2 * Video Decode and Presentation API for UNIX (VDPAU) is used for
3 * HW decode acceleration for MPEG-1/2, MPEG-4 ASP, H.264 and VC-1.
5 * Copyright (c) 2008 NVIDIA
7 * This file is part of FFmpeg.
9 * FFmpeg is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
14 * FFmpeg is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with FFmpeg; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
33 #include "vdpau_internal.h"
36 * @addtogroup VDPAU_Decoding
41 AVVDPAUContext *av_alloc_vdpaucontext(void)
43 return av_vdpau_alloc_context();
46 MAKE_ACCESSORS(AVVDPAUContext, vdpau_hwaccel, AVVDPAU_Render2, render2)
48 int ff_vdpau_common_start_frame(Picture *pic,
49 av_unused const uint8_t *buffer,
50 av_unused uint32_t size)
52 struct vdpau_picture_context *pic_ctx = pic->hwaccel_picture_private;
54 pic_ctx->bitstream_buffers_allocated = 0;
55 pic_ctx->bitstream_buffers_used = 0;
56 pic_ctx->bitstream_buffers = NULL;
60 #if CONFIG_H263_VDPAU_HWACCEL || CONFIG_MPEG1_VDPAU_HWACCEL || \
61 CONFIG_MPEG2_VDPAU_HWACCEL || CONFIG_MPEG4_VDPAU_HWACCEL || \
62 CONFIG_VC1_VDPAU_HWACCEL || CONFIG_WMV3_VDPAU_HWACCEL
63 int ff_vdpau_mpeg_end_frame(AVCodecContext *avctx)
66 AVVDPAUContext *hwctx = avctx->hwaccel_context;
67 MpegEncContext *s = avctx->priv_data;
68 Picture *pic = s->current_picture_ptr;
69 struct vdpau_picture_context *pic_ctx = pic->hwaccel_picture_private;
70 VdpVideoSurface surf = ff_vdpau_get_surface_id(pic);
73 res = hwctx->render2(avctx, &pic->f, (void *)&pic_ctx->info,
74 pic_ctx->bitstream_buffers_used, pic_ctx->bitstream_buffers);
76 hwctx->render(hwctx->decoder, surf, (void *)&pic_ctx->info,
77 pic_ctx->bitstream_buffers_used, pic_ctx->bitstream_buffers);
79 ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
80 av_freep(&pic_ctx->bitstream_buffers);
86 int ff_vdpau_add_buffer(Picture *pic, const uint8_t *buf, uint32_t size)
88 struct vdpau_picture_context *pic_ctx = pic->hwaccel_picture_private;
89 VdpBitstreamBuffer *buffers = pic_ctx->bitstream_buffers;
91 buffers = av_fast_realloc(buffers, &pic_ctx->bitstream_buffers_allocated,
92 (pic_ctx->bitstream_buffers_used + 1) * sizeof(*buffers));
94 return AVERROR(ENOMEM);
96 pic_ctx->bitstream_buffers = buffers;
97 buffers += pic_ctx->bitstream_buffers_used++;
99 buffers->struct_version = VDP_BITSTREAM_BUFFER_VERSION;
100 buffers->bitstream = buf;
101 buffers->bitstream_bytes = size;
105 /* Obsolete non-hwaccel VDPAU support below... */
107 void ff_vdpau_h264_set_reference_frames(H264Context *h)
109 struct vdpau_render_state *render, *render_ref;
110 VdpReferenceFrameH264 *rf, *rf2;
112 int i, list, pic_frame_idx;
114 render = (struct vdpau_render_state *)h->cur_pic_ptr->f.data[0];
117 rf = &render->info.h264.referenceFrames[0];
118 #define H264_RF_COUNT FF_ARRAY_ELEMS(render->info.h264.referenceFrames)
120 for (list = 0; list < 2; ++list) {
121 Picture **lp = list ? h->long_ref : h->short_ref;
122 int ls = list ? 16 : h->short_ref_count;
124 for (i = 0; i < ls; ++i) {
126 if (!pic || !pic->reference)
128 pic_frame_idx = pic->long_ref ? pic->pic_id : pic->frame_num;
130 render_ref = (struct vdpau_render_state *)pic->f.data[0];
133 rf2 = &render->info.h264.referenceFrames[0];
136 (rf2->surface == render_ref->surface)
137 && (rf2->is_long_term == pic->long_ref)
138 && (rf2->frame_idx == pic_frame_idx)
