8773de33930572ead9d3a0fdf92647be14301033
[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 "libavutil/pixdesc.h"
30 #include "bytestream.h"
31 #include "decode.h"
32 #include "h264dec.h"
33 #include "hevcdec.h"
34 #include "mpegvideo.h"
35 #include <TargetConditionals.h>
36
37 #ifndef kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder
38 #  define kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder CFSTR("RequireHardwareAcceleratedVideoDecoder")
39 #endif
40 #ifndef kVTVideoDecoderSpecification_EnableHardwareAcceleratedVideoDecoder
41 #  define kVTVideoDecoderSpecification_EnableHardwareAcceleratedVideoDecoder CFSTR("EnableHardwareAcceleratedVideoDecoder")
42 #endif
43
44 #if !HAVE_KCMVIDEOCODECTYPE_HEVC
45 enum { kCMVideoCodecType_HEVC = 'hvc1' };
46 #endif
47
48 #define VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING  12
49
50 typedef struct VTHWFrame {
51     CVPixelBufferRef pixbuf;
52     AVBufferRef *hw_frames_ctx;
53 } VTHWFrame;
54
55 static void videotoolbox_buffer_release(void *opaque, uint8_t *data)
56 {
57     VTHWFrame *ref = (VTHWFrame *)data;
58     av_buffer_unref(&ref->hw_frames_ctx);
59     CVPixelBufferRelease(ref->pixbuf);
60
61     av_free(data);
62 }
63
64 static int videotoolbox_buffer_copy(VTContext *vtctx,
65                                     const uint8_t *buffer,
66                                     uint32_t size)
67 {
68     void *tmp;
69
70     tmp = av_fast_realloc(vtctx->bitstream,
71                          &vtctx->allocated_size,
72                          size);
73
74     if (!tmp)
75         return AVERROR(ENOMEM);
76
77     vtctx->bitstream = tmp;
78     memcpy(vtctx->bitstream, buffer, size);
79     vtctx->bitstream_size = size;
80
81     return 0;
82 }
83
84 static int videotoolbox_postproc_frame(void *avctx, AVFrame *frame)
85 {
86     VTHWFrame *ref = (VTHWFrame *)frame->buf[0]->data;
87
88     if (!ref->pixbuf) {
89         av_log(avctx, AV_LOG_ERROR, "No frame decoded?\n");
90         av_frame_unref(frame);
91         return AVERROR_EXTERNAL;
92     }
93
94     frame->crop_right = 0;
95     frame->crop_left = 0;
96     frame->crop_top = 0;
97     frame->crop_bottom = 0;
98
99     frame->data[3] = (uint8_t*)ref->pixbuf;
100
101     if (ref->hw_frames_ctx) {
102         av_buffer_unref(&frame->hw_frames_ctx);
103         frame->hw_frames_ctx = av_buffer_ref(ref->hw_frames_ctx);
104         if (!frame->hw_frames_ctx)
105             return AVERROR(ENOMEM);
106     }
107
108     return 0;
109 }
110
111 int ff_videotoolbox_alloc_frame(AVCodecContext *avctx, AVFrame *frame)
112 {
113     size_t      size = sizeof(VTHWFrame);
114     uint8_t    *data = NULL;
115     AVBufferRef *buf = NULL;
116     int ret = ff_attach_decode_data(frame);
117     FrameDecodeData *fdd;
118     if (ret < 0)
119         return ret;
120
121     data = av_mallocz(size);
122     if (!data)
123         return AVERROR(ENOMEM);
124     buf = av_buffer_create(data, size, videotoolbox_buffer_release, NULL, 0);
125     if (!buf) {
126         av_freep(&data);
127         return AVERROR(ENOMEM);
128     }
129     frame->buf[0] = buf;
130
131     fdd = (FrameDecodeData*)frame->private_ref->data;
132     fdd->post_process = videotoolbox_postproc_frame;
133
134     frame->width  = avctx->width;
135     frame->height = avctx->height;
136     frame->format = avctx->pix_fmt;
