git.sesse.net Git - ffmpeg/blob - libavcodec/videotoolboxenc.c

   1 /*
   2  * copyright (c) 2015 Rick Kern <kernrj@gmail.com>
   3  *
   4  * This file is part of FFmpeg.
   5  *
   6  * FFmpeg is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * FFmpeg is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with FFmpeg; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 #include <VideoToolbox/VideoToolbox.h>
  22 #include <CoreVideo/CoreVideo.h>
  23 #include <CoreMedia/CoreMedia.h>
  24 #include <TargetConditionals.h>
  25 #include <Availability.h>
  26 #include "avcodec.h"
  27 #include "libavutil/opt.h"
  28 #include "libavutil/avassert.h"
  29 #include "libavutil/avstring.h"
  30 #include "libavcodec/avcodec.h"
  31 #include "libavutil/pixdesc.h"
  32 #include "internal.h"
  33 #include <pthread.h>
  34 #include "h264.h"
  35 #include "h264_sei.h"
  36 #include <dlfcn.h>
  37
  38 #if !HAVE_KCMVIDEOCODECTYPE_HEVC
  39 enum { kCMVideoCodecType_HEVC = 'hvc1' };
  40 #endif
  41
  42 #if !HAVE_KCVPIXELFORMATTYPE_420YPCBCR10BIPLANARVIDEORANGE
  43 enum { kCVPixelFormatType_420YpCbCr10BiPlanarFullRange = 'xf20' };
  44 enum { kCVPixelFormatType_420YpCbCr10BiPlanarVideoRange = 'x420' };
  45 #endif
  46
  47 typedef OSStatus (*getParameterSetAtIndex)(CMFormatDescriptionRef videoDesc,
  48                                            size_t parameterSetIndex,
  49                                            const uint8_t **parameterSetPointerOut,
  50                                            size_t *parameterSetSizeOut,
  51                                            size_t *parameterSetCountOut,
  52                                            int *NALUnitHeaderLengthOut);
  53
  54 //These symbols may not be present
  55 static struct{
  56     CFStringRef kCVImageBufferColorPrimaries_ITU_R_2020;
  57     CFStringRef kCVImageBufferTransferFunction_ITU_R_2020;
  58     CFStringRef kCVImageBufferYCbCrMatrix_ITU_R_2020;
  59
  60     CFStringRef kVTCompressionPropertyKey_H264EntropyMode;
  61     CFStringRef kVTH264EntropyMode_CAVLC;
  62     CFStringRef kVTH264EntropyMode_CABAC;
  63
  64     CFStringRef kVTProfileLevel_H264_Baseline_4_0;
  65     CFStringRef kVTProfileLevel_H264_Baseline_4_2;
  66     CFStringRef kVTProfileLevel_H264_Baseline_5_0;
  67     CFStringRef kVTProfileLevel_H264_Baseline_5_1;
  68     CFStringRef kVTProfileLevel_H264_Baseline_5_2;
  69     CFStringRef kVTProfileLevel_H264_Baseline_AutoLevel;
  70     CFStringRef kVTProfileLevel_H264_Main_4_2;
  71     CFStringRef kVTProfileLevel_H264_Main_5_1;
  72     CFStringRef kVTProfileLevel_H264_Main_5_2;
  73     CFStringRef kVTProfileLevel_H264_Main_AutoLevel;
  74     CFStringRef kVTProfileLevel_H264_High_3_0;
  75     CFStringRef kVTProfileLevel_H264_High_3_1;
  76     CFStringRef kVTProfileLevel_H264_High_3_2;
  77     CFStringRef kVTProfileLevel_H264_High_4_0;
  78     CFStringRef kVTProfileLevel_H264_High_4_1;
  79     CFStringRef kVTProfileLevel_H264_High_4_2;
  80     CFStringRef kVTProfileLevel_H264_High_5_1;
  81     CFStringRef kVTProfileLevel_H264_High_5_2;
  82     CFStringRef kVTProfileLevel_H264_High_AutoLevel;
  83
  84     CFStringRef kVTProfileLevel_HEVC_Main_AutoLevel;
  85     CFStringRef kVTProfileLevel_HEVC_Main10_AutoLevel;
  86
  87     CFStringRef kVTCompressionPropertyKey_RealTime;
  88
  89     CFStringRef kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder;
  90     CFStringRef kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder;
  91
  92     getParameterSetAtIndex CMVideoFormatDescriptionGetHEVCParameterSetAtIndex;
  93 } compat_keys;
  94
  95 #define GET_SYM(symbol, defaultVal)                                     \
  96 do{                                                                     \
  97     CFStringRef* handle = (CFStringRef*)dlsym(RTLD_DEFAULT, #symbol);   \
  98     if(!handle)                                                         \
  99         compat_keys.symbol = CFSTR(defaultVal);                         \
 100     else                                                                \
 101         compat_keys.symbol = *handle;                                   \
 102 }while(0)
 103
 104 static pthread_once_t once_ctrl = PTHREAD_ONCE_INIT;
 105
 106 static void loadVTEncSymbols(){
 107     compat_keys.CMVideoFormatDescriptionGetHEVCParameterSetAtIndex =
 108         (getParameterSetAtIndex)dlsym(
 109             RTLD_DEFAULT,
 110             "CMVideoFormatDescriptionGetHEVCParameterSetAtIndex"
 111         );
 112
 113     GET_SYM(kCVImageBufferColorPrimaries_ITU_R_2020,   "ITU_R_2020");
 114     GET_SYM(kCVImageBufferTransferFunction_ITU_R_2020, "ITU_R_2020");
 115     GET_SYM(kCVImageBufferYCbCrMatrix_ITU_R_2020,      "ITU_R_2020");
 116
 117     GET_SYM(kVTCompressionPropertyKey_H264EntropyMode, "H264EntropyMode");
 118     GET_SYM(kVTH264EntropyMode_CAVLC, "CAVLC");
 119     GET_SYM(kVTH264EntropyMode_CABAC, "CABAC");
 120
 121     GET_SYM(kVTProfileLevel_H264_Baseline_4_0,       "H264_Baseline_4_0");
 122     GET_SYM(kVTProfileLevel_H264_Baseline_4_2,       "H264_Baseline_4_2");
 123     GET_SYM(kVTProfileLevel_H264_Baseline_5_0,       "H264_Baseline_5_0");
 124     GET_SYM(kVTProfileLevel_H264_Baseline_5_1,       "H264_Baseline_5_1");
 125     GET_SYM(kVTProfileLevel_H264_Baseline_5_2,       "H264_Baseline_5_2");
 126     GET_SYM(kVTProfileLevel_H264_Baseline_AutoLevel, "H264_Baseline_AutoLevel");
 127     GET_SYM(kVTProfileLevel_H264_Main_4_2,           "H264_Main_4_2");
 128     GET_SYM(kVTProfileLevel_H264_Main_5_1,           "H264_Main_5_1");
 129     GET_SYM(kVTProfileLevel_H264_Main_5_2,           "H264_Main_5_2");
 130     GET_SYM(kVTProfileLevel_H264_Main_AutoLevel,     "H264_Main_AutoLevel");
 131     GET_SYM(kVTProfileLevel_H264_High_3_0,           "H264_High_3_0");
 132     GET_SYM(kVTProfileLevel_H264_High_3_1,           "H264_High_3_1");
 133     GET_SYM(kVTProfileLevel_H264_High_3_2,           "H264_High_3_2");
 134     GET_SYM(kVTProfileLevel_H264_High_4_0,           "H264_High_4_0");
 135     GET_SYM(kVTProfileLevel_H264_High_4_1,           "H264_High_4_1");
 136     GET_SYM(kVTProfileLevel_H264_High_4_2,           "H264_High_4_2");
 137     GET_SYM(kVTProfileLevel_H264_High_5_1,           "H264_High_5_1");
 138     GET_SYM(kVTProfileLevel_H264_High_5_2,           "H264_High_5_2");
 139     GET_SYM(kVTProfileLevel_H264_High_AutoLevel,     "H264_High_AutoLevel");
 140
 141     GET_SYM(kVTProfileLevel_HEVC_Main_AutoLevel,     "HEVC_Main_AutoLevel");
 142     GET_SYM(kVTProfileLevel_HEVC_Main10_AutoLevel,   "HEVC_Main10_AutoLevel");
 143
 144     GET_SYM(kVTCompressionPropertyKey_RealTime, "RealTime");
 145
 146     GET_SYM(kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder,
 147             "EnableHardwareAcceleratedVideoEncoder");
 148     GET_SYM(kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder,
 149             "RequireHardwareAcceleratedVideoEncoder");
 150 }
 151
 152 typedef enum VT_H264Profile {
 153     H264_PROF_AUTO,
 154     H264_PROF_BASELINE,
 155     H264_PROF_MAIN,
 156     H264_PROF_HIGH,
 157     H264_PROF_COUNT
 158 } VT_H264Profile;
 159
 160 typedef enum VTH264Entropy{
 161     VT_ENTROPY_NOT_SET,
 162     VT_CAVLC,
 163     VT_CABAC
 164 } VTH264Entropy;
 165
 166 typedef enum VT_HEVCProfile {
 167     HEVC_PROF_AUTO,
 168     HEVC_PROF_MAIN,
 169     HEVC_PROF_MAIN10,
 170     HEVC_PROF_COUNT
 171 } VT_HEVCProfile;
 172
 173 static const uint8_t start_code[] = { 0, 0, 0, 1 };
 174
 175 typedef struct ExtraSEI {
 176   void *data;
 177   size_t size;
 178 } ExtraSEI;
 179
 180 typedef struct BufNode {
 181     CMSampleBufferRef cm_buffer;
 182     ExtraSEI *sei;
 183     struct BufNode* next;
