da7236f1002c75ecb0a7773cb2f437743b32d0d6
[ffmpeg.git] / libavcodec / videotoolbox.c
1 /*
2  * Videotoolbox hardware acceleration
3  *
4  * copyright (c) 2012 Sebastien Zwickert
5  *
6  * This file is part of FFmpeg.
7  *
8  * FFmpeg is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * FFmpeg is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with FFmpeg; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22
23 #include "config.h"
24 #include "videotoolbox.h"
25 #include "libavutil/hwcontext_videotoolbox.h"
26 #include "vt_internal.h"
27 #include "libavutil/avutil.h"
28 #include "libavutil/hwcontext.h"
29 #include "bytestream.h"
30 #include "decode.h"
31 #include "h264dec.h"
32 #include "hevcdec.h"
33 #include "mpegvideo.h"
34 #include <TargetConditionals.h>
35
36 #ifndef kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder
37 #  define kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder CFSTR("RequireHardwareAcceleratedVideoDecoder")
38 #endif
39 #ifndef kVTVideoDecoderSpecification_EnableHardwareAcceleratedVideoDecoder
40 #  define kVTVideoDecoderSpecification_EnableHardwareAcceleratedVideoDecoder CFSTR("EnableHardwareAcceleratedVideoDecoder")
41 #endif
42
43 #if !HAVE_KCMVIDEOCODECTYPE_HEVC
44 enum { kCMVideoCodecType_HEVC = 'hvc1' };
45 #endif
46
47 #define VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING  12
48
49 typedef struct VTHWFrame {
50     CVPixelBufferRef pixbuf;
51     AVBufferRef *hw_frames_ctx;
52 } VTHWFrame;
53
54 static void videotoolbox_buffer_release(void *opaque, uint8_t *data)
55 {
56     VTHWFrame *ref = (VTHWFrame *)data;
57     av_buffer_unref(&ref->hw_frames_ctx);
58     CVPixelBufferRelease(ref->pixbuf);
59
60     av_free(data);
61 }
62
63 static int videotoolbox_buffer_copy(VTContext *vtctx,
64                                     const uint8_t *buffer,
65                                     uint32_t size)
66 {
67     void *tmp;
68
69     tmp = av_fast_realloc(vtctx->bitstream,
70                          &vtctx->allocated_size,
71                          size);
72
73     if (!tmp)
74         return AVERROR(ENOMEM);
75
76     vtctx->bitstream = tmp;
77     memcpy(vtctx->bitstream, buffer, size);
78     vtctx->bitstream_size = size;
79
80     return 0;
81 }
82
83 static int videotoolbox_postproc_frame(void *avctx, AVFrame *frame)
84 {
85     VTHWFrame *ref = (VTHWFrame *)frame->buf[0]->data;
86
87     if (!ref->pixbuf) {
88         av_log(avctx, AV_LOG_ERROR, "No frame decoded?\n");
89         av_frame_unref(frame);
90         return AVERROR_EXTERNAL;
91     }
92
93     frame->data[3] = (uint8_t*)ref->pixbuf;
94
95     if (ref->hw_frames_ctx) {
96         av_buffer_unref(&frame->hw_frames_ctx);
97         frame->hw_frames_ctx = av_buffer_ref(ref->hw_frames_ctx);
98         if (!frame->hw_frames_ctx)
99             return AVERROR(ENOMEM);
100     }
101
102     return 0;
103 }
104
105 int ff_videotoolbox_alloc_frame(AVCodecContext *avctx, AVFrame *frame)
106 {
107     size_t      size = sizeof(VTHWFrame);
108     uint8_t    *data = NULL;
109     AVBufferRef *buf = NULL;
110     int ret = ff_attach_decode_data(frame);
111     FrameDecodeData *fdd;
112     if (ret < 0)
113         return ret;
114
115     data = av_mallocz(size);
116     if (!data)
117         return AVERROR(ENOMEM);
118     buf = av_buffer_create(data, size, videotoolbox_buffer_release, NULL, 0);
119     if (!buf) {
120         av_freep(&data);
121         return AVERROR(ENOMEM);
122     }
123     frame->buf[0] = buf;
124
125     fdd = (FrameDecodeData*)frame->private_ref->data;
126     fdd->post_process = videotoolbox_postproc_frame;
127
128     frame->width  = avctx->width;
129     frame->height = avctx->height;
130     frame->format = avctx->pix_fmt;
131
132     return 0;
133 }
134
135 #define AV_W8(p, v) *(p) = (v)
136
137 CFDataRef ff_videotoolbox_avcc_extradata_create(AVCodecContext *avctx)
138 {
139     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
140     H264Context *h = avctx->priv_data;
141     CFDataRef data = NULL;
142     uint8_t *p;
143     int vt_extradata_size = 6 + 2 + h->ps.sps->data_size + 3 + h->ps.pps->data_size;
144     uint8_t *vt_extradata = av_malloc(vt_extradata_size);
145     if (!vt_extradata)
146         return NULL;
147
148     p = vt_extradata;
149
150     AV_W8(p + 0, 1); /* version */
151     AV_W8(p + 1, h->ps.sps->data[1]); /* profile */
152     AV_W8(p + 2, h->ps.sps->data[2]); /* profile compat */
153     AV_W8(p + 3, h->ps.sps->data[3]); /* level */
154     AV_W8(p + 4, 0xff); /* 6 bits reserved (111111) + 2 bits nal size length - 3 (11) */
155     AV_W8(p + 5, 0xe1); /* 3 bits reserved (111) + 5 bits number of sps (00001) */
156     AV_WB16(p + 6, h->ps.sps->data_size);
157     memcpy(p + 8, h->ps.sps->data, h->ps.sps->data_size);
158     p += 8 + h->ps.sps->data_size;
159     AV_W8(p + 0, 1); /* number of pps */
160     AV_WB16(p + 1, h->ps.pps->data_size);
161     memcpy(p + 3, h->ps.pps->data, h->ps.pps->data_size);
162
163     p += 3 + h->ps.pps->data_size;
164     av_assert0(p - vt_extradata == vt_extradata_size);
165
166     // save sps header (profile/level) used to create decoder session,
167     // so we can detect changes and recreate it.