144 rf2->top_is_reference |= (pic->reference & PICT_TOP_FIELD) ? VDP_TRUE : VDP_FALSE;
145 rf2->bottom_is_reference |= (pic->reference & PICT_BOTTOM_FIELD) ? VDP_TRUE : VDP_FALSE;
149 if (rf >= &render->info.h264.referenceFrames[H264_RF_COUNT])
152 rf->surface = render_ref->surface;
153 rf->is_long_term = pic->long_ref;
154 rf->top_is_reference = (pic->reference & PICT_TOP_FIELD) ? VDP_TRUE : VDP_FALSE;
155 rf->bottom_is_reference = (pic->reference & PICT_BOTTOM_FIELD) ? VDP_TRUE : VDP_FALSE;
156 rf->field_order_cnt[0] = pic->field_poc[0];
157 rf->field_order_cnt[1] = pic->field_poc[1];
158 rf->frame_idx = pic_frame_idx;
164 for (; rf < &render->info.h264.referenceFrames[H264_RF_COUNT]; ++rf) {
165 rf->surface = VDP_INVALID_HANDLE;
166 rf->is_long_term = 0;
167 rf->top_is_reference = 0;
168 rf->bottom_is_reference = 0;
169 rf->field_order_cnt[0] = 0;
170 rf->field_order_cnt[1] = 0;
175 void ff_vdpau_add_data_chunk(uint8_t *data, const uint8_t *buf, int buf_size)
177 struct vdpau_render_state *render = (struct vdpau_render_state*)data;
180 render->bitstream_buffers= av_fast_realloc(
181 render->bitstream_buffers,
182 &render->bitstream_buffers_allocated,
183 sizeof(*render->bitstream_buffers)*(render->bitstream_buffers_used + 1)
186 render->bitstream_buffers[render->bitstream_buffers_used].struct_version = VDP_BITSTREAM_BUFFER_VERSION;
187 render->bitstream_buffers[render->bitstream_buffers_used].bitstream = buf;
188 render->bitstream_buffers[render->bitstream_buffers_used].bitstream_bytes = buf_size;
189 render->bitstream_buffers_used++;
192 #if CONFIG_H264_VDPAU_DECODER
193 void ff_vdpau_h264_picture_start(H264Context *h)
195 struct vdpau_render_state *render;
198 render = (struct vdpau_render_state *)h->cur_pic_ptr->f.data[0];
201 for (i = 0; i < 2; ++i) {
202 int foc = h->cur_pic_ptr->field_poc[i];
205 render->info.h264.field_order_cnt[i] = foc;
208 render->info.h264.frame_num = h->frame_num;
211 void ff_vdpau_h264_picture_complete(H264Context *h)
213 struct vdpau_render_state *render;
215 render = (struct vdpau_render_state *)h->cur_pic_ptr->f.data[0];
218 render->info.h264.slice_count = h->slice_num;
219 if (render->info.h264.slice_count < 1)
222 render->info.h264.is_reference = (h->cur_pic_ptr->reference & 3) ? VDP_TRUE : VDP_FALSE;
223 render->info.h264.field_pic_flag = h->picture_structure != PICT_FRAME;
224 render->info.h264.bottom_field_flag = h->picture_structure == PICT_BOTTOM_FIELD;
225 render->info.h264.num_ref_frames = h->sps.ref_frame_count;
226 render->info.h264.mb_adaptive_frame_field_flag = h->sps.mb_aff && !render->info.h264.field_pic_flag;
227 render->info.h264.constrained_intra_pred_flag = h->pps.constrained_intra_pred;
228 render->info.h264.weighted_pred_flag = h->pps.weighted_pred;
229 render->info.h264.weighted_bipred_idc = h->pps.weighted_bipred_idc;
230 render->info.h264.frame_mbs_only_flag = h->sps.frame_mbs_only_flag;
231 render->info.h264.transform_8x8_mode_flag = h->pps.transform_8x8_mode;
232 render->info.h264.chroma_qp_index_offset = h->pps.chroma_qp_index_offset[0];
233 render->info.h264.second_chroma_qp_index_offset = h->pps.chroma_qp_index_offset[1];
234 render->info.h264.pic_init_qp_minus26 = h->pps.init_qp - 26;
235 render->info.h264.num_ref_idx_l0_active_minus1 = h->pps.ref_count[0] - 1;
236 render->info.h264.num_ref_idx_l1_active_minus1 = h->pps.ref_count[1] - 1;
