range;
1148 -} StdVideoAV1MESAFilmGrainFlags;
1150 -typedef struct StdVideoAV1MESAFilmGrainParameters {
1151 - StdVideoAV1MESAFilmGrainFlags flags;
1152 - uint32_t grain_scaling_minus_8;
1153 - uint32_t ar_coeff_lag;
1154 - uint32_t ar_coeff_shift_minus_6;
1155 - uint32_t grain_scale_shift;
1157 - uint16_t grain_seed;
1158 - uint8_t num_y_points;
1159 - uint8_t point_y_value[14];
1160 - uint8_t point_y_scaling[14];
1162 - uint8_t num_cb_points;
1163 - uint8_t point_cb_value[10];
1164 - uint8_t point_cb_scaling[10];
1166 - uint8_t num_cr_points;
1167 - uint8_t point_cr_value[10];
1168 - uint8_t point_cr_scaling[10];
1170 - int8_t ar_coeffs_y_plus_128[24];
1171 - int8_t ar_coeffs_cb_plus_128[25];
1172 - int8_t ar_coeffs_cr_plus_128[25];
1174 - uint8_t cb_luma_mult;
1175 - uint16_t cb_offset;
1177 - uint8_t cr_luma_mult;
1178 - uint16_t cr_offset;
1179 -} StdVideoAV1MESAFilmGrainParameters;
1181 -typedef struct StdVideoAV1MESAGlobalMotionFlags {
1182 - uint8_t gm_invalid;
1183 -} StdVideoAV1MESAGlobalMotionFlags;
1185 -typedef struct StdVideoAV1MESAGlobalMotion {
1186 - StdVideoAV1MESAGlobalMotionFlags flags;
1188 - uint32_t gm_params[6];
1189 -} StdVideoAV1MESAGlobalMotion;
1191 -typedef struct StdVideoAV1MESALoopRestoration {
1192 - uint8_t lr_type[3];
1193 - uint8_t lr_unit_shift;
1194 - uint8_t lr_uv_shift;
1195 -} StdVideoAV1MESALoopRestoration;
1197 -typedef struct StdVideoAV1MESATileInfoFlags {
1198 - uint8_t uniform_tile_spacing_flag;
1199 -} StdVideoAV1MESATileInfoFlags;
1201 -typedef struct StdVideoAV1MESATileInfo {
1202 - StdVideoAV1MESATileInfoFlags flags;
1203 - uint8_t tile_cols;
1204 - uint8_t tile_rows;
1205 - uint8_t tile_start_col_sb[64];
1206 - uint8_t tile_start_row_sb[64];
1207 - uint8_t width_in_sbs_minus_1[64];
1208 - uint8_t height_in_sbs_minus_1[64];
1209 - uint16_t context_update_tile_id;
1210 - uint8_t tile_size_bytes_minus1;
1211 -} StdVideoAV1MESATileInfo;
1213 -typedef struct StdVideoAV1MESAQuantizationFlags {
1214 - uint8_t using_qmatrix;
1215 -} StdVideoAV1MESAQuantizationFlags;
1217 -typedef struct StdVideoAV1MESAQuantization {
1218 - StdVideoAV1MESAQuantizationFlags flags;
1219 - uint8_t base_q_idx;
1220 - int8_t delta_q_y_dc;
1221 - uint8_t diff_uv_delta;
1222 - int8_t delta_q_u_dc;
1223 - int8_t delta_q_u_ac;
1224 - int8_t delta_q_v_dc;
1225 - int8_t delta_q_v_ac;
1229 -} StdVideoAV1MESAQuantization;
1231 -typedef struct StdVideoAV1MESACDEF {
1232 - uint8_t damping_minus_3;
1234 - uint8_t y_pri_strength[8];
1235 - uint8_t y_sec_strength[8];
1236 - uint8_t uv_pri_strength[8];
1237 - uint8_t uv_sec_strength[8];
1238 -} StdVideoAV1MESACDEF;
1240 -typedef struct StdVideoAV1MESADeltaQFlags {