168     if (vtctx)
169         memcpy(vtctx->sps, h->ps.sps->data + 1, 3);
170
171     data = CFDataCreate(kCFAllocatorDefault, vt_extradata, vt_extradata_size);
172     av_free(vt_extradata);
173     return data;
174 }
175
176 CFDataRef ff_videotoolbox_hvcc_extradata_create(AVCodecContext *avctx)
177 {
178     HEVCContext *h = avctx->priv_data;
179     const HEVCVPS *vps = (const HEVCVPS *)h->ps.vps_list[0]->data;
180     const HEVCSPS *sps = (const HEVCSPS *)h->ps.sps_list[0]->data;
181     int i, num_pps = 0;
182     const HEVCPPS *pps = h->ps.pps;
183     PTLCommon ptlc = vps->ptl.general_ptl;
184     VUI vui = sps->vui;
185     uint8_t parallelismType;
186     CFDataRef data = NULL;
187     uint8_t *p;
188     int vt_extradata_size = 23 + 5 + vps->data_size + 5 + sps->data_size + 3;
189     uint8_t *vt_extradata;
190
191     for (i = 0; i < HEVC_MAX_PPS_COUNT; i++) {
192         if (h->ps.pps_list[i]) {
193             const HEVCPPS *pps = (const HEVCPPS *)h->ps.pps_list[i]->data;
194             vt_extradata_size += 2 + pps->data_size;
195             num_pps++;
196         }
197     }
198
199     vt_extradata = av_malloc(vt_extradata_size);
200     if (!vt_extradata)
201         return NULL;
202     p = vt_extradata;
203
204     /* unsigned int(8) configurationVersion = 1; */
205     AV_W8(p + 0, 1);
206
207     /*
208      * unsigned int(2) general_profile_space;
209      * unsigned int(1) general_tier_flag;
210      * unsigned int(5) general_profile_idc;
211      */
212     AV_W8(p + 1, ptlc.profile_space << 6 |
213                  ptlc.tier_flag     << 5 |
214                  ptlc.profile_idc);
215
216     /* unsigned int(32) general_profile_compatibility_flags; */
217     memcpy(p + 2, ptlc.profile_compatibility_flag, 4);
218
219     /* unsigned int(48) general_constraint_indicator_flags; */
220     AV_W8(p + 6, ptlc.progressive_source_flag    << 7 |
221                  ptlc.interlaced_source_flag     << 6 |
222                  ptlc.non_packed_constraint_flag << 5 |
223                  ptlc.frame_only_constraint_flag << 4);
224     AV_W8(p + 7, 0);
225     AV_WN32(p + 8, 0);
226
227     /* unsigned int(8) general_level_idc; */
228     AV_W8(p + 12, ptlc.level_idc);
229
230     /*
231      * bit(4) reserved = ‘1111’b;
232      * unsigned int(12) min_spatial_segmentation_idc;
233      */
234     AV_W8(p + 13, 0xf0 | (vui.min_spatial_segmentation_idc >> 4));
235     AV_W8(p + 14, vui.min_spatial_segmentation_idc & 0xff);
236
237     /*
238      * bit(6) reserved = ‘111111’b;
239      * unsigned int(2) parallelismType;
240      */
241     if (!vui.min_spatial_segmentation_idc)
242         parallelismType = 0;
243     else if (pps->entropy_coding_sync_enabled_flag && pps->tiles_enabled_flag)
244         parallelismType = 0;
245     else if (pps->entropy_coding_sync_enabled_flag)
246         parallelismType = 3;
247     else if (pps->tiles_enabled_flag)
248         parallelismType = 2;
249     else
250         parallelismType = 1;
251     AV_W8(p + 15, 0xfc | parallelismType);
252
253     /*
254      * bit(6) reserved = ‘111111’b;
255      * unsigned int(2) chromaFormat;
256      */
257     AV_W8(p + 16, sps->chroma_format_idc | 0xfc);
258
259     /*
260      * bit(5) reserved = ‘11111’b;
261      * unsigned int(3) bitDepthLumaMinus8;
262      */
263     AV_W8(p + 17, (sps->bit_depth - 8) | 0xfc);
264
265     /*
266      * bit(5) reserved = ‘11111’b;
267      * unsigned int(3) bitDepthChromaMinus8;
268      */
269     AV_W8(p + 18, (sps->bit_depth_chroma - 8) | 0xfc);
270
271     /* bit(16) avgFrameRate; */
272     AV_WB16(p + 19, 0);
273
274     /*
275      * bit(2) constantFrameRate;
276      * bit(3) numTemporalLayers;
277      * bit(1) temporalIdNested;
278      * unsigned int(2) lengthSizeMinusOne;
279      */
280     AV_W8(p + 21, 0                             << 6 |
281                   sps->max_sub_layers           << 3 |
282                   sps->temporal_id_nesting_flag << 2 |
283                   3);
284
285     /* unsigned int(8) numOfArrays; */
286     AV_W8(p + 22, 3);
287
288     p += 23;
289     /* vps */
290     /*
291      * bit(1) array_completeness;
292      * unsigned int(1) reserved = 0;
293      * unsigned int(6) NAL_unit_type;
294      */
295     AV_W8(p, 1 << 7 |
296              HEVC_NAL_VPS & 0x3f);
297     /* unsigned int(16) numNalus; */
298     AV_WB16(p + 1, 1);
299     /* unsigned int(16) nalUnitLength; */
300     AV_WB16(p + 3, vps->data_size);
301     /* bit(8*nalUnitLength) nalUnit; */
302     memcpy(p + 5, vps->data, vps->data_size);
303     p += 5 + vps->data_size;
304
305     /* sps */
306     AV_W8(p, 1 << 7 |
307              HEVC_NAL_SPS & 0x3f);
308     AV_WB16(p + 1, 1);
309     AV_WB16(p + 3, sps->data_size);
310     memcpy(p + 5, sps->data, sps->data_size);
311     p += 5 + sps->data_size;
312
313     /* pps */
314     AV_W8(p, 1 << 7 |
315              HEVC_NAL_PPS & 0x3f);
316     AV_WB16(p + 1, num_pps);
317     p += 3;