 184     int error;
 185 } BufNode;
 186
 187 typedef struct VTEncContext {
 188     AVClass *class;
 189     enum AVCodecID codec_id;
 190     VTCompressionSessionRef session;
 191     CFStringRef ycbcr_matrix;
 192     CFStringRef color_primaries;
 193     CFStringRef transfer_function;
 194     getParameterSetAtIndex get_param_set_func;
 195
 196     pthread_mutex_t lock;
 197     pthread_cond_t  cv_sample_sent;
 198
 199     int async_error;
 200
 201     BufNode *q_head;
 202     BufNode *q_tail;
 203
 204     int64_t frame_ct_out;
 205     int64_t frame_ct_in;
 206
 207     int64_t first_pts;
 208     int64_t dts_delta;
 209
 210     int64_t profile;
 211     int64_t level;
 212     int64_t entropy;
 213     int64_t realtime;
 214     int64_t frames_before;
 215     int64_t frames_after;
 216
 217     int64_t allow_sw;
 218     int64_t require_sw;
 219
 220     bool flushing;
 221     bool has_b_frames;
 222     bool warned_color_range;
 223     bool a53_cc;
 224 } VTEncContext;
 225
 226 static int vtenc_populate_extradata(AVCodecContext   *avctx,
 227                                     CMVideoCodecType codec_type,
 228                                     CFStringRef      profile_level,
 229                                     CFNumberRef      gamma_level,
 230                                     CFDictionaryRef  enc_info,
 231                                     CFDictionaryRef  pixel_buffer_info);
 232
 233 /**
 234  * NULL-safe release of *refPtr, and sets value to NULL.
 235  */
 236 static void vt_release_num(CFNumberRef* refPtr){
 237     if (!*refPtr) {
 238         return;
 239     }
 240
 241     CFRelease(*refPtr);
 242     *refPtr = NULL;
 243 }
 244
 245 static void set_async_error(VTEncContext *vtctx, int err)
 246 {
 247     BufNode *info;
 248
 249     pthread_mutex_lock(&vtctx->lock);
 250
 251     vtctx->async_error = err;
 252
 253     info = vtctx->q_head;
 254     vtctx->q_head = vtctx->q_tail = NULL;
 255
 256     while (info) {
 257         BufNode *next = info->next;
 258         CFRelease(info->cm_buffer);
 259         av_free(info);
 260         info = next;
 261     }
 262
 263     pthread_mutex_unlock(&vtctx->lock);
 264 }
 265
 266 static void clear_frame_queue(VTEncContext *vtctx)
 267 {
 268     set_async_error(vtctx, 0);
 269 }
 270
 271 static int vtenc_q_pop(VTEncContext *vtctx, bool wait, CMSampleBufferRef *buf, ExtraSEI **sei)
 272 {
 273     BufNode *info;
 274
 275     pthread_mutex_lock(&vtctx->lock);
 276
 277     if (vtctx->async_error) {
 278         pthread_mutex_unlock(&vtctx->lock);
 279         return vtctx->async_error;
 280     }
 281
 282     if (vtctx->flushing && vtctx->frame_ct_in == vtctx->frame_ct_out) {
 283         *buf = NULL;
 284
 285         pthread_mutex_unlock(&vtctx->lock);
 286         return 0;
 287     }
 288
 289     while (!vtctx->q_head && !vtctx->async_error && wait) {
 290         pthread_cond_wait(&vtctx->cv_sample_sent, &vtctx->lock);
 291     }
 292
 293     if (!vtctx->q_head) {
 294         pthread_mutex_unlock(&vtctx->lock);
 295         *buf = NULL;
 296         return 0;
 297     }
 298
 299     info = vtctx->q_head;
 300     vtctx->q_head = vtctx->q_head->next;
 301     if (!vtctx->q_head) {
 302         vtctx->q_tail = NULL;
 303     }
 304
 305     pthread_mutex_unlock(&vtctx->lock);
 306
 307     *buf = info->cm_buffer;
 308     if (sei && *buf) {
 309         *sei = info->sei;
 310     } else if (info->sei) {
 311         if (info->sei->data) av_free(info->sei->data);
 312         av_free(info->sei);
 313     }
 314     av_free(info);
 315
 316     vtctx->frame_ct_out++;
 317
 318     return 0;
 319 }
 320
 321 static void vtenc_q_push(VTEncContext *vtctx, CMSampleBufferRef buffer, ExtraSEI *sei)
 322 {
 323     BufNode *info = av_malloc(sizeof(BufNode));
 324     if (!info) {
 325         set_async_error(vtctx, AVERROR(ENOMEM));
 326         return;
 327     }
 328
 329     CFRetain(buffer);
 330     info->cm_buffer = buffer;
 331     info->sei = sei;
 332     info->next = NULL;
 333
 334     pthread_mutex_lock(&vtctx->lock);
 335     pthread_cond_signal(&vtctx->cv_sample_sent);
 336
 337     if (!vtctx->q_head) {
 338         vtctx->q_head = info;
 339     } else {
 340         vtctx->q_tail->next = info;
 341     }
 342
 343     vtctx->q_tail = info;
 344
 345     pthread_mutex_unlock(&vtctx->lock);
 346 }
 347
 348 static int count_nalus(size_t length_code_size,
 349                        CMSampleBufferRef sample_buffer,
 350                        int *count)
 351 {
 352     size_t offset = 0;
 353     int status;
 354     int nalu_ct = 0;
 355     uint8_t size_buf[4];
 356     size_t src_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
 357     CMBlockBufferRef block = CMSampleBufferGetDataBuffer(sample_buffer);
 358
 359     if (length_code_size > 4)
 360         return AVERROR_INVALIDDATA;
 361
 362     while (offset < src_size) {
 363         size_t curr_src_len;
 364         size_t box_len = 0;
 365         size_t i;
 366
 367         status = CMBlockBufferCopyDataBytes(block,
 368                                             offset,
 369                                             length_code_size,
 370                                             size_buf);
 371
 372         for (i = 0; i < length_code_size; i++) {
 373             box_len <<= 8;
 374             box_len |= size_buf[i];
 375         }
 376
 377         curr_src_len = box_len + length_code_size;
 378         offset += curr_src_len;
 379
 380         nalu_ct++;
 381     }
 382
 383     *count = nalu_ct;
 384     return 0;
 385 }
 386
 387 static CMVideoCodecType get_cm_codec_type(enum AVCodecID id)
 388 {
 389     switch (id) {
 390     case AV_CODEC_ID_H264: return kCMVideoCodecType_H264;
 391     case AV_CODEC_ID_HEVC: return kCMVideoCodecType_HEVC;
 392     default:               return 0;
 393     }
 394 }
 395
 396 /**
 397  * Get the parameter sets from a CMSampleBufferRef.
 398  * @param dst If *dst isn't NULL, the parameters are copied into existing
 399  *            memory. *dst_size must be set accordingly when *dst != NULL.
 400  *            If *dst is NULL, it will be allocated.
 401  *            In all cases, *dst_size is set to the number of bytes used starting
 402  *            at *dst.
 403  */
 404 static int get_params_size(
 405     AVCodecContext              *avctx,
 406     CMVideoFormatDescriptionRef vid_fmt,
 407     size_t                      *size)
 408 {
 409     VTEncContext *vtctx = avctx->priv_data;
 410     size_t total_size = 0;
 411     size_t ps_count;
 412     int is_count_bad = 0;
 413     size_t i;
 414     int status;
 415     status = vtctx->get_param_set_func(vid_fmt,
 416                                        0,
 417                                        NULL,
 418                                        NULL,
 419                                        &ps_count,
 420                                        NULL);
 421     if (status) {
 422         is_count_bad = 1;
 423         ps_count     = 0;
 424         status       = 0;
 425     }
 426
 427     for (i = 0; i < ps_count || is_count_bad; i++) {
 428         const uint8_t *ps;
 429         size_t ps_size;
 430         status = vtctx->get_param_set_func(vid_fmt,
 431                                            i,
 432                                            &ps,
 433                                            &ps_size,
 434                                            NULL,
 435                                            NULL);
 436         if (status) {
 437             /*
 438              * When ps_count is invalid, status != 0 ends the loop normally
 439              * unless we didn't get any parameter sets.