1241 - uint8_t delta_lf_present;
1242 - uint8_t delta_lf_multi;
1243 -} StdVideoAV1MESADeltaQFlags;
1245 -typedef struct StdVideoAV1MESADeltaQ {
1246 - StdVideoAV1MESADeltaQFlags flags;
1247 - uint8_t delta_q_res;
1248 - uint8_t delta_lf_res;
1249 -} StdVideoAV1MESADeltaQ;
1251 -typedef struct StdVideoAV1MESASegmentationFlags {
1253 - uint8_t update_map;
1254 - uint8_t temporal_update;
1255 - uint8_t update_data;
1256 -} StdVideoAV1MESASegmentationFlags;
1258 -typedef struct StdVideoAV1MESASegmentation {
1259 - StdVideoAV1MESASegmentationFlags flags;
1260 - uint8_t feature_enabled_bits[8];
1261 - int16_t feature_data[8][8];
1262 -} StdVideoAV1MESASegmentation;
1264 -typedef struct StdVideoAV1MESALoopFilterFlags {
1265 - uint8_t delta_enabled;
1266 - uint8_t delta_update;
1267 -} StdVideoAV1MESALoopFilterFlags;
1269 -typedef struct StdVideoAV1MESALoopFilter {
1270 - StdVideoAV1MESALoopFilterFlags flags;
1272 - uint8_t sharpness;
1273 - int8_t ref_deltas[8];
1274 - int8_t mode_deltas[2];
1275 -} StdVideoAV1MESALoopFilter;
1277 -typedef struct StdVideoAV1MESAFrameHeaderFlags {
1278 - uint8_t error_resilient_mode;
1279 - uint8_t disable_cdf_update;
1280 - uint8_t use_superres;
1281 - uint8_t render_and_frame_size_different;
1282 - uint8_t allow_screen_content_tools;
1283 - uint8_t is_filter_switchable;
1284 - uint8_t force_integer_mv;
1285 - uint8_t frame_size_override_flag;
1286 - uint8_t buffer_removal_time_present_flag;
1287 - uint8_t allow_intrabc;
1288 - uint8_t frame_refs_short_signaling;
1289 - uint8_t allow_high_precision_mv;
1290 - uint8_t is_motion_mode_switchable;
1291 - uint8_t use_ref_frame_mvs;
1292 - uint8_t disable_frame_end_update_cdf;
1293 - uint8_t allow_warped_motion;
1294 - uint8_t reduced_tx_set;
1295 - uint8_t reference_select;
1296 - uint8_t skip_mode_present;
1297 - uint8_t delta_q_present;
1299 -} StdVideoAV1MESAFrameHeaderFlags;
1301 -typedef struct StdVideoAV1MESAFrameHeader {
1302 - StdVideoAV1MESAFrameHeaderFlags flags;
1304 - uint32_t frame_presentation_time;
1305 - uint32_t display_frame_id;
1306 - uint32_t current_frame_id;
1307 - uint8_t frame_to_show_map_idx;
1308 - uint8_t frame_type;
1309 - uint8_t order_hint;
1310 - uint8_t primary_ref_frame;
1311 - uint16_t frame_width_minus_1;
1312 - uint16_t frame_height_minus_1;
1313 - uint16_t render_width_minus_1;
1314 - uint16_t render_height_minus_1;
1315 - uint8_t coded_denom;
1317 - uint8_t refresh_frame_flags;
1318 - uint8_t ref_order_hint[8];
1319 - int8_t ref_frame_idx[7];
1320 - uint32_t delta_frame_id_minus1[7];
1322 - uint8_t interpolation_filter;
1325 - StdVideoAV1MESATileInfo tiling;
1326 - StdVideoAV1MESAQuantization quantization;