318     for (i = 0; i < HEVC_MAX_PPS_COUNT; i++) {
319         if (h->ps.pps_list[i]) {
320             const HEVCPPS *pps = (const HEVCPPS *)h->ps.pps_list[i]->data;
321             AV_WB16(p, pps->data_size);
322             memcpy(p + 2, pps->data, pps->data_size);
323             p += 2 + pps->data_size;
324         }
325     }
326
327     av_assert0(p - vt_extradata == vt_extradata_size);
328
329     data = CFDataCreate(kCFAllocatorDefault, vt_extradata, vt_extradata_size);
330     av_free(vt_extradata);
331     return data;
332 }
333
334 int ff_videotoolbox_h264_start_frame(AVCodecContext *avctx,
335                                      const uint8_t *buffer,
336                                      uint32_t size)
337 {
338     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
339     H264Context *h = avctx->priv_data;
340
341     if (h->is_avc == 1) {
342         return videotoolbox_buffer_copy(vtctx, buffer, size);
343     }
344
345     return 0;
346 }
347
348 static int videotoolbox_h264_decode_params(AVCodecContext *avctx,
349                                            int type,
350                                            const uint8_t *buffer,
351                                            uint32_t size)
352 {
353     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
354     H264Context *h = avctx->priv_data;
355
356     // save sps header (profile/level) used to create decoder session
357     if (!vtctx->sps[0])
358         memcpy(vtctx->sps, h->ps.sps->data + 1, 3);
359
360     if (type == H264_NAL_SPS) {
361         if (size > 4 && memcmp(vtctx->sps, buffer + 1, 3) != 0) {
362             vtctx->reconfig_needed = true;
363             memcpy(vtctx->sps, buffer + 1, 3);
364         }
365     }
366
367     // pass-through SPS/PPS changes to the decoder
368     return ff_videotoolbox_h264_decode_slice(avctx, buffer, size);
369 }
370
371 static int videotoolbox_common_decode_slice(AVCodecContext *avctx,
372                                             const uint8_t *buffer,
373                                             uint32_t size)
374 {
375     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
376     void *tmp;
377
378     tmp = av_fast_realloc(vtctx->bitstream,
379                           &vtctx->allocated_size,
380                           vtctx->bitstream_size+size+4);
381     if (!tmp)
382         return AVERROR(ENOMEM);
383
384     vtctx->bitstream = tmp;
385
386     AV_WB32(vtctx->bitstream + vtctx->bitstream_size, size);
387     memcpy(vtctx->bitstream + vtctx->bitstream_size + 4, buffer, size);
388
389     vtctx->bitstream_size += size + 4;
390
391     return 0;
392 }
393
394 int ff_videotoolbox_h264_decode_slice(AVCodecContext *avctx,
395                                       const uint8_t *buffer,
396                                       uint32_t size)
397 {
398     H264Context *h = avctx->priv_data;
399
400     if (h->is_avc == 1)
401         return 0;
402
403     return videotoolbox_common_decode_slice(avctx, buffer, size);
404 }
405
406 int ff_videotoolbox_uninit(AVCodecContext *avctx)
407 {
408     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
409     if (vtctx) {
410         av_freep(&vtctx->bitstream);
411         if (vtctx->frame)
412             CVPixelBufferRelease(vtctx->frame);
413     }
414
415     return 0;
416 }
417
418 #if CONFIG_VIDEOTOOLBOX
419 // Return the AVVideotoolboxContext that matters currently. Where it comes from
420 // depends on the API used.
421 static AVVideotoolboxContext *videotoolbox_get_context(AVCodecContext *avctx)
422 {
423     // Somewhat tricky because the user can call av_videotoolbox_default_free()
424     // at any time, even when the codec is closed.
425     if (avctx->internal && avctx->internal->hwaccel_priv_data) {
426         VTContext *vtctx = avctx->internal->hwaccel_priv_data;
427         if (vtctx->vt_ctx)
428             return vtctx->vt_ctx;
429     }
430     return avctx->hwaccel_context;
431 }
432
433 static int videotoolbox_buffer_create(AVCodecContext *avctx, AVFrame *frame)
434 {
435     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
436     CVPixelBufferRef pixbuf = (CVPixelBufferRef)vtctx->frame;
437     OSType pixel_format = CVPixelBufferGetPixelFormatType(pixbuf);
438     enum AVPixelFormat sw_format = av_map_videotoolbox_format_to_pixfmt(pixel_format);
439     int width = CVPixelBufferGetWidth(pixbuf);
440     int height = CVPixelBufferGetHeight(pixbuf);
441     AVHWFramesContext *cached_frames;
442     VTHWFrame *ref;
443     int ret;
444
445     if (!frame->buf[0] || frame->data[3]) {
446         av_log(avctx, AV_LOG_ERROR, "videotoolbox: invalid state\n");
447         av_frame_unref(frame);
448         return AVERROR_EXTERNAL;
449     }
450
451     ref = (VTHWFrame *)frame->buf[0]->data;
452
453     if (ref->pixbuf)
454         CVPixelBufferRelease(ref->pixbuf);
455     ref->pixbuf = vtctx->frame;
456     vtctx->frame = NULL;
457
458     // Old API code path.