 440              */
 441             if (i > 0 && is_count_bad) status = 0;
 442
 443             break;
 444         }
 445
 446         total_size += ps_size + sizeof(start_code);
 447     }
 448
 449     if (status) {
 450         av_log(avctx, AV_LOG_ERROR, "Error getting parameter set sizes: %d\n", status);
 451         return AVERROR_EXTERNAL;
 452     }
 453
 454     *size = total_size;
 455     return 0;
 456 }
 457
 458 static int copy_param_sets(
 459     AVCodecContext              *avctx,
 460     CMVideoFormatDescriptionRef vid_fmt,
 461     uint8_t                     *dst,
 462     size_t                      dst_size)
 463 {
 464     VTEncContext *vtctx = avctx->priv_data;
 465     size_t ps_count;
 466     int is_count_bad = 0;
 467     int status;
 468     size_t offset = 0;
 469     size_t i;
 470
 471     status = vtctx->get_param_set_func(vid_fmt,
 472                                        0,
 473                                        NULL,
 474                                        NULL,
 475                                        &ps_count,
 476                                        NULL);
 477     if (status) {
 478         is_count_bad = 1;
 479         ps_count     = 0;
 480         status       = 0;
 481     }
 482
 483
 484     for (i = 0; i < ps_count || is_count_bad; i++) {
 485         const uint8_t *ps;
 486         size_t ps_size;
 487         size_t next_offset;
 488
 489         status = vtctx->get_param_set_func(vid_fmt,
 490                                            i,
 491                                            &ps,
 492                                            &ps_size,
 493                                            NULL,
 494                                            NULL);
 495         if (status) {
 496             if (i > 0 && is_count_bad) status = 0;
 497
 498             break;
 499         }
 500
 501         next_offset = offset + sizeof(start_code) + ps_size;
 502         if (dst_size < next_offset) {
 503             av_log(avctx, AV_LOG_ERROR, "Error: buffer too small for parameter sets.\n");
 504             return AVERROR_BUFFER_TOO_SMALL;
 505         }
 506
 507         memcpy(dst + offset, start_code, sizeof(start_code));
 508         offset += sizeof(start_code);
 509
 510         memcpy(dst + offset, ps, ps_size);
 511         offset = next_offset;
 512     }
 513
 514     if (status) {
 515         av_log(avctx, AV_LOG_ERROR, "Error getting parameter set data: %d\n", status);
 516         return AVERROR_EXTERNAL;
 517     }
 518
 519     return 0;
 520 }
 521
 522 static int set_extradata(AVCodecContext *avctx, CMSampleBufferRef sample_buffer)
 523 {
 524     CMVideoFormatDescriptionRef vid_fmt;
 525     size_t total_size;
 526     int status;
 527
 528     vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
 529     if (!vid_fmt) {
 530         av_log(avctx, AV_LOG_ERROR, "No video format.\n");
 531         return AVERROR_EXTERNAL;
 532     }
 533
 534     status = get_params_size(avctx, vid_fmt, &total_size);
 535     if (status) {
 536         av_log(avctx, AV_LOG_ERROR, "Could not get parameter sets.\n");
 537         return status;
 538     }
 539
 540     avctx->extradata = av_mallocz(total_size + AV_INPUT_BUFFER_PADDING_SIZE);
 541     if (!avctx->extradata) {
 542         return AVERROR(ENOMEM);
 543     }
 544     avctx->extradata_size = total_size;
 545
 546     status = copy_param_sets(avctx, vid_fmt, avctx->extradata, total_size);
 547
 548     if (status) {
 549         av_log(avctx, AV_LOG_ERROR, "Could not copy param sets.\n");
 550         return status;
 551     }
 552
 553     return 0;
 554 }
 555
 556 static void vtenc_output_callback(
 557     void *ctx,
 558     void *sourceFrameCtx,
 559     OSStatus status,
 560     VTEncodeInfoFlags flags,
 561     CMSampleBufferRef sample_buffer)
 562 {
 563     AVCodecContext *avctx = ctx;
 564     VTEncContext   *vtctx = avctx->priv_data;
 565     ExtraSEI *sei = sourceFrameCtx;
 566
 567     if (vtctx->async_error) {
 568         if(sample_buffer) CFRelease(sample_buffer);
 569         return;
 570     }
 571
 572     if (status || !sample_buffer) {
 573         av_log(avctx, AV_LOG_ERROR, "Error encoding frame: %d\n", (int)status);
 574         set_async_error(vtctx, AVERROR_EXTERNAL);
 575         return;
 576     }
 577
 578     if (!avctx->extradata && (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)) {
 579         int set_status = set_extradata(avctx, sample_buffer);
 580         if (set_status) {
 581             set_async_error(vtctx, set_status);
 582             return;
 583         }
 584     }
 585
 586     vtenc_q_push(vtctx, sample_buffer, sei);
 587 }
 588
 589 static int get_length_code_size(
 590     AVCodecContext    *avctx,
 591     CMSampleBufferRef sample_buffer,
 592     size_t            *size)
 593 {
 594     VTEncContext *vtctx = avctx->priv_data;
 595     CMVideoFormatDescriptionRef vid_fmt;
 596     int isize;
 597     int status;
 598
 599     vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
 600     if (!vid_fmt) {
 601         av_log(avctx, AV_LOG_ERROR, "Error getting buffer format description.\n");
 602         return AVERROR_EXTERNAL;
 603     }
 604
 605     status = vtctx->get_param_set_func(vid_fmt,
 606                                        0,
 607                                        NULL,
 608                                        NULL,
 609                                        NULL,
 610                                        &isize);
 611     if (status) {
 612         av_log(avctx, AV_LOG_ERROR, "Error getting length code size: %d\n", status);
 613         return AVERROR_EXTERNAL;
 614     }
 615
 616     *size = isize;
 617     return 0;
 618 }
 619
 620 /*
 621  * Returns true on success.
 622  *
 623  * If profile_level_val is NULL and this method returns true, don't specify the
 624  * profile/level to the encoder.
 625  */
 626 static bool get_vt_h264_profile_level(AVCodecContext *avctx,
 627                                       CFStringRef    *profile_level_val)
 628 {
 629     VTEncContext *vtctx = avctx->priv_data;
 630     int64_t profile = vtctx->profile;
 631
 632     if (profile == H264_PROF_AUTO && vtctx->level) {
 633         //Need to pick a profile if level is not auto-selected.