237 render->info.h264.log2_max_frame_num_minus4 = h->sps.log2_max_frame_num - 4;
238 render->info.h264.pic_order_cnt_type = h->sps.poc_type;
239 render->info.h264.log2_max_pic_order_cnt_lsb_minus4 = h->sps.poc_type ? 0 : h->sps.log2_max_poc_lsb - 4;
240 render->info.h264.delta_pic_order_always_zero_flag = h->sps.delta_pic_order_always_zero_flag;
241 render->info.h264.direct_8x8_inference_flag = h->sps.direct_8x8_inference_flag;
242 render->info.h264.entropy_coding_mode_flag = h->pps.cabac;
243 render->info.h264.pic_order_present_flag = h->pps.pic_order_present;
244 render->info.h264.deblocking_filter_control_present_flag = h->pps.deblocking_filter_parameters_present;
245 render->info.h264.redundant_pic_cnt_present_flag = h->pps.redundant_pic_cnt_present;
246 memcpy(render->info.h264.scaling_lists_4x4, h->pps.scaling_matrix4, sizeof(render->info.h264.scaling_lists_4x4));
247 memcpy(render->info.h264.scaling_lists_8x8[0], h->pps.scaling_matrix8[0], sizeof(render->info.h264.scaling_lists_8x8[0]));
248 memcpy(render->info.h264.scaling_lists_8x8[1], h->pps.scaling_matrix8[3], sizeof(render->info.h264.scaling_lists_8x8[0]));
250 ff_h264_draw_horiz_band(h, 0, h->avctx->height);
251 render->bitstream_buffers_used = 0;
253 #endif /* CONFIG_H264_VDPAU_DECODER */
255 #if CONFIG_MPEG_VDPAU_DECODER || CONFIG_MPEG1_VDPAU_DECODER
256 void ff_vdpau_mpeg_picture_complete(MpegEncContext *s, const uint8_t *buf,
257 int buf_size, int slice_count)
259 struct vdpau_render_state *render, *last, *next;
262 if (!s->current_picture_ptr) return;
264 render = (struct vdpau_render_state *)s->current_picture_ptr->f.data[0];
267 /* fill VdpPictureInfoMPEG1Or2 struct */
268 render->info.mpeg.picture_structure = s->picture_structure;
269 render->info.mpeg.picture_coding_type = s->pict_type;
270 render->info.mpeg.intra_dc_precision = s->intra_dc_precision;
271 render->info.mpeg.frame_pred_frame_dct = s->frame_pred_frame_dct;
272 render->info.mpeg.concealment_motion_vectors = s->concealment_motion_vectors;
273 render->info.mpeg.intra_vlc_format = s->intra_vlc_format;
274 render->info.mpeg.alternate_scan = s->alternate_scan;
275 render->info.mpeg.q_scale_type = s->q_scale_type;
276 render->info.mpeg.top_field_first = s->top_field_first;
277 render->info.mpeg.full_pel_forward_vector = s->full_pel[0]; // MPEG-1 only. Set 0 for MPEG-2
278 render->info.mpeg.full_pel_backward_vector = s->full_pel[1]; // MPEG-1 only. Set 0 for MPEG-2
279 render->info.mpeg.f_code[0][0] = s->mpeg_f_code[0][0]; // For MPEG-1 fill both horiz. & vert.
280 render->info.mpeg.f_code[0][1] = s->mpeg_f_code[0][1];
281 render->info.mpeg.f_code[1][0] = s->mpeg_f_code[1][0];
282 render->info.mpeg.f_code[1][1] = s->mpeg_f_code[1][1];
283 for (i = 0; i < 64; ++i) {
284 render->info.mpeg.intra_quantizer_matrix[i] = s->intra_matrix[i];
285 render->info.mpeg.non_intra_quantizer_matrix[i] = s->inter_matrix[i];
288 render->info.mpeg.forward_reference = VDP_INVALID_HANDLE;
289 render->info.mpeg.backward_reference = VDP_INVALID_HANDLE;
291 switch(s->pict_type){
292 case AV_PICTURE_TYPE_B:
293 next = (struct vdpau_render_state *)s->next_picture.f.data[0];
295 render->info.mpeg.backward_reference = next->surface;
296 // no return here, going to set forward prediction
297 case AV_PICTURE_TYPE_P:
298 last = (struct vdpau_render_state *)s->last_picture.f.data[0];
299 if (!last) // FIXME: Does this test make sense?