1327 - StdVideoAV1MESASegmentation segmentation;
1328 - StdVideoAV1MESADeltaQ delta_q;
1329 - StdVideoAV1MESALoopFilter loop_filter;
1330 - StdVideoAV1MESACDEF cdef;
1331 - StdVideoAV1MESALoopRestoration lr;
1332 - StdVideoAV1MESAGlobalMotion global_motion[8]; // One per ref frame
1333 - StdVideoAV1MESAFilmGrainParameters film_grain;
1334 -} StdVideoAV1MESAFrameHeader;
1336 -typedef struct StdVideoAV1MESAScreenCoding {
1337 - uint8_t seq_force_screen_content_tools;
1338 -} StdVideoAV1MESAScreenCoding;
1340 -typedef struct StdVideoAV1MESATimingInfoFlags {
1341 - uint8_t equal_picture_interval;
1342 -} StdVideoAV1MESATimingInfoFlags;
1344 -typedef struct StdVideoAV1MESATimingInfo {
1345 - StdVideoAV1MESATimingInfoFlags flags;
1346 - uint32_t num_units_in_display_tick;
1347 - uint32_t time_scale;
1348 - uint32_t num_ticks_per_picture_minus_1;
1349 -} StdVideoAV1MESATimingInfo;
1351 -typedef struct StdVideoAV1MESAColorConfigFlags {
1352 - uint8_t mono_chrome;
1353 - uint8_t color_range;
1354 - uint8_t separate_uv_delta_q;
1355 -} StdVideoAV1MESAColorConfigFlags;
1357 -typedef struct StdVideoAV1MESAColorConfig {
1358 - StdVideoAV1MESAColorConfigFlags flags;
1359 - uint8_t bit_depth;
1360 - uint8_t subsampling_x;
1361 - uint8_t subsampling_y;
1362 -} StdVideoAV1MESAColorConfig;
1364 -typedef struct StdVideoAV1MESASequenceHeaderFlags {
1365 - uint8_t still_picture;
1366 - uint8_t reduced_still_picture_header;
1367 - uint8_t use_128x128_superblock;
1368 - uint8_t enable_filter_intra;
1369 - uint8_t enable_intra_edge_filter;
1370 - uint8_t enable_interintra_compound;
1371 - uint8_t enable_masked_compound;
1372 - uint8_t enable_warped_motion;
1373 - uint8_t enable_dual_filter;
1374 - uint8_t enable_order_hint;
1375 - uint8_t enable_jnt_comp;
1376 - uint8_t enable_ref_frame_mvs;
1377 - uint8_t frame_id_numbers_present_flag;
1378 - uint8_t enable_superres;
1379 - uint8_t enable_cdef;
1380 - uint8_t enable_restoration;
1381 - uint8_t film_grain_params_present;
1382 - uint8_t timing_info_present_flag;
1383 - uint8_t initial_display_delay_present_flag;
1384 -} StdVideoAV1MESASequenceHeaderFlags;
1386 -typedef struct StdVideoAV1MESASequenceHeader {
1387 - StdVideoAV1MESASequenceHeaderFlags flags;
1389 - StdVideoAV1MESAProfile seq_profile;
1390 - uint8_t frame_width_bits_minus_1;
1391 - uint8_t frame_height_bits_minus_1;
1392 - uint16_t max_frame_width_minus_1;
1393 - uint16_t max_frame_height_minus_1;
1394 - uint8_t delta_frame_id_length_minus_2;
1395 - uint8_t additional_frame_id_length_minus_1;
1396 - uint8_t order_hint_bits_minus_1;
1397 - uint8_t seq_choose_integer_mv;
1398 - uint8_t seq_force_integer_mv;
1400 - StdVideoAV1MESATimingInfo timing_info;