 634         profile = vtctx->has_b_frames ? H264_PROF_MAIN : H264_PROF_BASELINE;
 635     }
 636
 637     *profile_level_val = NULL;
 638
 639     switch (profile) {
 640         case H264_PROF_AUTO:
 641             return true;
 642
 643         case H264_PROF_BASELINE:
 644             switch (vtctx->level) {
 645                 case  0: *profile_level_val =
 646                                   compat_keys.kVTProfileLevel_H264_Baseline_AutoLevel; break;
 647                 case 13: *profile_level_val = kVTProfileLevel_H264_Baseline_1_3;       break;
 648                 case 30: *profile_level_val = kVTProfileLevel_H264_Baseline_3_0;       break;
 649                 case 31: *profile_level_val = kVTProfileLevel_H264_Baseline_3_1;       break;
 650                 case 32: *profile_level_val = kVTProfileLevel_H264_Baseline_3_2;       break;
 651                 case 40: *profile_level_val =
 652                                   compat_keys.kVTProfileLevel_H264_Baseline_4_0;       break;
 653                 case 41: *profile_level_val = kVTProfileLevel_H264_Baseline_4_1;       break;
 654                 case 42: *profile_level_val =
 655                                   compat_keys.kVTProfileLevel_H264_Baseline_4_2;       break;
 656                 case 50: *profile_level_val =
 657                                   compat_keys.kVTProfileLevel_H264_Baseline_5_0;       break;
 658                 case 51: *profile_level_val =
 659                                   compat_keys.kVTProfileLevel_H264_Baseline_5_1;       break;
 660                 case 52: *profile_level_val =
 661                                   compat_keys.kVTProfileLevel_H264_Baseline_5_2;       break;
 662             }
 663             break;
 664
 665         case H264_PROF_MAIN:
 666             switch (vtctx->level) {
 667                 case  0: *profile_level_val =
 668                                   compat_keys.kVTProfileLevel_H264_Main_AutoLevel; break;
 669                 case 30: *profile_level_val = kVTProfileLevel_H264_Main_3_0;       break;
 670                 case 31: *profile_level_val = kVTProfileLevel_H264_Main_3_1;       break;
 671                 case 32: *profile_level_val = kVTProfileLevel_H264_Main_3_2;       break;
 672                 case 40: *profile_level_val = kVTProfileLevel_H264_Main_4_0;       break;
 673                 case 41: *profile_level_val = kVTProfileLevel_H264_Main_4_1;       break;
 674                 case 42: *profile_level_val =
 675                                   compat_keys.kVTProfileLevel_H264_Main_4_2;       break;
 676                 case 50: *profile_level_val = kVTProfileLevel_H264_Main_5_0;       break;
 677                 case 51: *profile_level_val =
 678                                   compat_keys.kVTProfileLevel_H264_Main_5_1;       break;
 679                 case 52: *profile_level_val =
 680                                   compat_keys.kVTProfileLevel_H264_Main_5_2;       break;
 681             }
 682             break;
 683
 684         case H264_PROF_HIGH:
 685             switch (vtctx->level) {
 686                 case  0: *profile_level_val =
 687                                   compat_keys.kVTProfileLevel_H264_High_AutoLevel; break;
 688                 case 30: *profile_level_val =
 689                                   compat_keys.kVTProfileLevel_H264_High_3_0;       break;
 690                 case 31: *profile_level_val =
 691                                   compat_keys.kVTProfileLevel_H264_High_3_1;       break;
 692                 case 32: *profile_level_val =
 693                                   compat_keys.kVTProfileLevel_H264_High_3_2;       break;
 694                 case 40: *profile_level_val =
 695                                   compat_keys.kVTProfileLevel_H264_High_4_0;       break;
 696                 case 41: *profile_level_val =
 697                                   compat_keys.kVTProfileLevel_H264_High_4_1;       break;
 698                 case 42: *profile_level_val =
 699                                   compat_keys.kVTProfileLevel_H264_High_4_2;       break;
 700                 case 50: *profile_level_val = kVTProfileLevel_H264_High_5_0;       break;
 701                 case 51: *profile_level_val =
 702                                   compat_keys.kVTProfileLevel_H264_High_5_1;       break;
 703                 case 52: *profile_level_val =
 704                                   compat_keys.kVTProfileLevel_H264_High_5_2;       break;
 705             }
 706             break;
 707     }
 708
 709     if (!*profile_level_val) {
 710         av_log(avctx, AV_LOG_ERROR, "Invalid Profile/Level.\n");
 711         return false;
 712     }
 713
 714     return true;
 715 }
 716
 717 /*
 718  * Returns true on success.
 719  *
 720  * If profile_level_val is NULL and this method returns true, don't specify the
 721  * profile/level to the encoder.
 722  */
 723 static bool get_vt_hevc_profile_level(AVCodecContext *avctx,
 724                                       CFStringRef    *profile_level_val)
 725 {
 726     VTEncContext *vtctx = avctx->priv_data;
 727     int64_t profile = vtctx->profile;
 728
 729     *profile_level_val = NULL;
 730
 731     switch (profile) {
 732         case HEVC_PROF_AUTO:
 733             return true;
 734         case HEVC_PROF_MAIN:
 735             *profile_level_val =
 736                 compat_keys.kVTProfileLevel_HEVC_Main_AutoLevel;
 737             break;
 738         case HEVC_PROF_MAIN10:
 739             *profile_level_val =
 740                 compat_keys.kVTProfileLevel_HEVC_Main10_AutoLevel;
 741             break;
 742     }
 743
 744     if (!*profile_level_val) {
 745         av_log(avctx, AV_LOG_ERROR, "Invalid Profile/Level.\n");
 746         return false;
 747     }
 748
 749     return true;
 750 }
 751
 752 static int get_cv_pixel_format(AVCodecContext* avctx,
 753                                enum AVPixelFormat fmt,
 754                                enum AVColorRange range,
 755                                int* av_pixel_format,
 756                                int* range_guessed)
 757 {
 758     if (range_guessed) *range_guessed = range != AVCOL_RANGE_MPEG &&
 759                                         range != AVCOL_RANGE_JPEG;
 760
 761     //MPEG range is used when no range is set
 762     if (fmt == AV_PIX_FMT_NV12) {
 763         *av_pixel_format = range == AVCOL_RANGE_JPEG ?
 764                                         kCVPixelFormatType_420YpCbCr8BiPlanarFullRange :
 765                                         kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange;
 766     } else if (fmt == AV_PIX_FMT_YUV420P) {
 767         *av_pixel_format = range == AVCOL_RANGE_JPEG ?
 768                                         kCVPixelFormatType_420YpCbCr8PlanarFullRange :
 769                                         kCVPixelFormatType_420YpCbCr8Planar;
 770     } else if (fmt == AV_PIX_FMT_P010LE) {
 771         *av_pixel_format = range == AVCOL_RANGE_JPEG ?
 772                                         kCVPixelFormatType_420YpCbCr10BiPlanarFullRange :
 773                                         kCVPixelFormatType_420YpCbCr10BiPlanarVideoRange;
 774         *av_pixel_format = kCVPixelFormatType_420YpCbCr10BiPlanarVideoRange;
 775     } else {
 776         return AVERROR(EINVAL);
 777     }
 778
 779     return 0;
 780 }
 781
 782 static void add_color_attr(AVCodecContext *avctx, CFMutableDictionaryRef dict) {
 783     VTEncContext *vtctx = avctx->priv_data;
 784
 785     if (vtctx->color_primaries) {
 786         CFDictionarySetValue(dict,
 787                              kCVImageBufferColorPrimariesKey,
 788                              vtctx->color_primaries);
 789     }
 790
 791     if (vtctx->transfer_function) {
 792         CFDictionarySetValue(dict,
 793                              kCVImageBufferTransferFunctionKey,
 794                              vtctx->transfer_function);
 795     }
 796
 797     if (vtctx->ycbcr_matrix) {
 798         CFDictionarySetValue(dict,
 799                              kCVImageBufferYCbCrMatrixKey,
 800                              vtctx->ycbcr_matrix);
 801     }
 802 }
 803
 804 static int create_cv_pixel_buffer_info(AVCodecContext* avctx,
 805                                        CFMutableDictionaryRef* dict)
 806 {
 807     CFNumberRef cv_color_format_num = NULL;
 808     CFNumberRef width_num = NULL;
 809     CFNumberRef height_num = NULL;
 810     CFMutableDictionaryRef pixel_buffer_info = NULL;
 811     int cv_color_format;
 812     int status = get_cv_pixel_format(avctx,
 813                                      avctx->pix_fmt,
 814                                      avctx->color_range,
 815                                      &cv_color_format,