137
138     return 0;
139 }
140
141 #define AV_W8(p, v) *(p) = (v)
142
143 CFDataRef ff_videotoolbox_avcc_extradata_create(AVCodecContext *avctx)
144 {
145     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
146     H264Context *h = avctx->priv_data;
147     CFDataRef data = NULL;
148     uint8_t *p;
149     int vt_extradata_size = 6 + 2 + h->ps.sps->data_size + 3 + h->ps.pps->data_size;
150     uint8_t *vt_extradata = av_malloc(vt_extradata_size);
151     if (!vt_extradata)
152         return NULL;
153
154     p = vt_extradata;
155
156     AV_W8(p + 0, 1); /* version */
157     AV_W8(p + 1, h->ps.sps->data[1]); /* profile */
158     AV_W8(p + 2, h->ps.sps->data[2]); /* profile compat */
159     AV_W8(p + 3, h->ps.sps->data[3]); /* level */
160     AV_W8(p + 4, 0xff); /* 6 bits reserved (111111) + 2 bits nal size length - 3 (11) */
161     AV_W8(p + 5, 0xe1); /* 3 bits reserved (111) + 5 bits number of sps (00001) */
162     AV_WB16(p + 6, h->ps.sps->data_size);
163     memcpy(p + 8, h->ps.sps->data, h->ps.sps->data_size);
164     p += 8 + h->ps.sps->data_size;
165     AV_W8(p + 0, 1); /* number of pps */
166     AV_WB16(p + 1, h->ps.pps->data_size);
167     memcpy(p + 3, h->ps.pps->data, h->ps.pps->data_size);
168
169     p += 3 + h->ps.pps->data_size;
170     av_assert0(p - vt_extradata == vt_extradata_size);
171
172     // save sps header (profile/level) used to create decoder session,
173     // so we can detect changes and recreate it.
174     if (vtctx)
175         memcpy(vtctx->sps, h->ps.sps->data + 1, 3);
176
177     data = CFDataCreate(kCFAllocatorDefault, vt_extradata, vt_extradata_size);
178     av_free(vt_extradata);
179     return data;
180 }
181
182 CFDataRef ff_videotoolbox_hvcc_extradata_create(AVCodecContext *avctx)
183 {
184     HEVCContext *h = avctx->priv_data;
185     int i, num_vps = 0, num_sps = 0, num_pps = 0;
186     const HEVCVPS *vps = h->ps.vps;
187     const HEVCSPS *sps = h->ps.sps;
188     const HEVCPPS *pps = h->ps.pps;
189     PTLCommon ptlc = vps->ptl.general_ptl;
190     VUI vui = sps->vui;
191     uint8_t parallelismType;
192     CFDataRef data = NULL;
193     uint8_t *p;
194     int vt_extradata_size = 23 + 3 + 3 + 3;
195     uint8_t *vt_extradata;
196
197 #define COUNT_SIZE_PS(T, t) \
198     for (i = 0; i < HEVC_MAX_##T##PS_COUNT; i++) { \
199         if (h->ps.t##ps_list[i]) { \
200             const HEVC##T##PS *lps = (const HEVC##T##PS *)h->ps.t##ps_list[i]->data; \
201             vt_extradata_size += 2 + lps->data_size; \
202             num_##t##ps++; \
203         } \
204     }
205
206     COUNT_SIZE_PS(V, v)
207     COUNT_SIZE_PS(S, s)
208     COUNT_SIZE_PS(P, p)
209
210     vt_extradata = av_malloc(vt_extradata_size);
211     if (!vt_extradata)
212         return NULL;
213     p = vt_extradata;
214
215     /* unsigned int(8) configurationVersion = 1; */
216     AV_W8(p + 0, 1);
217
218     /*
219      * unsigned int(2) general_profile_space;
220      * unsigned int(1) general_tier_flag;
221      * unsigned int(5) general_profile_idc;
222      */
223     AV_W8(p + 1, ptlc.profile_space << 6 |
224                  ptlc.tier_flag     << 5 |
225                  ptlc.profile_idc);
226
227     /* unsigned int(32) general_profile_compatibility_flags; */
228     memcpy(p + 2, ptlc.profile_compatibility_flag, 4);
229