1401 - StdVideoAV1MESAColorConfig color_config;
1402 -} StdVideoAV1MESASequenceHeader;
1404 -typedef struct StdVideoAV1MESATile {
1405 - uint16_t tg_start;
1411 -} StdVideoAV1MESATile;
1413 -typedef struct StdVideoAV1MESATileList {
1414 - StdVideoAV1MESATile *tile_list;
1415 - uint32_t nb_tiles;
1416 -} StdVideoAV1MESATileList;
1418 -typedef struct VkVideoDecodeAV1PictureInfoMESA {
1419 - VkStructureType sType;
1420 - const void *pNext;
1421 - StdVideoAV1MESAFrameHeader *frame_header;
1422 - StdVideoAV1MESATileList *tile_list;
1423 - uint8_t skip_mode_frame_idx[2];
1424 -} VkVideoDecodeAV1PictureInfoMESA;
1426 -typedef struct VkVideoDecodeAV1DpbSlotInfoMESA {
1427 - VkStructureType sType;
1428 - const void *pNext;
1430 - uint8_t ref_order_hint[7];
1431 - uint8_t disable_frame_end_update_cdf;
1432 -} VkVideoDecodeAV1DpbSlotInfoMESA;
1434 -typedef struct VkVideoDecodeAV1SessionParametersAddInfoMESA {
1435 - VkStructureType sType;
1436 - const void *pNext;
1437 - StdVideoAV1MESASequenceHeader *sequence_header;
1438 -} VkVideoDecodeAV1SessionParametersAddInfoMESA;
1440 -typedef struct VkVideoDecodeAV1SessionParametersCreateInfoMESA {
1441 - VkStructureType sType;
1442 - const void *pNext;
1443 - const VkVideoDecodeAV1SessionParametersAddInfoMESA *pParametersAddInfo;
1444 -} VkVideoDecodeAV1SessionParametersCreateInfoMESA;
1446 -typedef struct VkVideoDecodeAV1ProfileInfoMESA {
1447 - VkStructureType sType;
1448 - const void *pNext;
1449 - StdVideoAV1MESAProfile stdProfileIdc;
1450 -} VkVideoDecodeAV1ProfileInfoMESA;
1452 -typedef enum VkVideoDecodeAV1CapabilityFlagBitsMESA {
1453 - VK_VIDEO_DECODE_AV1_CAPABILITY_EXTERNAL_FILM_GRAIN_MESA = 0x00000001,
1454 - VK_VIDEO_DECODE_AV1_CAPABILITY_FLAG_BITS_MAX_ENUM_MESA = 0x7FFFFFFF
1455 -} VkVideoDecodeAV1CapabilityFlagBitsMESA;
1456 -typedef VkFlags VkVideoDecodeAV1CapabilityFlagsMESA;
1458 -typedef struct VkVideoDecodeAV1CapabilitiesMESA {
1459 - VkStructureType sType;
1460 - const void *pNext;
1461 - VkVideoDecodeAV1CapabilityFlagsMESA flags;
1462 - StdVideoAV1MESALevel maxLevelIdc;
1463 -} VkVideoDecodeAV1CapabilitiesMESA;
1465 -#define VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_PICTURE_INFO_MESA 1000509000
1466 -#define VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_SESSION_PARAMETERS_CREATE_INFO_MESA 1000509001
1467 -#define VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_SESSION_PARAMETERS_ADD_INFO_MESA 1000509002
1468 -#define VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_DPB_SLOT_INFO_MESA 1000509003
1469 -#define VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_CAPABILITIES_MESA 1000509004