 816                                      NULL);
 817     if (status) return status;
 818
 819     pixel_buffer_info = CFDictionaryCreateMutable(
 820                             kCFAllocatorDefault,
 821                             20,
 822                             &kCFCopyStringDictionaryKeyCallBacks,
 823                             &kCFTypeDictionaryValueCallBacks);
 824
 825     if (!pixel_buffer_info) goto pbinfo_nomem;
 826
 827     cv_color_format_num = CFNumberCreate(kCFAllocatorDefault,
 828                                          kCFNumberSInt32Type,
 829                                          &cv_color_format);
 830     if (!cv_color_format_num) goto pbinfo_nomem;
 831
 832     CFDictionarySetValue(pixel_buffer_info,
 833                          kCVPixelBufferPixelFormatTypeKey,
 834                          cv_color_format_num);
 835     vt_release_num(&cv_color_format_num);
 836
 837     width_num = CFNumberCreate(kCFAllocatorDefault,
 838                                kCFNumberSInt32Type,
 839                                &avctx->width);
 840     if (!width_num) return AVERROR(ENOMEM);
 841
 842     CFDictionarySetValue(pixel_buffer_info,
 843                          kCVPixelBufferWidthKey,
 844                          width_num);
 845     vt_release_num(&width_num);
 846
 847     height_num = CFNumberCreate(kCFAllocatorDefault,
 848                                 kCFNumberSInt32Type,
 849                                 &avctx->height);
 850     if (!height_num) goto pbinfo_nomem;
 851
 852     CFDictionarySetValue(pixel_buffer_info,
 853                          kCVPixelBufferHeightKey,
 854                          height_num);
 855     vt_release_num(&height_num);
 856
 857     add_color_attr(avctx, pixel_buffer_info);
 858
 859     *dict = pixel_buffer_info;
 860     return 0;
 861
 862 pbinfo_nomem:
 863     vt_release_num(&cv_color_format_num);
 864     vt_release_num(&width_num);
 865     vt_release_num(&height_num);
 866     if (pixel_buffer_info) CFRelease(pixel_buffer_info);
 867
 868     return AVERROR(ENOMEM);
 869 }
 870
 871 static int get_cv_color_primaries(AVCodecContext *avctx,
 872                                   CFStringRef *primaries)
 873 {
 874     enum AVColorPrimaries pri = avctx->color_primaries;
 875     switch (pri) {
 876         case AVCOL_PRI_UNSPECIFIED:
 877             *primaries = NULL;
 878             break;
 879
 880         case AVCOL_PRI_BT709:
 881             *primaries = kCVImageBufferColorPrimaries_ITU_R_709_2;
 882             break;
 883
 884         case AVCOL_PRI_BT2020:
 885             *primaries = compat_keys.kCVImageBufferColorPrimaries_ITU_R_2020;
 886             break;
 887
 888         default:
 889             av_log(avctx, AV_LOG_ERROR, "Color primaries %s is not supported.\n", av_color_primaries_name(pri));
 890             *primaries = NULL;
 891             return -1;
 892     }
 893
 894     return 0;
 895 }
 896
 897 static int get_cv_transfer_function(AVCodecContext *avctx,
 898                                     CFStringRef *transfer_fnc,
 899                                     CFNumberRef *gamma_level)
 900 {
 901     enum AVColorTransferCharacteristic trc = avctx->color_trc;
 902     Float32 gamma;
 903     *gamma_level = NULL;
 904
 905     switch (trc) {
 906         case AVCOL_TRC_UNSPECIFIED:
 907             *transfer_fnc = NULL;
 908             break;
 909
 910         case AVCOL_TRC_BT709:
 911             *transfer_fnc = kCVImageBufferTransferFunction_ITU_R_709_2;
 912             break;
 913
 914         case AVCOL_TRC_SMPTE240M:
 915             *transfer_fnc = kCVImageBufferTransferFunction_SMPTE_240M_1995;
 916             break;
 917
 918 #if HAVE_KCVIMAGEBUFFERTRANSFERFUNCTION_SMPTE_ST_2084_PQ
 919         case AVCOL_TRC_SMPTE2084:
 920             *transfer_fnc = kCVImageBufferTransferFunction_SMPTE_ST_2084_PQ;
 921             break;
 922 #endif
 923 #if HAVE_KCVIMAGEBUFFERTRANSFERFUNCTION_LINEAR
 924         case AVCOL_TRC_LINEAR:
 925             *transfer_fnc = kCVImageBufferTransferFunction_Linear;
 926             break;
 927 #endif
 928 #if HAVE_KCVIMAGEBUFFERTRANSFERFUNCTION_ITU_R_2100_HLG
 929         case AVCOL_TRC_ARIB_STD_B67:
 930             *transfer_fnc = kCVImageBufferTransferFunction_ITU_R_2100_HLG;
 931             break;
 932 #endif
 933
 934         case AVCOL_TRC_GAMMA22:
 935             gamma = 2.2;
 936             *transfer_fnc = kCVImageBufferTransferFunction_UseGamma;
 937             *gamma_level = CFNumberCreate(NULL, kCFNumberFloat32Type, &gamma);
 938             break;
 939
 940         case AVCOL_TRC_GAMMA28:
 941             gamma = 2.8;
 942             *transfer_fnc = kCVImageBufferTransferFunction_UseGamma;
 943             *gamma_level = CFNumberCreate(NULL, kCFNumberFloat32Type, &gamma);
 944             break;
 945
 946         case AVCOL_TRC_BT2020_10:
 947         case AVCOL_TRC_BT2020_12:
 948             *transfer_fnc = compat_keys.kCVImageBufferTransferFunction_ITU_R_2020;
 949             break;
 950
 951         default:
 952             *transfer_fnc = NULL;
 953             av_log(avctx, AV_LOG_ERROR, "Transfer function %s is not supported.\n", av_color_transfer_name(trc));
 954             return -1;
 955     }
 956
 957     return 0;
 958 }
 959
 960 static int get_cv_ycbcr_matrix(AVCodecContext *avctx, CFStringRef *matrix) {
 961     switch(avctx->colorspace) {
 962         case AVCOL_SPC_BT709:
 963             *matrix = kCVImageBufferYCbCrMatrix_ITU_R_709_2;
 964             break;
 965
 966         case AVCOL_SPC_UNSPECIFIED:
 967             *matrix = NULL;
 968             break;
 969
 970         case AVCOL_SPC_BT470BG:
 971         case AVCOL_SPC_SMPTE170M:
 972             *matrix = kCVImageBufferYCbCrMatrix_ITU_R_601_4;
 973             break;
 974
 975         case AVCOL_SPC_SMPTE240M:
 976             *matrix = kCVImageBufferYCbCrMatrix_SMPTE_240M_1995;
 977             break;
 978
 979         case AVCOL_SPC_BT2020_NCL:
 980             *matrix = compat_keys.kCVImageBufferYCbCrMatrix_ITU_R_2020;
 981             break;
 982
 983         default:
 984             av_log(avctx, AV_LOG_ERROR, "Color space %s is not supported.\n", av_color_space_name(avctx->colorspace));
 985             return -1;
 986     }
 987
 988     return 0;
 989 }
 990
 991 static int vtenc_create_encoder(AVCodecContext   *avctx,
 992                                 CMVideoCodecType codec_type,
 993                                 CFStringRef      profile_level,
 994                                 CFNumberRef      gamma_level,
 995                                 CFDictionaryRef  enc_info,
 996                                 CFDictionaryRef  pixel_buffer_info,
 997                                 VTCompressionSessionRef *session)
 998 {
 999     VTEncContext *vtctx = avctx->priv_data;
1000     SInt32       bit_rate = avctx->bit_rate;
1001     SInt32       max_rate = avctx->rc_max_rate;
1002     CFNumberRef  bit_rate_num;
1003     CFNumberRef  bytes_per_second;
1004     CFNumberRef  one_second;
1005     CFArrayRef   data_rate_limits;
1006     int64_t      bytes_per_second_value = 0;
1007     int64_t      one_second_value = 0;
1008     void         *nums[2];
1009
1010     int status = VTCompressionSessionCreate(kCFAllocatorDefault,
1011                                             avctx->width,
1012                                             avctx->height,
1013                                             codec_type,
1014                                             enc_info,
1015                                             pixel_buffer_info,
1016                                             kCFAllocatorDefault,
1017                                             vtenc_output_callback,
1018                                             avctx,
1019                                             session);
1020
1021     if (status || !vtctx->session) {
1022         av_log(avctx, AV_LOG_ERROR, "Error: cannot create compression session: %d\n", status);
1023
1024 #if !TARGET_OS_IPHONE
1025         if (!vtctx->allow_sw) {
1026             av_log(avctx, AV_LOG_ERROR, "Try -allow_sw 1. The hardware encoder may be busy, or not supported.\n");
1027         }
1028 #endif
1029
1030         return AVERROR_EXTERNAL;
1031     }
1032
1033     bit_rate_num = CFNumberCreate(kCFAllocatorDefault,
1034                                   kCFNumberSInt32Type,
1035                                   &bit_rate);
1036     if (!bit_rate_num) return AVERROR(ENOMEM);
1037
1038     status = VTSessionSetProperty(vtctx->session,
1039                                   kVTCompressionPropertyKey_AverageBitRate,
1040                                   bit_rate_num);
1041     CFRelease(bit_rate_num);
1042
1043     if (status) {