230     /* unsigned int(48) general_constraint_indicator_flags; */
231     AV_W8(p + 6, ptlc.progressive_source_flag    << 7 |
232                  ptlc.interlaced_source_flag     << 6 |
233                  ptlc.non_packed_constraint_flag << 5 |
234                  ptlc.frame_only_constraint_flag << 4);
235     AV_W8(p + 7, 0);
236     AV_WN32(p + 8, 0);
237
238     /* unsigned int(8) general_level_idc; */
239     AV_W8(p + 12, ptlc.level_idc);
240
241     /*
242      * bit(4) reserved = ‘1111’b;
243      * unsigned int(12) min_spatial_segmentation_idc;
244      */
245     AV_W8(p + 13, 0xf0 | (vui.min_spatial_segmentation_idc >> 4));
246     AV_W8(p + 14, vui.min_spatial_segmentation_idc & 0xff);
247
248     /*
249      * bit(6) reserved = ‘111111’b;
250      * unsigned int(2) parallelismType;
251      */
252     if (!vui.min_spatial_segmentation_idc)
253         parallelismType = 0;
254     else if (pps->entropy_coding_sync_enabled_flag && pps->tiles_enabled_flag)
255         parallelismType = 0;
256     else if (pps->entropy_coding_sync_enabled_flag)
257         parallelismType = 3;
258     else if (pps->tiles_enabled_flag)
259         parallelismType = 2;
260     else
261         parallelismType = 1;
262     AV_W8(p + 15, 0xfc | parallelismType);
263
264     /*
265      * bit(6) reserved = ‘111111’b;
266      * unsigned int(2) chromaFormat;
267      */
268     AV_W8(p + 16, sps->chroma_format_idc | 0xfc);
269
270     /*
271      * bit(5) reserved = ‘11111’b;
272      * unsigned int(3) bitDepthLumaMinus8;
273      */
274     AV_W8(p + 17, (sps->bit_depth - 8) | 0xfc);
275
276     /*
277      * bit(5) reserved = ‘11111’b;
278      * unsigned int(3) bitDepthChromaMinus8;
279      */
280     AV_W8(p + 18, (sps->bit_depth_chroma - 8) | 0xfc);
281
282     /* bit(16) avgFrameRate; */
283     AV_WB16(p + 19, 0);
284
285     /*
286      * bit(2) constantFrameRate;
287      * bit(3) numTemporalLayers;
288      * bit(1) temporalIdNested;
289      * unsigned int(2) lengthSizeMinusOne;
290      */
291     AV_W8(p + 21, 0                             << 6 |
292                   sps->max_sub_layers           << 3 |
293                   sps->temporal_id_nesting_flag << 2 |
294                   3);
295
296     /* unsigned int(8) numOfArrays; */
297     AV_W8(p + 22, 3);
298
299     p += 23;
300
301 #define APPEND_PS(T, t) \
302     /* \
303      * bit(1) array_completeness; \
304      * unsigned int(1) reserved = 0; \
305      * unsigned int(6) NAL_unit_type; \
306      */ \
307     AV_W8(p, 1 << 7 | \
308              HEVC_NAL_##T##PS & 0x3f); \
309     /* unsigned int(16) numNalus; */ \
310     AV_WB16(p + 1, num_##t##ps); \
311     p += 3; \
312     for (i = 0; i < HEVC_MAX_##T##PS_COUNT; i++) { \
313         if (h->ps.t##ps_list[i]) { \
314             const HEVC##T##PS *lps = (const HEVC##T##PS *)h->ps.t##ps_list[i]->data; \
315             /* unsigned int(16) nalUnitLength; */ \
316             AV_WB16(p, lps->data_size); \
317             /* bit(8*nalUnitLength) nalUnit; */ \
318             memcpy(p + 2, lps->data, lps->data_size); \
319             p += 2 + lps->data_size; \
320         } \
321     }
322
323     APPEND_PS(V, v)
324     APPEND_PS(S, s)
325     APPEND_PS(P, p)
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(avctx, 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     if (vtctx->reconfig_needed == true) {