1470 -#define VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_PROFILE_INFO_MESA 1000509005
1477 diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
1478 index 91b9f96ccf..6e3b96b73a 100644
1479 --- a/libavutil/hwcontext_vulkan.c
1480 +++ b/libavutil/hwcontext_vulkan.c
1481 @@ -446,7 +446,7 @@ static const VulkanOptExtension optional_device_exts[] = {
1482 { VK_KHR_VIDEO_DECODE_QUEUE_EXTENSION_NAME, FF_VK_EXT_VIDEO_DECODE_QUEUE },
1483 { VK_KHR_VIDEO_DECODE_H264_EXTENSION_NAME, FF_VK_EXT_VIDEO_DECODE_H264 },
1484 { VK_KHR_VIDEO_DECODE_H265_EXTENSION_NAME, FF_VK_EXT_VIDEO_DECODE_H265 },
1485 - { "VK_MESA_video_decode_av1", FF_VK_EXT_VIDEO_DECODE_AV1 },
1486 + { VK_KHR_VIDEO_DECODE_AV1_EXTENSION_NAME, FF_VK_EXT_VIDEO_DECODE_AV1 },
1489 static VkBool32 VKAPI_CALL vk_dbg_callback(VkDebugUtilsMessageSeverityFlagBitsEXT severity,
1490 diff --git a/libavutil/vulkan_functions.h b/libavutil/vulkan_functions.h
1491 index 65021b04b1..6b379acf93 100644
1492 --- a/libavutil/vulkan_functions.h
1493 +++ b/libavutil/vulkan_functions.h
1494 @@ -43,7 +43,7 @@ typedef enum FFVulkanExtensions {
1495 FF_VK_EXT_VIDEO_DECODE_QUEUE = 1ULL << 11, /* VK_KHR_video_decode_queue */
1496 FF_VK_EXT_VIDEO_DECODE_H264 = 1ULL << 12, /* VK_EXT_video_decode_h264 */
1497 FF_VK_EXT_VIDEO_DECODE_H265 = 1ULL << 13, /* VK_EXT_video_decode_h265 */
1498 - FF_VK_EXT_VIDEO_DECODE_AV1 = 1ULL << 14, /* VK_MESA_video_decode_av1 */
1499 + FF_VK_EXT_VIDEO_DECODE_AV1 = 1ULL << 14, /* VK_KHR_video_decode_av1 */
1500 FF_VK_EXT_ATOMIC_FLOAT = 1ULL << 15, /* VK_EXT_shader_atomic_float */
1501 FF_VK_EXT_COOP_MATRIX = 1ULL << 16, /* VK_KHR_cooperative_matrix */
1503 diff --git a/libavutil/vulkan_loader.h b/libavutil/vulkan_loader.h
1504 index f9e739e1e3..73cf03935d 100644
1505 --- a/libavutil/vulkan_loader.h
1506 +++ b/libavutil/vulkan_loader.h
1507 @@ -58,7 +58,7 @@ static inline uint64_t ff_vk_extensions_to_mask(const char * const *extensions,
1508 { VK_KHR_VIDEO_DECODE_QUEUE_EXTENSION_NAME, FF_VK_EXT_VIDEO_DECODE_QUEUE },
1509 { VK_KHR_VIDEO_DECODE_H264_EXTENSION_NAME, FF_VK_EXT_VIDEO_DECODE_H264 },
1510 { VK_KHR_VIDEO_DECODE_H265_EXTENSION_NAME, FF_VK_EXT_VIDEO_DECODE_H265 },
1511 - { "VK_MESA_video_decode_av1", FF_VK_EXT_VIDEO_DECODE_AV1 },
1512 + { VK_KHR_VIDEO_DECODE_AV1_EXTENSION_NAME, FF_VK_EXT_VIDEO_DECODE_AV1 },
1515 FFVulkanExtensions mask = 0x0;
1516 --- ffmpeg-6.1.1/libavcodec/vulkan_video.c.orig 2023-11-11 01:25:17.000000000 +0100
1517 +++ ffmpeg-6.1.1/libavcodec/vulkan_video.c 2024-04-15 21:49:23.744852316 +0200
1518 @@ -37,7 +37,7 @@ const FFVkCodecMap ff_vk_codec_map[AV_CO
1521 FF_VK_EXT_VIDEO_DECODE_AV1,
1522 - 0x01000000 /* TODO fix this */
1523 + VK_VIDEO_CODEC_OPERATION_DECODE_AV1_BIT_KHR,