300 last = render; // predict second field from the first
301 render->info.mpeg.forward_reference = last->surface;
304 ff_vdpau_add_data_chunk(s->current_picture_ptr->f.data[0], buf, buf_size);
306 render->info.mpeg.slice_count = slice_count;
309 ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
310 render->bitstream_buffers_used = 0;
312 #endif /* CONFIG_MPEG_VDPAU_DECODER || CONFIG_MPEG1_VDPAU_DECODER */
314 #if CONFIG_VC1_VDPAU_DECODER
315 void ff_vdpau_vc1_decode_picture(MpegEncContext *s, const uint8_t *buf,
318 VC1Context *v = s->avctx->priv_data;
319 struct vdpau_render_state *render, *last, *next;
321 render = (struct vdpau_render_state *)s->current_picture.f.data[0];
324 /* fill LvPictureInfoVC1 struct */
325 render->info.vc1.frame_coding_mode = v->fcm ? v->fcm + 1 : 0;
326 render->info.vc1.postprocflag = v->postprocflag;
327 render->info.vc1.pulldown = v->broadcast;
328 render->info.vc1.interlace = v->interlace;
329 render->info.vc1.tfcntrflag = v->tfcntrflag;
330 render->info.vc1.finterpflag = v->finterpflag;
331 render->info.vc1.psf = v->psf;
332 render->info.vc1.dquant = v->dquant;
333 render->info.vc1.panscan_flag = v->panscanflag;
334 render->info.vc1.refdist_flag = v->refdist_flag;
335 render->info.vc1.quantizer = v->quantizer_mode;
336 render->info.vc1.extended_mv = v->extended_mv;
337 render->info.vc1.extended_dmv = v->extended_dmv;
338 render->info.vc1.overlap = v->overlap;
339 render->info.vc1.vstransform = v->vstransform;
340 render->info.vc1.loopfilter = v->s.loop_filter;
341 render->info.vc1.fastuvmc = v->fastuvmc;
342 render->info.vc1.range_mapy_flag = v->range_mapy_flag;
343 render->info.vc1.range_mapy = v->range_mapy;
344 render->info.vc1.range_mapuv_flag = v->range_mapuv_flag;
345 render->info.vc1.range_mapuv = v->range_mapuv;
346 /* Specific to simple/main profile only */
347 render->info.vc1.multires = v->multires;
348 render->info.vc1.syncmarker = v->s.resync_marker;
349 render->info.vc1.rangered = v->rangered | (v->rangeredfrm << 1);
350 render->info.vc1.maxbframes = v->s.max_b_frames;
352 render->info.vc1.deblockEnable = v->postprocflag & 1;
353 render->info.vc1.pquant = v->pq;
355 render->info.vc1.forward_reference = VDP_INVALID_HANDLE;
356 render->info.vc1.backward_reference = VDP_INVALID_HANDLE;
359 render->info.vc1.picture_type = 4;
361 render->info.vc1.picture_type = s->pict_type - 1 + s->pict_type / 3;
363 switch(s->pict_type){
364 case AV_PICTURE_TYPE_B:
365 next = (struct vdpau_render_state *)s->next_picture.f.data[0];
367 render->info.vc1.backward_reference = next->surface;
368 // no break here, going to set forward prediction
369 case AV_PICTURE_TYPE_P:
370 last = (struct vdpau_render_state *)s->last_picture.f.data[0];
371 if (!last) // FIXME: Does this test make sense?
372 last = render; // predict second field from the first
373 render->info.vc1.forward_reference = last->surface;
376 ff_vdpau_add_data_chunk(s->current_picture_ptr->f.data[0], buf, buf_size);
378 render->info.vc1.slice_count = 1;
380 ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
381 render->bitstream_buffers_used = 0;
383 #endif /* (CONFIG_VC1_VDPAU_DECODER */
385 #if CONFIG_MPEG4_VDPAU_DECODER
386 void ff_vdpau_mpeg4_decode_picture(MpegEncContext *s, const uint8_t *buf,
389 struct vdpau_render_state *render, *last, *next;
392 if (!s->current_picture_ptr) return;
394 render = (struct vdpau_render_state *)s->current_picture_ptr->f.data[0];
397 /* fill VdpPictureInfoMPEG4Part2 struct */
398 render->info.mpeg4.trd[0] = s->pp_time;