1044         av_log(avctx, AV_LOG_ERROR, "Error setting bitrate property: %d\n", status);
1045         return AVERROR_EXTERNAL;
1046     }
1047
1048     if (vtctx->codec_id == AV_CODEC_ID_H264 && max_rate > 0) {
1049         // kVTCompressionPropertyKey_DataRateLimits is not available for HEVC
1050         bytes_per_second_value = max_rate >> 3;
1051         bytes_per_second = CFNumberCreate(kCFAllocatorDefault,
1052                                           kCFNumberSInt64Type,
1053                                           &bytes_per_second_value);
1054         if (!bytes_per_second) {
1055             return AVERROR(ENOMEM);
1056         }
1057         one_second_value = 1;
1058         one_second = CFNumberCreate(kCFAllocatorDefault,
1059                                     kCFNumberSInt64Type,
1060                                     &one_second_value);
1061         if (!one_second) {
1062             CFRelease(bytes_per_second);
1063             return AVERROR(ENOMEM);
1064         }
1065         nums[0] = (void *)bytes_per_second;
1066         nums[1] = (void *)one_second;
1067         data_rate_limits = CFArrayCreate(kCFAllocatorDefault,
1068                                          (const void **)nums,
1069                                          2,
1070                                          &kCFTypeArrayCallBacks);
1071
1072         if (!data_rate_limits) {
1073             CFRelease(bytes_per_second);
1074             CFRelease(one_second);
1075             return AVERROR(ENOMEM);
1076         }
1077         status = VTSessionSetProperty(vtctx->session,
1078                                       kVTCompressionPropertyKey_DataRateLimits,
1079                                       data_rate_limits);
1080
1081         CFRelease(bytes_per_second);
1082         CFRelease(one_second);
1083         CFRelease(data_rate_limits);
1084
1085         if (status) {
1086             av_log(avctx, AV_LOG_ERROR, "Error setting max bitrate property: %d\n", status);
1087             return AVERROR_EXTERNAL;
1088         }
1089     }
1090
1091     if (vtctx->codec_id == AV_CODEC_ID_H264) {
1092         // kVTCompressionPropertyKey_ProfileLevel is not available for HEVC
1093         if (profile_level) {
1094             status = VTSessionSetProperty(vtctx->session,
1095                                         kVTCompressionPropertyKey_ProfileLevel,
1096                                         profile_level);
1097             if (status) {
1098                 av_log(avctx, AV_LOG_ERROR, "Error setting profile/level property: %d\n", status);
1099             }
1100         }
1101     }
1102
1103     if (avctx->gop_size > 0) {
1104         CFNumberRef interval = CFNumberCreate(kCFAllocatorDefault,
1105                                               kCFNumberIntType,
1106                                               &avctx->gop_size);
1107         if (!interval) {
1108             return AVERROR(ENOMEM);
1109         }
1110
1111         status = VTSessionSetProperty(vtctx->session,
1112                                       kVTCompressionPropertyKey_MaxKeyFrameInterval,
1113                                       interval);
1114         CFRelease(interval);
1115
1116         if (status) {
1117             av_log(avctx, AV_LOG_ERROR, "Error setting 'max key-frame interval' property: %d\n", status);
1118             return AVERROR_EXTERNAL;
1119         }
1120     }
1121
1122     if (vtctx->frames_before) {
1123         status = VTSessionSetProperty(vtctx->session,
1124                                       kVTCompressionPropertyKey_MoreFramesBeforeStart,
1125                                       kCFBooleanTrue);
1126
1127         if (status == kVTPropertyNotSupportedErr) {
1128             av_log(avctx, AV_LOG_WARNING, "frames_before property is not supported on this device. Ignoring.\n");
1129         } else if (status) {
1130             av_log(avctx, AV_LOG_ERROR, "Error setting frames_before property: %d\n", status);
1131         }
1132     }
1133
1134     if (vtctx->frames_after) {
1135         status = VTSessionSetProperty(vtctx->session,
1136                                       kVTCompressionPropertyKey_MoreFramesAfterEnd,
1137                                       kCFBooleanTrue);
1138
1139         if (status == kVTPropertyNotSupportedErr) {
1140             av_log(avctx, AV_LOG_WARNING, "frames_after property is not supported on this device. Ignoring.\n");
1141         } else if (status) {
1142             av_log(avctx, AV_LOG_ERROR, "Error setting frames_after property: %d\n", status);
1143         }
1144     }
1145
1146     if (avctx->sample_aspect_ratio.num != 0) {
1147         CFNumberRef num;
1148         CFNumberRef den;
1149         CFMutableDictionaryRef par;
1150         AVRational *avpar = &avctx->sample_aspect_ratio;
1151
1152         av_reduce(&avpar->num, &avpar->den,
1153                    avpar->num,  avpar->den,
1154                   0xFFFFFFFF);
1155
1156         num = CFNumberCreate(kCFAllocatorDefault,
1157                              kCFNumberIntType,
1158                              &avpar->num);
1159
1160         den = CFNumberCreate(kCFAllocatorDefault,
1161                              kCFNumberIntType,
1162                              &avpar->den);
1163
1164
1165
1166         par = CFDictionaryCreateMutable(kCFAllocatorDefault,
1167                                         2,
1168                                         &kCFCopyStringDictionaryKeyCallBacks,
1169                                         &kCFTypeDictionaryValueCallBacks);
1170
1171         if (!par || !num || !den) {
1172             if (par) CFRelease(par);
1173             if (num) CFRelease(num);
1174             if (den) CFRelease(den);
1175
1176             return AVERROR(ENOMEM);
1177         }
1178
1179         CFDictionarySetValue(
1180             par,
1181             kCMFormatDescriptionKey_PixelAspectRatioHorizontalSpacing,
1182             num);
1183
1184         CFDictionarySetValue(
1185             par,
1186             kCMFormatDescriptionKey_PixelAspectRatioVerticalSpacing,
1187             den);
1188
1189         status = VTSessionSetProperty(vtctx->session,
1190                                       kVTCompressionPropertyKey_PixelAspectRatio,
1191                                       par);
1192
1193         CFRelease(par);
1194         CFRelease(num);
1195         CFRelease(den);
1196
1197         if (status) {
1198             av_log(avctx,
1199                    AV_LOG_ERROR,
1200                    "Error setting pixel aspect ratio to %d:%d: %d.\n",
1201                    avctx->sample_aspect_ratio.num,
1202                    avctx->sample_aspect_ratio.den,
1203                    status);
1204
1205             return AVERROR_EXTERNAL;
1206         }
1207     }
1208
1209
1210     if (vtctx->transfer_function) {
1211         status = VTSessionSetProperty(vtctx->session,
1212                                       kVTCompressionPropertyKey_TransferFunction,
1213                                       vtctx->transfer_function);
1214
1215         if (status) {
1216             av_log(avctx, AV_LOG_WARNING, "Could not set transfer function: %d\n", status);
1217         }
1218     }
1219
1220
1221     if (vtctx->ycbcr_matrix) {
1222         status = VTSessionSetProperty(vtctx->session,
1223                                       kVTCompressionPropertyKey_YCbCrMatrix,
1224                                       vtctx->ycbcr_matrix);
1225
1226         if (status) {
1227             av_log(avctx, AV_LOG_WARNING, "Could not set ycbcr matrix: %d\n", status);
1228         }
1229     }
1230
1231
1232     if (vtctx->color_primaries) {
1233         status = VTSessionSetProperty(vtctx->session,
1234                                       kVTCompressionPropertyKey_ColorPrimaries,
1235                                       vtctx->color_primaries);
1236
1237         if (status) {
1238             av_log(avctx, AV_LOG_WARNING, "Could not set color primaries: %d\n", status);
1239         }
1240     }
1241
1242     if (gamma_level) {
1243         status = VTSessionSetProperty(vtctx->session,
1244                                       kCVImageBufferGammaLevelKey,
1245                                       gamma_level);
1246
1247         if (status) {
1248             av_log(avctx, AV_LOG_WARNING, "Could not set gamma level: %d\n", status);
1249         }
1250     }
1251
1252     if (!vtctx->has_b_frames) {
1253         status = VTSessionSetProperty(vtctx->session,
1254                                       kVTCompressionPropertyKey_AllowFrameReordering,
1255                                       kCFBooleanFalse);
1256
1257         if (status) {
1258             av_log(avctx, AV_LOG_ERROR, "Error setting 'allow frame reordering' property: %d\n", status);
1259             return AVERROR_EXTERNAL;