459     if (!vtctx->cached_hw_frames_ctx)
460         return 0;
461
462     cached_frames = (AVHWFramesContext*)vtctx->cached_hw_frames_ctx->data;
463
464     if (cached_frames->sw_format != sw_format ||
465         cached_frames->width != width ||
466         cached_frames->height != height) {
467         AVBufferRef *hw_frames_ctx = av_hwframe_ctx_alloc(cached_frames->device_ref);
468         AVHWFramesContext *hw_frames;
469         if (!hw_frames_ctx)
470             return AVERROR(ENOMEM);
471
472         hw_frames = (AVHWFramesContext*)hw_frames_ctx->data;
473         hw_frames->format = cached_frames->format;
474         hw_frames->sw_format = sw_format;
475         hw_frames->width = width;
476         hw_frames->height = height;
477
478         ret = av_hwframe_ctx_init(hw_frames_ctx);
479         if (ret < 0) {
480             av_buffer_unref(&hw_frames_ctx);
481             return ret;
482         }
483
484         av_buffer_unref(&vtctx->cached_hw_frames_ctx);
485         vtctx->cached_hw_frames_ctx = hw_frames_ctx;
486     }
487
488     av_buffer_unref(&ref->hw_frames_ctx);
489     ref->hw_frames_ctx = av_buffer_ref(vtctx->cached_hw_frames_ctx);
490     if (!ref->hw_frames_ctx)
491         return AVERROR(ENOMEM);
492
493     return 0;
494 }
495
496 static void videotoolbox_write_mp4_descr_length(PutByteContext *pb, int length)
497 {
498     int i;
499     uint8_t b;
500
501     for (i = 3; i >= 0; i--) {
502         b = (length >> (i * 7)) & 0x7F;
503         if (i != 0)
504             b |= 0x80;
505
506         bytestream2_put_byteu(pb, b);
507     }
508 }
509
510 static CFDataRef videotoolbox_esds_extradata_create(AVCodecContext *avctx)
511 {
512     CFDataRef data;
513     uint8_t *rw_extradata;
514     PutByteContext pb;
515     int full_size = 3 + 5 + 13 + 5 + avctx->extradata_size + 3;
516     // ES_DescrTag data + DecoderConfigDescrTag + data + DecSpecificInfoTag + size + SLConfigDescriptor
517     int config_size = 13 + 5 + avctx->extradata_size;
518     int s;
519
520     if (!(rw_extradata = av_mallocz(full_size + VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING)))
521         return NULL;
522
523     bytestream2_init_writer(&pb, rw_extradata, full_size + VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING);
524     bytestream2_put_byteu(&pb, 0);        // version
525     bytestream2_put_ne24(&pb, 0);         // flags
526
527     // elementary stream descriptor
528     bytestream2_put_byteu(&pb, 0x03);     // ES_DescrTag
529     videotoolbox_write_mp4_descr_length(&pb, full_size);
530     bytestream2_put_ne16(&pb, 0);         // esid
531     bytestream2_put_byteu(&pb, 0);        // stream priority (0-32)
532
533     // decoder configuration descriptor
534     bytestream2_put_byteu(&pb, 0x04);     // DecoderConfigDescrTag
535     videotoolbox_write_mp4_descr_length(&pb, config_size);
536     bytestream2_put_byteu(&pb, 32);       // object type indication. 32 = AV_CODEC_ID_MPEG4
537     bytestream2_put_byteu(&pb, 0x11);     // stream type
538     bytestream2_put_ne24(&pb, 0);         // buffer size
539     bytestream2_put_ne32(&pb, 0);         // max bitrate
540     bytestream2_put_ne32(&pb, 0);         // avg bitrate
541
542     // decoder specific descriptor
543     bytestream2_put_byteu(&pb, 0x05);     ///< DecSpecificInfoTag
544     videotoolbox_write_mp4_descr_length(&pb, avctx->extradata_size);
545
546     bytestream2_put_buffer(&pb, avctx->extradata, avctx->extradata_size);
547
548     // SLConfigDescriptor
549     bytestream2_put_byteu(&pb, 0x06);     // SLConfigDescrTag
550     bytestream2_put_byteu(&pb, 0x01);     // length
551     bytestream2_put_byteu(&pb, 0x02);     //
552
553     s = bytestream2_size_p(&pb);
554
555     data = CFDataCreate(kCFAllocatorDefault, rw_extradata, s);
556
557     av_freep(&rw_extradata);
558     return data;
559 }
560
561 static CMSampleBufferRef videotoolbox_sample_buffer_create(CMFormatDescriptionRef fmt_desc,
562                                                            void *buffer,
563                                                            int size)
564 {
565     OSStatus status;
566     CMBlockBufferRef  block_buf;
567     CMSampleBufferRef sample_buf;
568
569     block_buf  = NULL;
570     sample_buf = NULL;
571
572     status = CMBlockBufferCreateWithMemoryBlock(kCFAllocatorDefault,// structureAllocator
573                                                 buffer,             // memoryBlock
574                                                 size,               // blockLength
575                                                 kCFAllocatorNull,   // blockAllocator
576                                                 NULL,               // customBlockSource
577                                                 0,                  // offsetToData
578                                                 size,               // dataLength
579                                                 0,                  // flags
580                                                 &block_buf);
581
582     if (!status) {
583         status = CMSampleBufferCreate(kCFAllocatorDefault,  // allocator
584                                       block_buf,            // dataBuffer
585                                       TRUE,                 // dataReady
586                                       0,                    // makeDataReadyCallback
587                                       0,                    // makeDataReadyRefcon
588                                       fmt_desc,             // formatDescription
589                                       1,                    // numSamples
590                                       0,                    // numSampleTimingEntries
591                                       NULL,                 // sampleTimingArray
592                                       0,                    // numSampleSizeEntries
593                                       NULL,                 // sampleSizeArray
594                                       &sample_buf);
595     }
596
597     if (block_buf)
598         CFRelease(block_buf);
599
600     return sample_buf;
601 }
602
603 static void videotoolbox_decoder_callback(void *opaque,
604                                           void *sourceFrameRefCon,