907         vtctx->reconfig_needed = false;
908         av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox decoder needs reconfig, restarting..\n");
909         videotoolbox_stop(avctx);
910         if (videotoolbox_start(avctx) != 0) {
911             return AVERROR_EXTERNAL;
912         }
913     }
914
915     if (!videotoolbox->session || !vtctx->bitstream || !vtctx->bitstream_size)
916         return AVERROR_INVALIDDATA;
917
918     status = videotoolbox_session_decode_frame(avctx);
919     if (status != noErr) {
920         if (status == kVTVideoDecoderMalfunctionErr || status == kVTInvalidSessionErr)
921             vtctx->reconfig_needed = true;
922         av_log(avctx, AV_LOG_ERROR, "Failed to decode frame (%s, %d)\n", videotoolbox_error_string(status), (int)status);
923         return AVERROR_UNKNOWN;
924     }
925
926     if (!vtctx->frame) {
927         vtctx->reconfig_needed = true;
928         return AVERROR_UNKNOWN;
929     }
930
931     return videotoolbox_buffer_create(avctx, frame);
932 }
933
934 static int videotoolbox_h264_end_frame(AVCodecContext *avctx)
935 {
936     H264Context *h = avctx->priv_data;
937     AVFrame *frame = h->cur_pic_ptr->f;
938     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
939     int ret = videotoolbox_common_end_frame(avctx, frame);
940     vtctx->bitstream_size = 0;
941     return ret;
942 }
943
944 static int videotoolbox_hevc_start_frame(AVCodecContext *avctx,
945                                          const uint8_t *buffer,
946                                          uint32_t size)
947 {
948     return 0;
949 }
950
951 static int videotoolbox_hevc_decode_slice(AVCodecContext *avctx,
952                                           const uint8_t *buffer,
953                                           uint32_t size)
954 {
955     return videotoolbox_common_decode_slice(avctx, buffer, size);
956 }
957
958
959 static int videotoolbox_hevc_decode_params(AVCodecContext *avctx,
960                                            int type,
961                                            const uint8_t *buffer,
962                                            uint32_t size)
963 {
964     return videotoolbox_common_decode_slice(avctx, buffer, size);
965 }
966
967 static int videotoolbox_hevc_end_frame(AVCodecContext *avctx)
968 {
969     HEVCContext *h = avctx->priv_data;
970     AVFrame *frame = h->ref->frame;
971     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
972
973     h->output_frame->crop_right = 0;
974     h->output_frame->crop_left = 0;
975     h->output_frame->crop_top = 0;
976     h->output_frame->crop_bottom = 0;
977
978     int ret = videotoolbox_common_end_frame(avctx, frame);
979     vtctx->bitstream_size = 0;
980     return ret;
981 }
982
983 static int videotoolbox_mpeg_start_frame(AVCodecContext *avctx,
984                                          const uint8_t *buffer,
985                                          uint32_t size)
986 {
987     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
988
989     return videotoolbox_buffer_copy(vtctx, buffer, size);
990 }
991
992 static int videotoolbox_mpeg_decode_slice(AVCodecContext *avctx,
993                                           const uint8_t *buffer,
994                                           uint32_t size)
995 {
996     return 0;