1260         }
1261     }
1262
1263     if (vtctx->entropy != VT_ENTROPY_NOT_SET) {
1264         CFStringRef entropy = vtctx->entropy == VT_CABAC ?
1265                                 compat_keys.kVTH264EntropyMode_CABAC:
1266                                 compat_keys.kVTH264EntropyMode_CAVLC;
1267
1268         status = VTSessionSetProperty(vtctx->session,
1269                                       compat_keys.kVTCompressionPropertyKey_H264EntropyMode,
1270                                       entropy);
1271
1272         if (status) {
1273             av_log(avctx, AV_LOG_ERROR, "Error setting entropy property: %d\n", status);
1274         }
1275     }
1276
1277     if (vtctx->realtime) {
1278         status = VTSessionSetProperty(vtctx->session,
1279                                       compat_keys.kVTCompressionPropertyKey_RealTime,
1280                                       kCFBooleanTrue);
1281
1282         if (status) {
1283             av_log(avctx, AV_LOG_ERROR, "Error setting realtime property: %d\n", status);
1284         }
1285     }
1286
1287     status = VTCompressionSessionPrepareToEncodeFrames(vtctx->session);
1288     if (status) {
1289         av_log(avctx, AV_LOG_ERROR, "Error: cannot prepare encoder: %d\n", status);
1290         return AVERROR_EXTERNAL;
1291     }
1292
1293     return 0;
1294 }
1295
1296 static int vtenc_configure_encoder(AVCodecContext *avctx)
1297 {
1298     CFMutableDictionaryRef enc_info;
1299     CFMutableDictionaryRef pixel_buffer_info;
1300     CMVideoCodecType       codec_type;
1301     VTEncContext           *vtctx = avctx->priv_data;
1302     CFStringRef            profile_level;
1303     CFNumberRef            gamma_level = NULL;
1304     int                    status;
1305
1306     codec_type = get_cm_codec_type(avctx->codec_id);
1307     if (!codec_type) {
1308         av_log(avctx, AV_LOG_ERROR, "Error: no mapping for AVCodecID %d\n", avctx->codec_id);
1309         return AVERROR(EINVAL);
1310     }
1311
1312     vtctx->codec_id = avctx->codec_id;
1313
1314     if (vtctx->codec_id == AV_CODEC_ID_H264) {
1315         vtctx->get_param_set_func = CMVideoFormatDescriptionGetH264ParameterSetAtIndex;
1316
1317         vtctx->has_b_frames = avctx->max_b_frames > 0;
1318         if(vtctx->has_b_frames && vtctx->profile == H264_PROF_BASELINE){
1319             av_log(avctx, AV_LOG_WARNING, "Cannot use B-frames with baseline profile. Output will not contain B-frames.\n");
1320             vtctx->has_b_frames = false;
1321         }
1322
1323         if (vtctx->entropy == VT_CABAC && vtctx->profile == H264_PROF_BASELINE) {
1324             av_log(avctx, AV_LOG_WARNING, "CABAC entropy requires 'main' or 'high' profile, but baseline was requested. Encode will not use CABAC entropy.\n");
1325             vtctx->entropy = VT_ENTROPY_NOT_SET;
1326         }
1327
1328         if (!get_vt_h264_profile_level(avctx, &profile_level)) return AVERROR(EINVAL);
1329     } else {
1330         vtctx->get_param_set_func = compat_keys.CMVideoFormatDescriptionGetHEVCParameterSetAtIndex;
1331         if (!vtctx->get_param_set_func) return AVERROR(EINVAL);
1332         if (!get_vt_hevc_profile_level(avctx, &profile_level)) return AVERROR(EINVAL);
1333     }
1334
1335     enc_info = CFDictionaryCreateMutable(
1336         kCFAllocatorDefault,
1337         20,
1338         &kCFCopyStringDictionaryKeyCallBacks,
1339         &kCFTypeDictionaryValueCallBacks
1340     );
1341
1342     if (!enc_info) return AVERROR(ENOMEM);
1343
1344 #if !TARGET_OS_IPHONE
1345     if(vtctx->require_sw) {
1346         CFDictionarySetValue(enc_info,
1347                              compat_keys.kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder,
1348                              kCFBooleanFalse);
1349     } else if (!vtctx->allow_sw) {
1350         CFDictionarySetValue(enc_info,
1351                              compat_keys.kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder,
1352                              kCFBooleanTrue);
1353     } else {
1354         CFDictionarySetValue(enc_info,
1355                              compat_keys.kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder,
1356                              kCFBooleanTrue);
1357     }
1358 #endif
1359
1360     if (avctx->pix_fmt != AV_PIX_FMT_VIDEOTOOLBOX) {
1361         status = create_cv_pixel_buffer_info(avctx, &pixel_buffer_info);
1362         if (status)
1363             goto init_cleanup;
1364     } else {
1365         pixel_buffer_info = NULL;
1366     }
1367
1368     vtctx->dts_delta = vtctx->has_b_frames ? -1 : 0;
1369
1370     get_cv_transfer_function(avctx, &vtctx->transfer_function, &gamma_level);
1371     get_cv_ycbcr_matrix(avctx, &vtctx->ycbcr_matrix);
1372     get_cv_color_primaries(avctx, &vtctx->color_primaries);
1373
1374
1375     if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
1376         status = vtenc_populate_extradata(avctx,
1377                                           codec_type,
1378                                           profile_level,
1379                                           gamma_level,
1380                                           enc_info,
1381                                           pixel_buffer_info);
1382         if (status)
1383             goto init_cleanup;
1384     }
1385
1386     status = vtenc_create_encoder(avctx,
1387                                   codec_type,
1388                                   profile_level,
1389                                   gamma_level,
1390                                   enc_info,
1391                                   pixel_buffer_info,
1392                                   &vtctx->session);
1393
1394 init_cleanup:
1395     if (gamma_level)
1396         CFRelease(gamma_level);
1397
1398     if (pixel_buffer_info)
1399         CFRelease(pixel_buffer_info);
1400
1401     CFRelease(enc_info);
1402
1403     return status;
1404 }
1405
1406 static av_cold int vtenc_init(AVCodecContext *avctx)
1407 {
1408     VTEncContext    *vtctx = avctx->priv_data;
1409     CFBooleanRef    has_b_frames_cfbool;
1410     int             status;
1411
1412     pthread_once(&once_ctrl, loadVTEncSymbols);
1413
1414     pthread_mutex_init(&vtctx->lock, NULL);
1415     pthread_cond_init(&vtctx->cv_sample_sent, NULL);
1416
1417     vtctx->session = NULL;
1418     status = vtenc_configure_encoder(avctx);
1419     if (status) return status;
1420
1421     status = VTSessionCopyProperty(vtctx->session,
1422                                    kVTCompressionPropertyKey_AllowFrameReordering,
1423                                    kCFAllocatorDefault,
1424                                    &has_b_frames_cfbool);
1425
1426     if (!status && has_b_frames_cfbool) {
1427         //Some devices don't output B-frames for main profile, even if requested.
1428         vtctx->has_b_frames = CFBooleanGetValue(has_b_frames_cfbool);
1429         CFRelease(has_b_frames_cfbool);
1430     }
1431     avctx->has_b_frames = vtctx->has_b_frames;
1432
1433     return 0;
1434 }
1435
1436 static void vtenc_get_frame_info(CMSampleBufferRef buffer, bool *is_key_frame)
1437 {
1438     CFArrayRef      attachments;
1439     CFDictionaryRef attachment;
1440     CFBooleanRef    not_sync;
1441     CFIndex         len;
1442
1443     attachments = CMSampleBufferGetSampleAttachmentsArray(buffer, false);
1444     len = !attachments ? 0 : CFArrayGetCount(attachments);
1445
1446     if (!len) {
1447         *is_key_frame = true;
1448         return;
1449     }
1450
1451     attachment = CFArrayGetValueAtIndex(attachments, 0);
1452
1453     if (CFDictionaryGetValueIfPresent(attachment,
1454                                       kCMSampleAttachmentKey_NotSync,
1455                                       (const void **)&not_sync))
1456     {
1457         *is_key_frame = !CFBooleanGetValue(not_sync);
1458     } else {
1459         *is_key_frame = true;
1460     }
1461 }
1462
1463 static int is_post_sei_nal_type(int nal_type){
1464     return nal_type != H264_NAL_SEI &&
1465            nal_type != H264_NAL_SPS &&
1466            nal_type != H264_NAL_PPS &&
1467            nal_type != H264_NAL_AUD;
1468 }
1469
1470 /*
1471  * Finds the sei message start/size of type find_sei_type.
1472  * If more than one of that type exists, the last one is returned.
1473  */
1474 static int find_sei_end(AVCodecContext *avctx,
1475                         uint8_t        *nal_data,
1476                         size_t          nal_size,
1477                         uint8_t       **sei_end)
1478 {
1479     int nal_type;
1480     size_t sei_payload_size = 0;
1481     int sei_payload_type = 0;
1482     *sei_end = NULL;
1483     uint8_t *nal_start = nal_data;
1484
1485     if (!nal_size)
1486         return 0;
1487
1488     nal_type = *nal_data & 0x1F;
1489     if (nal_type != H264_NAL_SEI)
1490         return 0;
1491
1492     nal_data++;
1493     nal_size--;
1494
1495     if (nal_data[nal_size - 1] == 0x80)
1496         nal_size--;
1497
1498     while (nal_size > 0 && *nal_data > 0) {
1499         do{
1500             sei_payload_type += *nal_data;
1501             nal_data++;
1502             nal_size--;
1503         } while (nal_size > 0 && *nal_data == 0xFF);
1504
1505         if (!nal_size) {
1506             av_log(avctx, AV_LOG_ERROR, "Unexpected end of SEI NAL Unit parsing type.\n");
1507             return AVERROR_INVALIDDATA;
1508         }
1509
1510         do{
1511             sei_payload_size += *nal_data;
1512             nal_data++;
1513             nal_size--;
1514         } while (nal_size > 0 && *nal_data == 0xFF);
1515
1516         if (nal_size < sei_payload_size) {
1517             av_log(avctx, AV_LOG_ERROR, "Unexpected end of SEI NAL Unit parsing size.\n");
1518             return AVERROR_INVALIDDATA;
1519         }
1520
1521         nal_data += sei_payload_size;
1522         nal_size -= sei_payload_size;
1523     }
1524
1525     *sei_end = nal_data;
1526
1527     return nal_data - nal_start + 1;
1528 }
1529
1530 /**
1531  * Copies the data inserting emulation prevention bytes as needed.