399 render->info.mpeg4.trb[0] = s->pb_time;
400 render->info.mpeg4.trd[1] = s->pp_field_time >> 1;
401 render->info.mpeg4.trb[1] = s->pb_field_time >> 1;
402 render->info.mpeg4.vop_time_increment_resolution = s->avctx->time_base.den;
403 render->info.mpeg4.vop_coding_type = 0;
404 render->info.mpeg4.vop_fcode_forward = s->f_code;
405 render->info.mpeg4.vop_fcode_backward = s->b_code;
406 render->info.mpeg4.resync_marker_disable = !s->resync_marker;
407 render->info.mpeg4.interlaced = !s->progressive_sequence;
408 render->info.mpeg4.quant_type = s->mpeg_quant;
409 render->info.mpeg4.quarter_sample = s->quarter_sample;
410 render->info.mpeg4.short_video_header = s->avctx->codec->id == AV_CODEC_ID_H263;
411 render->info.mpeg4.rounding_control = s->no_rounding;
412 render->info.mpeg4.alternate_vertical_scan_flag = s->alternate_scan;
413 render->info.mpeg4.top_field_first = s->top_field_first;
414 for (i = 0; i < 64; ++i) {
415 render->info.mpeg4.intra_quantizer_matrix[i] = s->intra_matrix[i];
416 render->info.mpeg4.non_intra_quantizer_matrix[i] = s->inter_matrix[i];
418 render->info.mpeg4.forward_reference = VDP_INVALID_HANDLE;
419 render->info.mpeg4.backward_reference = VDP_INVALID_HANDLE;
421 switch (s->pict_type) {
422 case AV_PICTURE_TYPE_B:
423 next = (struct vdpau_render_state *)s->next_picture.f.data[0];
425 render->info.mpeg4.backward_reference = next->surface;
426 render->info.mpeg4.vop_coding_type = 2;
427 // no break here, going to set forward prediction
428 case AV_PICTURE_TYPE_P:
429 last = (struct vdpau_render_state *)s->last_picture.f.data[0];
431 render->info.mpeg4.forward_reference = last->surface;
434 ff_vdpau_add_data_chunk(s->current_picture_ptr->f.data[0], buf, buf_size);
436 ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
437 render->bitstream_buffers_used = 0;
439 #endif /* CONFIG_MPEG4_VDPAU_DECODER */
441 int av_vdpau_get_profile(AVCodecContext *avctx, VdpDecoderProfile *profile)
443 #define PROFILE(prof) \
449 switch (avctx->codec_id) {
450 case AV_CODEC_ID_MPEG1VIDEO: PROFILE(VDP_DECODER_PROFILE_MPEG1);
451 case AV_CODEC_ID_MPEG2VIDEO:
452 switch (avctx->profile) {
453 case FF_PROFILE_MPEG2_MAIN: PROFILE(VDP_DECODER_PROFILE_MPEG2_MAIN);
454 case FF_PROFILE_MPEG2_SIMPLE: PROFILE(VDP_DECODER_PROFILE_MPEG2_SIMPLE);
455 default: return AVERROR(EINVAL);
457 case AV_CODEC_ID_H263: PROFILE(VDP_DECODER_PROFILE_MPEG4_PART2_ASP);
458 case AV_CODEC_ID_MPEG4:
459 switch (avctx->profile) {
460 case FF_PROFILE_MPEG4_SIMPLE: PROFILE(VDP_DECODER_PROFILE_MPEG4_PART2_SP);
461 case FF_PROFILE_MPEG4_ADVANCED_SIMPLE: PROFILE(VDP_DECODER_PROFILE_MPEG4_PART2_ASP);
462 default: return AVERROR(EINVAL);
464 case AV_CODEC_ID_H264:
465 switch (avctx->profile) {
466 case FF_PROFILE_H264_CONSTRAINED_BASELINE:
467 case FF_PROFILE_H264_BASELINE: PROFILE(VDP_DECODER_PROFILE_H264_BASELINE);
468 case FF_PROFILE_H264_MAIN: PROFILE(VDP_DECODER_PROFILE_H264_MAIN);
469 case FF_PROFILE_H264_HIGH: PROFILE(VDP_DECODER_PROFILE_H264_HIGH);
470 default: return AVERROR(EINVAL);
472 case AV_CODEC_ID_WMV3:
473 case AV_CODEC_ID_VC1:
474 switch (avctx->profile) {
475 case FF_PROFILE_VC1_SIMPLE: PROFILE(VDP_DECODER_PROFILE_VC1_SIMPLE);
476 case FF_PROFILE_VC1_MAIN: PROFILE(VDP_DECODER_PROFILE_VC1_MAIN);
477 case FF_PROFILE_VC1_ADVANCED: PROFILE(VDP_DECODER_PROFILE_VC1_ADVANCED);
478 default: return AVERROR(EINVAL);
481 return AVERROR(EINVAL);
484 AVVDPAUContext *av_vdpau_alloc_context(void)
486 return av_mallocz(sizeof(AVVDPAUContext));