997 }
998
999 static int videotoolbox_mpeg_end_frame(AVCodecContext *avctx)
1000 {
1001     MpegEncContext *s = avctx->priv_data;
1002     AVFrame *frame = s->current_picture_ptr->f;
1003
1004     return videotoolbox_common_end_frame(avctx, frame);
1005 }
1006
1007 static int videotoolbox_uninit(AVCodecContext *avctx)
1008 {
1009     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
1010     if (!vtctx)
1011         return 0;
1012
1013     ff_videotoolbox_uninit(avctx);
1014
1015     if (vtctx->vt_ctx)
1016         videotoolbox_stop(avctx);
1017
1018     av_buffer_unref(&vtctx->cached_hw_frames_ctx);
1019     av_freep(&vtctx->vt_ctx);
1020
1021     return 0;
1022 }
1023
1024 static enum AVPixelFormat videotoolbox_best_pixel_format(AVCodecContext *avctx) {
1025     const AVPixFmtDescriptor *descriptor = av_pix_fmt_desc_get(avctx->pix_fmt);
1026     if (!descriptor)
1027         return AV_PIX_FMT_NV12; // same as av_videotoolbox_alloc_context()
1028
1029     int depth = descriptor->comp[0].depth;
1030     if (depth > 8) {
1031         return AV_PIX_FMT_P010;
1032     }
1033
1034     return AV_PIX_FMT_NV12;
1035 }
1036
1037 static int videotoolbox_common_init(AVCodecContext *avctx)
1038 {
1039     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
1040     AVHWFramesContext *hw_frames;
1041     int err;
1042
1043     // Old API - do nothing.
1044     if (avctx->hwaccel_context)
1045         return 0;
1046
1047     if (!avctx->hw_frames_ctx && !avctx->hw_device_ctx) {
1048         av_log(avctx, AV_LOG_ERROR,
1049                "Either hw_frames_ctx or hw_device_ctx must be set.\n");
1050         return AVERROR(EINVAL);
1051     }
1052
1053     vtctx->vt_ctx = av_videotoolbox_alloc_context();
1054     if (!vtctx->vt_ctx) {
1055         err = AVERROR(ENOMEM);
1056         goto fail;
1057     }
1058
1059     if (avctx->hw_frames_ctx) {
1060         hw_frames = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
1061     } else {
1062         avctx->hw_frames_ctx = av_hwframe_ctx_alloc(avctx->hw_device_ctx);
1063         if (!avctx->hw_frames_ctx) {
1064             err = AVERROR(ENOMEM);
1065             goto fail;
1066         }
1067
1068         hw_frames = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
1069         hw_frames->format = AV_PIX_FMT_VIDEOTOOLBOX;
1070         hw_frames->sw_format = videotoolbox_best_pixel_format(avctx);
1071         hw_frames->width = avctx->width;
1072         hw_frames->height = avctx->height;
1073
1074         err = av_hwframe_ctx_init(avctx->hw_frames_ctx);
1075         if (err < 0) {
1076             av_buffer_unref(&avctx->hw_frames_ctx);
1077             goto fail;
1078         }
1079     }
1080
1081     vtctx->cached_hw_frames_ctx = av_buffer_ref(avctx->hw_frames_ctx);
1082     if (!vtctx->cached_hw_frames_ctx) {
1083         err = AVERROR(ENOMEM);
1084         goto fail;
1085     }
1086
1087     bool full_range = avctx->color_range == AVCOL_RANGE_JPEG;
1088     vtctx->vt_ctx->cv_pix_fmt_type =
1089         av_map_videotoolbox_format_from_pixfmt2(hw_frames->sw_format, full_range);
1090     if (!vtctx->vt_ctx->cv_pix_fmt_type) {
1091         av_log(avctx, AV_LOG_ERROR, "Unknown sw_format.\n");
1092         err = AVERROR(EINVAL);
1093         goto fail;
1094     }
1095
1096     err = videotoolbox_start(avctx);
1097     if (err < 0)