605                                           OSStatus status,
606                                           VTDecodeInfoFlags flags,
607                                           CVImageBufferRef image_buffer,
608                                           CMTime pts,
609                                           CMTime duration)
610 {
611     AVCodecContext *avctx = opaque;
612     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
613
614     if (vtctx->frame) {
615         CVPixelBufferRelease(vtctx->frame);
616         vtctx->frame = NULL;
617     }
618
619     if (!image_buffer) {
620         av_log(NULL, AV_LOG_DEBUG, "vt decoder cb: output image buffer is null\n");
621         return;
622     }
623
624     vtctx->frame = CVPixelBufferRetain(image_buffer);
625 }
626
627 static OSStatus videotoolbox_session_decode_frame(AVCodecContext *avctx)
628 {
629     OSStatus status;
630     CMSampleBufferRef sample_buf;
631     AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
632     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
633
634     sample_buf = videotoolbox_sample_buffer_create(videotoolbox->cm_fmt_desc,
635                                                    vtctx->bitstream,
636                                                    vtctx->bitstream_size);
637
638     if (!sample_buf)
639         return -1;
640
641     status = VTDecompressionSessionDecodeFrame(videotoolbox->session,
642                                                sample_buf,
643                                                0,       // decodeFlags
644                                                NULL,    // sourceFrameRefCon
645                                                0);      // infoFlagsOut
646     if (status == noErr)
647         status = VTDecompressionSessionWaitForAsynchronousFrames(videotoolbox->session);
648
649     CFRelease(sample_buf);
650
651     return status;
652 }
653
654 static CMVideoFormatDescriptionRef videotoolbox_format_desc_create(CMVideoCodecType codec_type,
655                                                                    CFDictionaryRef decoder_spec,
656                                                                    int width,
657                                                                    int height)
658 {
659     CMFormatDescriptionRef cm_fmt_desc;
660     OSStatus status;
661
662     status = CMVideoFormatDescriptionCreate(kCFAllocatorDefault,
663                                             codec_type,
664                                             width,
665                                             height,
666                                             decoder_spec, // Dictionary of extension
667                                             &cm_fmt_desc);
668
669     if (status)
670         return NULL;
671
672     return cm_fmt_desc;
673 }
674
675 static CFDictionaryRef videotoolbox_buffer_attributes_create(int width,
676                                                              int height,
677                                                              OSType pix_fmt)
678 {
679     CFMutableDictionaryRef buffer_attributes;
680     CFMutableDictionaryRef io_surface_properties;
681     CFNumberRef cv_pix_fmt;
682     CFNumberRef w;
683     CFNumberRef h;
684
685     w = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &width);
686     h = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &height);
687     cv_pix_fmt = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &pix_fmt);
688
689     buffer_attributes = CFDictionaryCreateMutable(kCFAllocatorDefault,
690                                                   4,
691                                                   &kCFTypeDictionaryKeyCallBacks,
692                                                   &kCFTypeDictionaryValueCallBacks);
693     io_surface_properties = CFDictionaryCreateMutable(kCFAllocatorDefault,
694                                                       0,
695                                                       &kCFTypeDictionaryKeyCallBacks,
696                                                       &kCFTypeDictionaryValueCallBacks);
697
698     if (pix_fmt)
699         CFDictionarySetValue(buffer_attributes, kCVPixelBufferPixelFormatTypeKey, cv_pix_fmt);
700     CFDictionarySetValue(buffer_attributes, kCVPixelBufferIOSurfacePropertiesKey, io_surface_properties);
701     CFDictionarySetValue(buffer_attributes, kCVPixelBufferWidthKey, w);
702     CFDictionarySetValue(buffer_attributes, kCVPixelBufferHeightKey, h);
703 #if TARGET_OS_IPHONE
704     CFDictionarySetValue(buffer_attributes, kCVPixelBufferOpenGLESCompatibilityKey, kCFBooleanTrue);
705 #else
706     CFDictionarySetValue(buffer_attributes, kCVPixelBufferIOSurfaceOpenGLTextureCompatibilityKey, kCFBooleanTrue);
707 #endif
708
709     CFRelease(io_surface_properties);
710     CFRelease(cv_pix_fmt);
711     CFRelease(w);
712     CFRelease(h);
713
714     return buffer_attributes;
715 }
716
717 static CFDictionaryRef videotoolbox_decoder_config_create(CMVideoCodecType codec_type,
718                                                           AVCodecContext *avctx)
719 {
720     CFMutableDictionaryRef config_info = CFDictionaryCreateMutable(kCFAllocatorDefault,
721                                                                    0,
722                                                                    &kCFTypeDictionaryKeyCallBacks,
723                                                                    &kCFTypeDictionaryValueCallBacks);
724
725     CFDictionarySetValue(config_info,
726                          codec_type == kCMVideoCodecType_HEVC ?
727                             kVTVideoDecoderSpecification_EnableHardwareAcceleratedVideoDecoder :
728                             kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder,
729                          kCFBooleanTrue);
730
731     CFMutableDictionaryRef avc_info;
732     CFDataRef data = NULL;
733
734     avc_info = CFDictionaryCreateMutable(kCFAllocatorDefault,
735                                          1,
736                                          &kCFTypeDictionaryKeyCallBacks,
737                                          &kCFTypeDictionaryValueCallBacks);
738
739     switch (codec_type) {