1098         goto fail;
1099
1100     return 0;
1101
1102 fail:
1103     videotoolbox_uninit(avctx);
1104     return err;
1105 }
1106
1107 static int videotoolbox_frame_params(AVCodecContext *avctx,
1108                                      AVBufferRef *hw_frames_ctx)
1109 {
1110     AVHWFramesContext *frames_ctx = (AVHWFramesContext*)hw_frames_ctx->data;
1111
1112     frames_ctx->format            = AV_PIX_FMT_VIDEOTOOLBOX;
1113     frames_ctx->width             = avctx->coded_width;
1114     frames_ctx->height            = avctx->coded_height;
1115     frames_ctx->sw_format         = videotoolbox_best_pixel_format(avctx);
1116
1117     return 0;
1118 }
1119
1120 const AVHWAccel ff_h263_videotoolbox_hwaccel = {
1121     .name           = "h263_videotoolbox",
1122     .type           = AVMEDIA_TYPE_VIDEO,
1123     .id             = AV_CODEC_ID_H263,
1124     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1125     .alloc_frame    = ff_videotoolbox_alloc_frame,
1126     .start_frame    = videotoolbox_mpeg_start_frame,
1127     .decode_slice   = videotoolbox_mpeg_decode_slice,
1128     .end_frame      = videotoolbox_mpeg_end_frame,
1129     .frame_params   = videotoolbox_frame_params,
1130     .init           = videotoolbox_common_init,
1131     .uninit         = videotoolbox_uninit,
1132     .priv_data_size = sizeof(VTContext),
1133 };
1134
1135 const AVHWAccel ff_hevc_videotoolbox_hwaccel = {
1136     .name           = "hevc_videotoolbox",
1137     .type           = AVMEDIA_TYPE_VIDEO,
1138     .id             = AV_CODEC_ID_HEVC,
1139     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1140     .alloc_frame    = ff_videotoolbox_alloc_frame,
1141     .start_frame    = videotoolbox_hevc_start_frame,
1142     .decode_slice   = videotoolbox_hevc_decode_slice,
1143     .decode_params  = videotoolbox_hevc_decode_params,
1144     .end_frame      = videotoolbox_hevc_end_frame,
1145     .frame_params   = videotoolbox_frame_params,
1146     .init           = videotoolbox_common_init,
1147     .uninit         = videotoolbox_uninit,
1148     .priv_data_size = sizeof(VTContext),
1149 };
1150
1151 const AVHWAccel ff_h264_videotoolbox_hwaccel = {
1152     .name           = "h264_videotoolbox",
1153     .type           = AVMEDIA_TYPE_VIDEO,
1154     .id             = AV_CODEC_ID_H264,
1155     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1156     .alloc_frame    = ff_videotoolbox_alloc_frame,
1157     .start_frame    = ff_videotoolbox_h264_start_frame,
1158     .decode_slice   = ff_videotoolbox_h264_decode_slice,
1159     .decode_params  = videotoolbox_h264_decode_params,
1160     .end_frame      = videotoolbox_h264_end_frame,
1161     .frame_params   = videotoolbox_frame_params,
1162     .init           = videotoolbox_common_init,
1163     .uninit         = videotoolbox_uninit,
1164     .priv_data_size = sizeof(VTContext),
1165 };
1166
1167 const AVHWAccel ff_mpeg1_videotoolbox_hwaccel = {
1168     .name           = "mpeg1_videotoolbox",
1169     .type           = AVMEDIA_TYPE_VIDEO,
1170     .id             = AV_CODEC_ID_MPEG1VIDEO,
1171     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1172     .alloc_frame    = ff_videotoolbox_alloc_frame,
1173     .start_frame    = videotoolbox_mpeg_start_frame,
1174     .decode_slice   = videotoolbox_mpeg_decode_slice,
1175     .end_frame      = videotoolbox_mpeg_end_frame,
1176     .frame_params   = videotoolbox_frame_params,
1177     .init           = videotoolbox_common_init,
1178     .uninit         = videotoolbox_uninit,
1179     .priv_data_size = sizeof(VTContext),
1180 };
1181
1182 const AVHWAccel ff_mpeg2_videotoolbox_hwaccel = {
1183     .name           = "mpeg2_videotoolbox",
1184     .type           = AVMEDIA_TYPE_VIDEO,
1185     .id             = AV_CODEC_ID_MPEG2VIDEO,
1186     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1187     .alloc_frame    = ff_videotoolbox_alloc_frame,
1188     .start_frame    = videotoolbox_mpeg_start_frame,
1189     .decode_slice   = videotoolbox_mpeg_decode_slice,
1190     .end_frame      = videotoolbox_mpeg_end_frame,
1191     .frame_params   = videotoolbox_frame_params,
1192     .init           = videotoolbox_common_init,
1193     .uninit         = videotoolbox_uninit,
1194     .priv_data_size = sizeof(VTContext),
1195 };
1196
1197 const AVHWAccel ff_mpeg4_videotoolbox_hwaccel = {
1198     .name           = "mpeg4_videotoolbox",
1199     .type           = AVMEDIA_TYPE_VIDEO,
1200     .id             = AV_CODEC_ID_MPEG4,
1201     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
1202     .alloc_frame    = ff_videotoolbox_alloc_frame,
1203     .start_frame    = videotoolbox_mpeg_start_frame,
1204     .decode_slice   = videotoolbox_mpeg_decode_slice,
1205     .end_frame      = videotoolbox_mpeg_end_frame,
1206     .frame_params   = videotoolbox_frame_params,
1207     .init           = videotoolbox_common_init,
1208     .uninit         = videotoolbox_uninit,
1209     .priv_data_size = sizeof(VTContext),
1210 };
1211
1212 static AVVideotoolboxContext *av_videotoolbox_alloc_context_with_pix_fmt(enum AVPixelFormat pix_fmt,
1213                                                                          bool full_range)
1214 {
1215     AVVideotoolboxContext *ret = av_mallocz(sizeof(*ret));
1216
1217     if (ret) {
1218         ret->output_callback = videotoolbox_decoder_callback;
1219
1220         OSType cv_pix_fmt_type = av_map_videotoolbox_format_from_pixfmt2(pix_fmt, full_range);
1221         if (cv_pix_fmt_type == 0) {
1222             cv_pix_fmt_type = kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange;
1223         }
1224         ret->cv_pix_fmt_type = cv_pix_fmt_type;
1225     }
1226
1227     return ret;
1228 }
1229
1230 AVVideotoolboxContext *av_videotoolbox_alloc_context(void)
1231 {
1232     return av_videotoolbox_alloc_context_with_pix_fmt(AV_PIX_FMT_NONE, false);
1233 }
1234
1235 int av_videotoolbox_default_init(AVCodecContext *avctx)
1236 {
1237     return av_videotoolbox_default_init2(avctx, NULL);
1238 }
1239
1240 int av_videotoolbox_default_init2(AVCodecContext *avctx, AVVideotoolboxContext *vtctx)
1241 {
1242     enum AVPixelFormat pix_fmt = videotoolbox_best_pixel_format(avctx);
1243     bool full_range = avctx->color_range == AVCOL_RANGE_JPEG;
1244     avctx->hwaccel_context = vtctx ?: av_videotoolbox_alloc_context_with_pix_fmt(pix_fmt, full_range);
1245     if (!avctx->hwaccel_context)
1246         return AVERROR(ENOMEM);
1247     return videotoolbox_start(avctx);
1248 }
1249
1250 void av_videotoolbox_default_free(AVCodecContext *avctx)
1251 {
1252
1253     videotoolbox_stop(avctx);
1254     av_freep(&avctx->hwaccel_context);
1255 }
1256 #endif /* CONFIG_VIDEOTOOLBOX */