740     case kCMVideoCodecType_MPEG4Video :
741         if (avctx->extradata_size)
742             data = videotoolbox_esds_extradata_create(avctx);
743         if (data)
744             CFDictionarySetValue(avc_info, CFSTR("esds"), data);
745         break;
746     case kCMVideoCodecType_H264 :
747         data = ff_videotoolbox_avcc_extradata_create(avctx);
748         if (data)
749             CFDictionarySetValue(avc_info, CFSTR("avcC"), data);
750         break;
751     case kCMVideoCodecType_HEVC :
752         data = ff_videotoolbox_hvcc_extradata_create(avctx);
753         if (data)
754             CFDictionarySetValue(avc_info, CFSTR("hvcC"), data);
755         break;
756     default:
757         break;
758     }
759
760     CFDictionarySetValue(config_info,
761             kCMFormatDescriptionExtension_SampleDescriptionExtensionAtoms,
762             avc_info);
763
764     if (data)
765         CFRelease(data);
766
767     CFRelease(avc_info);
768     return config_info;
769 }
770
771 static int videotoolbox_start(AVCodecContext *avctx)
772 {
773     AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
774     OSStatus status;
775     VTDecompressionOutputCallbackRecord decoder_cb;
776     CFDictionaryRef decoder_spec;
777     CFDictionaryRef buf_attr;
778
779     if (!videotoolbox) {
780         av_log(avctx, AV_LOG_ERROR, "hwaccel context is not set\n");
781         return -1;
782     }
783
784     switch( avctx->codec_id ) {
785     case AV_CODEC_ID_H263 :
786         videotoolbox->cm_codec_type = kCMVideoCodecType_H263;
787         break;
788     case AV_CODEC_ID_H264 :
789         videotoolbox->cm_codec_type = kCMVideoCodecType_H264;
790         break;
791     case AV_CODEC_ID_HEVC :
792         videotoolbox->cm_codec_type = kCMVideoCodecType_HEVC;
793         break;
794     case AV_CODEC_ID_MPEG1VIDEO :
795         videotoolbox->cm_codec_type = kCMVideoCodecType_MPEG1Video;
796         break;
797     case AV_CODEC_ID_MPEG2VIDEO :
798         videotoolbox->cm_codec_type = kCMVideoCodecType_MPEG2Video;
799         break;
800     case AV_CODEC_ID_MPEG4 :
801         videotoolbox->cm_codec_type = kCMVideoCodecType_MPEG4Video;
802         break;
803     default :
804         break;
805     }
806
807     decoder_spec = videotoolbox_decoder_config_create(videotoolbox->cm_codec_type, avctx);
808
809     if (!decoder_spec) {
810         av_log(avctx, AV_LOG_ERROR, "decoder specification creation failed\n");
811         return -1;
812     }
813
814     videotoolbox->cm_fmt_desc = videotoolbox_format_desc_create(videotoolbox->cm_codec_type,
815                                                                 decoder_spec,
816                                                                 avctx->width,
817                                                                 avctx->height);
818     if (!videotoolbox->cm_fmt_desc) {
819         if (decoder_spec)
820             CFRelease(decoder_spec);
821
822         av_log(avctx, AV_LOG_ERROR, "format description creation failed\n");
823         return -1;
824     }
825
826     buf_attr = videotoolbox_buffer_attributes_create(avctx->width,
827                                                      avctx->height,
828                                                      videotoolbox->cv_pix_fmt_type);
829
830     decoder_cb.decompressionOutputCallback = videotoolbox_decoder_callback;
831     decoder_cb.decompressionOutputRefCon   = avctx;
832
833     status = VTDecompressionSessionCreate(NULL,                      // allocator
834                                           videotoolbox->cm_fmt_desc, // videoFormatDescription
835                                           decoder_spec,              // videoDecoderSpecification
836                                           buf_attr,                  // destinationImageBufferAttributes
837                                           &decoder_cb,               // outputCallback
838                                           &videotoolbox->session);   // decompressionSessionOut
839
840     if (decoder_spec)
841         CFRelease(decoder_spec);
842     if (buf_attr)
843         CFRelease(buf_attr);
844
845     switch (status) {
846     case kVTVideoDecoderNotAvailableNowErr:
847         av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox session not available.\n");
848         return AVERROR(ENOSYS);
849     case kVTVideoDecoderUnsupportedDataFormatErr:
850         av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox does not support this format.\n");
851         return AVERROR(ENOSYS);
852     case kVTCouldNotFindVideoDecoderErr:
853         av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox decoder for this format not found.\n");
854         return AVERROR(ENOSYS);
855     case kVTVideoDecoderMalfunctionErr:
856         av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox malfunction.\n");
857         return AVERROR(EINVAL);
858     case kVTVideoDecoderBadDataErr:
859         av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox reported invalid data.\n");
860         return AVERROR_INVALIDDATA;
861     case 0:
862         return 0;
863     default:
864         av_log(avctx, AV_LOG_VERBOSE, "Unknown VideoToolbox session creation error %d\n", (int)status);
865         return AVERROR_UNKNOWN;
866     }
867 }
868
869 static void videotoolbox_stop(AVCodecContext *avctx)
870 {
871     AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
872     if (!videotoolbox)
873         return;
874
875     if (videotoolbox->cm_fmt_desc) {
876         CFRelease(videotoolbox->cm_fmt_desc);
877         videotoolbox->cm_fmt_desc = NULL;
878     }
879
880     if (videotoolbox->session) {
881         VTDecompressionSessionInvalidate(videotoolbox->session);
882         CFRelease(videotoolbox->session);
883         videotoolbox->session = NULL;
884     }
885 }
886
887 static const char *videotoolbox_error_string(OSStatus status)
888 {
889     switch (status) {
890         case kVTVideoDecoderBadDataErr:
891             return "bad data";
892         case kVTVideoDecoderMalfunctionErr:
893             return "decoder malfunction";
894         case kVTInvalidSessionErr:
895             return "invalid session";
896     }
897     return "unknown";
898 }
899
900 static int videotoolbox_common_end_frame(AVCodecContext *avctx, AVFrame *frame)
901 {
902     OSStatus status;
903     AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
904     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
905
906     frame->crop_right = 0;
907     frame->crop_left = 0;
908     frame->crop_top = 0;
909     frame->crop_bottom = 0;
910
911     if (vtctx->reconfig_needed == true) {
912         vtctx->reconfig_needed = false;
913         av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox decoder needs reconfig, restarting..\n");
914         videotoolbox_stop(avctx);
915         if (videotoolbox_start(avctx) != 0) {
916             return AVERROR_EXTERNAL;
917         }
918     }
919
920     if (!videotoolbox->session || !vtctx->bitstream || !vtctx->bitstream_size)
921         return AVERROR_INVALIDDATA;
922
923     status = videotoolbox_session_decode_frame(avctx);
924     if (status != noErr) {
925         if (status == kVTVideoDecoderMalfunctionErr || status == kVTInvalidSessionErr)
926             vtctx->reconfig_needed = true;
927         av_log(avctx, AV_LOG_ERROR, "Failed to decode frame (%s, %d)\n", videotoolbox_error_string(status), (int)status);
928         return AVERROR_UNKNOWN;
929     }
930
931     if (!vtctx->frame) {
932         vtctx->reconfig_needed = true;
933         return AVERROR_UNKNOWN;
934     }
935
936     return videotoolbox_buffer_create(avctx, frame);
937 }
938
939 static int videotoolbox_h264_end_frame(AVCodecContext *avctx)
940 {
941     H264Context *h = avctx->priv_data;
942     AVFrame *frame = h->cur_pic_ptr->f;
943     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
944     int ret = videotoolbox_common_end_frame(avctx, frame);
945     vtctx->bitstream_size = 0;
946     return ret;
947 }
948
949 static int videotoolbox_hevc_start_frame(AVCodecContext *avctx,
950                                          const uint8_t *buffer,
951                                          uint32_t size)
952 {
953     return 0;
954 }
955
956 static int videotoolbox_hevc_decode_slice(AVCodecContext *avctx,
957                                           const uint8_t *buffer,
958                                           uint32_t size)
959 {
960     return videotoolbox_common_decode_slice(avctx, buffer, size);
961 }
962
963
964 static int videotoolbox_hevc_decode_params(AVCodecContext *avctx,
965                                            int type,
966                                            const uint8_t *buffer,
967                                            uint32_t size)
968 {
969     return videotoolbox_common_decode_slice(avctx, buffer, size);
970 }
971
972 static int videotoolbox_hevc_end_frame(AVCodecContext *avctx)
973 {
974     HEVCContext *h = avctx->priv_data;
975     AVFrame *frame = h->ref->frame;
976     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
977
978     h->output_frame->crop_right = 0;
979     h->output_frame->crop_left = 0;
980     h->output_frame->crop_top = 0;
981     h->output_frame->crop_bottom = 0;
982
983     int ret = videotoolbox_common_end_frame(avctx, frame);
984     vtctx->bitstream_size = 0;
985     return ret;
986 }
987
988 static int videotoolbox_mpeg_start_frame(AVCodecContext *avctx,
989                                          const uint8_t *buffer,
990                                          uint32_t size)
991 {
992     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
993
994     return videotoolbox_buffer_copy(vtctx, buffer, size);
995 }
996
997 static int videotoolbox_mpeg_decode_slice(AVCodecContext *avctx,
998                                           const uint8_t *buffer,
999                                           uint32_t size)
1000 {
1001     return 0;
1002 }
1003
1004 static int videotoolbox_mpeg_end_frame(AVCodecContext *avctx)
1005 {
1006     MpegEncContext *s = avctx->priv_data;
1007     AVFrame *frame = s->current_picture_ptr->f;
1008
1009     return videotoolbox_common_end_frame(avctx, frame);
1010 }
1011
1012 static int videotoolbox_uninit(AVCodecContext *avctx)
1013 {
1014     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
1015     if (!vtctx)
1016         return 0;
1017
1018     ff_videotoolbox_uninit(avctx);
1019
1020     if (vtctx->vt_ctx)
1021         videotoolbox_stop(avctx);
1022
1023     av_buffer_unref(&vtctx->cached_hw_frames_ctx);
1024     av_freep(&vtctx->vt_ctx);
1025
1026     return 0;
1027 }
1028
1029 static int videotoolbox_common_init(AVCodecContext *avctx)
1030 {
1031     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
1032     AVHWFramesContext *hw_frames;
1033     int err;
1034
1035     // Old API - do nothing.
1036     if (avctx->hwaccel_context)
1037         return 0;
1038
1039     if (!avctx->hw_frames_ctx && !avctx->hw_device_ctx) {
1040         av_log(avctx, AV_LOG_ERROR,
1041                "Either hw_frames_ctx or hw_device_ctx must be set.\n");
1042         return AVERROR(EINVAL);
1043     }
1044
1045     vtctx->vt_ctx = av_videotoolbox_alloc_context();
1046     if (!vtctx->vt_ctx) {
1047         err = AVERROR(ENOMEM);
1048         goto fail;
1049     }
1050
1051     if (avctx->hw_frames_ctx) {
1052         hw_frames = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
1053     } else {
1054         avctx->hw_frames_ctx = av_hwframe_ctx_alloc(avctx->hw_device_ctx);
1055         if (!avctx->hw_frames_ctx) {
1056             err = AVERROR(ENOMEM);
1057             goto fail;
1058         }
1059
1060         hw_frames = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
1061         hw_frames->format = AV_PIX_FMT_VIDEOTOOLBOX;
1062         hw_frames->sw_format = AV_PIX_FMT_NV12; // same as av_videotoolbox_alloc_context()
1063         hw_frames->width = avctx->width;
1064         hw_frames->height = avctx->height;
1065
1066         err = av_hwframe_ctx_init(avctx->hw_frames_ctx);
1067         if (err < 0) {
1068             av_buffer_unref(&avctx->hw_frames_ctx);
1069             goto fail;
1070         }
1071     }
1072
1073     vtctx->cached_hw_frames_ctx = av_buffer_ref(avctx->hw_frames_ctx);
1074     if (!vtctx->cached_hw_frames_ctx) {
1075         err = AVERROR(ENOMEM);
1076         goto fail;
1077     }
1078
1079     vtctx->vt_ctx->cv_pix_fmt_type =
1080         av_map_videotoolbox_format_from_pixfmt(hw_frames->sw_format);
1081     if (!vtctx->vt_ctx->cv_pix_fmt_type) {
1082         av_log(avctx, AV_LOG_ERROR, "Unknown sw_format.\n");
1083         err = AVERROR(EINVAL);
1084         goto fail;
1085     }
1086
1087     err = videotoolbox_start(avctx);
1088     if (err < 0)
1089         goto fail;
1090
1091     return 0;
1092
1093 fail:
1094     videotoolbox_uninit(avctx);
1095     return err;
1096 }
1097
1098 static int videotoolbox_frame_params(AVCodecContext *avctx,
1099                                      AVBufferRef *hw_frames_ctx)
1100 {
1101     AVHWFramesContext *frames_ctx = (AVHWFramesContext*)hw_frames_ctx->data;
1102
1103     frames_ctx->format            = AV_PIX_FMT_VIDEOTOOLBOX;
1104     frames_ctx->width             = avctx->coded_width;
1105     frames_ctx->height            = avctx->coded_height;
1106     frames_ctx->sw_format         = AV_PIX_FMT_NV12;
1107
1108     return 0;
1109 }
1110
1111 const AVHWAccel ff_h263_videotoolbox_hwaccel = {
1112     .name           = "h263_videotoolbox",
1113     .type           = AVMEDIA_TYPE_VIDEO,
1114     .id             = AV_CODEC_ID_H263,
1115     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1116     .alloc_frame    = ff_videotoolbox_alloc_frame,
1117     .start_frame    = videotoolbox_mpeg_start_frame,
1118     .decode_slice   = videotoolbox_mpeg_decode_slice,
1119     .end_frame      = videotoolbox_mpeg_end_frame,
1120     .frame_params   = videotoolbox_frame_params,
1121     .init           = videotoolbox_common_init,
1122     .uninit         = videotoolbox_uninit,
1123     .priv_data_size = sizeof(VTContext),
1124 };
1125
1126 const AVHWAccel ff_hevc_videotoolbox_hwaccel = {
1127     .name           = "hevc_videotoolbox",
1128     .type           = AVMEDIA_TYPE_VIDEO,
1129     .id             = AV_CODEC_ID_HEVC,
1130     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1131     .alloc_frame    = ff_videotoolbox_alloc_frame,
1132     .start_frame    = videotoolbox_hevc_start_frame,
1133     .decode_slice   = videotoolbox_hevc_decode_slice,
1134     .decode_params  = videotoolbox_hevc_decode_params,
1135     .end_frame      = videotoolbox_hevc_end_frame,
1136     .frame_params   = videotoolbox_frame_params,
1137     .init           = videotoolbox_common_init,
1138     .uninit         = ff_videotoolbox_uninit,
1139     .priv_data_size = sizeof(VTContext),
1140 };
1141
1142 const AVHWAccel ff_h264_videotoolbox_hwaccel = {
1143     .name           = "h264_videotoolbox",
1144     .type           = AVMEDIA_TYPE_VIDEO,
1145     .id             = AV_CODEC_ID_H264,
1146     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1147     .alloc_frame    = ff_videotoolbox_alloc_frame,
1148     .start_frame    = ff_videotoolbox_h264_start_frame,
1149     .decode_slice   = ff_videotoolbox_h264_decode_slice,
1150     .decode_params  = videotoolbox_h264_decode_params,
1151     .end_frame      = videotoolbox_h264_end_frame,
1152     .frame_params   = videotoolbox_frame_params,
1153     .init           = videotoolbox_common_init,
1154     .uninit         = videotoolbox_uninit,
1155     .priv_data_size = sizeof(VTContext),
1156 };
1157
1158 const AVHWAccel ff_mpeg1_videotoolbox_hwaccel = {
1159     .name           = "mpeg1_videotoolbox",
1160     .type           = AVMEDIA_TYPE_VIDEO,
1161     .id             = AV_CODEC_ID_MPEG1VIDEO,
1162     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1163     .alloc_frame    = ff_videotoolbox_alloc_frame,
1164     .start_frame    = videotoolbox_mpeg_start_frame,
1165     .decode_slice   = videotoolbox_mpeg_decode_slice,
1166     .end_frame      = videotoolbox_mpeg_end_frame,
1167     .frame_params   = videotoolbox_frame_params,
1168     .init           = videotoolbox_common_init,
1169     .uninit         = videotoolbox_uninit,
1170     .priv_data_size = sizeof(VTContext),
1171 };
1172
1173 const AVHWAccel ff_mpeg2_videotoolbox_hwaccel = {
1174     .name           = "mpeg2_videotoolbox",
1175     .type           = AVMEDIA_TYPE_VIDEO,
1176     .id             = AV_CODEC_ID_MPEG2VIDEO,
1177     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1178     .alloc_frame    = ff_videotoolbox_alloc_frame,
1179     .start_frame    = videotoolbox_mpeg_start_frame,
1180     .decode_slice   = videotoolbox_mpeg_decode_slice,
1181     .end_frame      = videotoolbox_mpeg_end_frame,
1182     .frame_params   = videotoolbox_frame_params,
1183     .init           = videotoolbox_common_init,
1184     .uninit         = videotoolbox_uninit,
1185     .priv_data_size = sizeof(VTContext),
1186 };
1187
1188 const AVHWAccel ff_mpeg4_videotoolbox_hwaccel = {
1189     .name           = "mpeg4_videotoolbox",
1190     .type           = AVMEDIA_TYPE_VIDEO,
1191     .id             = AV_CODEC_ID_MPEG4,
1192     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1193     .alloc_frame    = ff_videotoolbox_alloc_frame,
1194     .start_frame    = videotoolbox_mpeg_start_frame,
1195     .decode_slice   = videotoolbox_mpeg_decode_slice,
1196     .end_frame      = videotoolbox_mpeg_end_frame,
1197     .frame_params   = videotoolbox_frame_params,
1198     .init           = videotoolbox_common_init,
1199     .uninit         = videotoolbox_uninit,
1200     .priv_data_size = sizeof(VTContext),
1201 };
1202
1203 AVVideotoolboxContext *av_videotoolbox_alloc_context(void)
1204 {
1205     AVVideotoolboxContext *ret = av_mallocz(sizeof(*ret));
1206
1207     if (ret) {
1208         ret->output_callback = videotoolbox_decoder_callback;
1209         ret->cv_pix_fmt_type = kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange;
1210     }
1211
1212     return ret;
1213 }
1214
1215 int av_videotoolbox_default_init(AVCodecContext *avctx)
1216 {
1217     return av_videotoolbox_default_init2(avctx, NULL);
1218 }
1219
1220 int av_videotoolbox_default_init2(AVCodecContext *avctx, AVVideotoolboxContext *vtctx)
1221 {
1222     avctx->hwaccel_context = vtctx ?: av_videotoolbox_alloc_context();
1223     if (!avctx->hwaccel_context)
1224         return AVERROR(ENOMEM);
1225     return videotoolbox_start(avctx);
1226 }
1227
1228 void av_videotoolbox_default_free(AVCodecContext *avctx)
1229 {
1230
1231     videotoolbox_stop(avctx);
1232     av_freep(&avctx->hwaccel_context);
1233 }
1234 #endif /* CONFIG_VIDEOTOOLBOX */