1532  * Existing data in the destination can be taken into account by providing
1533  * dst with a dst_offset > 0.
1534  *
1535  * @return The number of bytes copied on success. On failure, the negative of
1536  *         the number of bytes needed to copy src is returned.
1537  */
1538 static int copy_emulation_prev(const uint8_t *src,
1539                                size_t         src_size,
1540                                uint8_t       *dst,
1541                                ssize_t        dst_offset,
1542                                size_t         dst_size)
1543 {
1544     int zeros = 0;
1545     int wrote_bytes;
1546     uint8_t* dst_start;
1547     uint8_t* dst_end = dst + dst_size;
1548     const uint8_t* src_end = src + src_size;
1549     int start_at = dst_offset > 2 ? dst_offset - 2 : 0;
1550     int i;
1551     for (i = start_at; i < dst_offset && i < dst_size; i++) {
1552         if (!dst[i])
1553             zeros++;
1554         else
1555             zeros = 0;
1556     }
1557
1558     dst += dst_offset;
1559     dst_start = dst;
1560     for (; src < src_end; src++, dst++) {
1561         if (zeros == 2) {
1562             int insert_ep3_byte = *src <= 3;
1563             if (insert_ep3_byte) {
1564                 if (dst < dst_end)
1565                     *dst = 3;
1566                 dst++;
1567             }
1568
1569             zeros = 0;
1570         }
1571
1572         if (dst < dst_end)
1573             *dst = *src;
1574
1575         if (!*src)
1576             zeros++;
1577         else
1578             zeros = 0;
1579     }
1580
1581     wrote_bytes = dst - dst_start;
1582
1583     if (dst > dst_end)
1584         return -wrote_bytes;
1585
1586     return wrote_bytes;
1587 }
1588
1589 static int write_sei(const ExtraSEI *sei,
1590                      int             sei_type,
1591                      uint8_t        *dst,
1592                      size_t          dst_size)
1593 {
1594     uint8_t *sei_start = dst;
1595     size_t remaining_sei_size = sei->size;
1596     size_t remaining_dst_size = dst_size;
1597     int header_bytes;
1598     int bytes_written;
1599     ssize_t offset;
1600
1601     if (!remaining_dst_size)
1602         return AVERROR_BUFFER_TOO_SMALL;
1603
1604     while (sei_type && remaining_dst_size != 0) {
1605         int sei_byte = sei_type > 255 ? 255 : sei_type;
1606         *dst = sei_byte;
1607
1608         sei_type -= sei_byte;
1609         dst++;
1610         remaining_dst_size--;
1611     }
1612
1613     if (!dst_size)
1614         return AVERROR_BUFFER_TOO_SMALL;
1615
1616     while (remaining_sei_size && remaining_dst_size != 0) {
1617         int size_byte = remaining_sei_size > 255 ? 255 : remaining_sei_size;
1618         *dst = size_byte;
1619
1620         remaining_sei_size -= size_byte;
1621         dst++;
1622         remaining_dst_size--;
1623     }
1624
1625     if (remaining_dst_size < sei->size)
1626         return AVERROR_BUFFER_TOO_SMALL;
1627
1628     header_bytes = dst - sei_start;
1629
1630     offset = header_bytes;
1631     bytes_written = copy_emulation_prev(sei->data,
1632                                         sei->size,
1633                                         sei_start,
1634                                         offset,
1635                                         dst_size);
1636     if (bytes_written < 0)
1637         return AVERROR_BUFFER_TOO_SMALL;
1638
1639     bytes_written += header_bytes;
1640     return bytes_written;
1641 }
1642
1643 /**
1644  * Copies NAL units and replaces length codes with
1645  * H.264 Annex B start codes. On failure, the contents of
1646  * dst_data may have been modified.
1647  *
1648  * @param length_code_size Byte length of each length code
1649  * @param sample_buffer NAL units prefixed with length codes.
1650  * @param sei Optional A53 closed captions SEI data.
1651  * @param dst_data Must be zeroed before calling this function.
1652  *                 Contains the copied NAL units prefixed with
1653  *                 start codes when the function returns
1654  *                 successfully.
1655  * @param dst_size Length of dst_data
1656  * @return 0 on success
1657  *         AVERROR_INVALIDDATA if length_code_size is invalid
1658  *         AVERROR_BUFFER_TOO_SMALL if dst_data is too small
1659  *         or if a length_code in src_data specifies data beyond
1660  *         the end of its buffer.
1661  */
1662 static int copy_replace_length_codes(
1663     AVCodecContext *avctx,
1664     size_t        length_code_size,
1665     CMSampleBufferRef sample_buffer,
1666     ExtraSEI      *sei,
1667     uint8_t       *dst_data,
1668     size_t        dst_size)
1669 {
1670     size_t src_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
1671     size_t remaining_src_size = src_size;
1672     size_t remaining_dst_size = dst_size;
1673     size_t src_offset = 0;
1674     int wrote_sei = 0;
1675     int status;
1676     uint8_t size_buf[4];
1677     uint8_t nal_type;
1678     CMBlockBufferRef block = CMSampleBufferGetDataBuffer(sample_buffer);
1679
1680     if (length_code_size > 4) {
1681         return AVERROR_INVALIDDATA;
1682     }
1683
1684     while (remaining_src_size > 0) {
1685         size_t curr_src_len;
1686         size_t curr_dst_len;
1687         size_t box_len = 0;
1688         size_t i;
1689
1690         uint8_t       *dst_box;
1691
1692         status = CMBlockBufferCopyDataBytes(block,
1693                                             src_offset,
1694                                             length_code_size,
1695                                             size_buf);
1696         if (status) {
1697             av_log(avctx, AV_LOG_ERROR, "Cannot copy length: %d\n", status);
1698             return AVERROR_EXTERNAL;
1699         }
1700
1701         status = CMBlockBufferCopyDataBytes(block,
1702                                             src_offset + length_code_size,
1703                                             1,
1704                                             &nal_type);
1705
1706         if (status) {
1707             av_log(avctx, AV_LOG_ERROR, "Cannot copy type: %d\n", status);
1708             return AVERROR_EXTERNAL;
1709         }
1710
1711         nal_type &= 0x1F;
1712
1713         for (i = 0; i < length_code_size; i++) {
1714             box_len <<= 8;
1715             box_len |= size_buf[i];
1716         }
1717
1718         if (sei && !wrote_sei && is_post_sei_nal_type(nal_type)) {
1719             //No SEI NAL unit - insert.
1720             int wrote_bytes;
1721
1722             memcpy(dst_data, start_code, sizeof(start_code));
1723             dst_data += sizeof(start_code);
1724             remaining_dst_size -= sizeof(start_code);
1725
1726             *dst_data = H264_NAL_SEI;
1727             dst_data++;
1728             remaining_dst_size--;
1729
1730             wrote_bytes = write_sei(sei,
1731                                     H264_SEI_TYPE_USER_DATA_REGISTERED,
1732                                     dst_data,
1733                                     remaining_dst_size);
1734
1735             if (wrote_bytes < 0)
1736                 return wrote_bytes;
1737
1738             remaining_dst_size -= wrote_bytes;
1739             dst_data += wrote_bytes;
1740
1741             if (remaining_dst_size <= 0)
1742                 return AVERROR_BUFFER_TOO_SMALL;
1743
1744             *dst_data = 0x80;
1745
1746             dst_data++;
1747             remaining_dst_size--;
1748
1749             wrote_sei = 1;
1750         }
1751
1752         curr_src_len = box_len + length_code_size;
1753         curr_dst_len = box_len + sizeof(start_code);
1754
1755         if (remaining_src_size < curr_src_len) {
1756             return AVERROR_BUFFER_TOO_SMALL;
1757         }
1758
1759         if (remaining_dst_size < curr_dst_len) {
1760             return AVERROR_BUFFER_TOO_SMALL;
1761         }
1762
1763         dst_box = dst_data + sizeof(start_code);
1764
1765         memcpy(dst_data, start_code, sizeof(start_code));
1766         status = CMBlockBufferCopyDataBytes(block,
1767                                             src_offset + length_code_size,
1768                                             box_len,
1769                                             dst_box);
1770
1771         if (status) {
1772             av_log(avctx, AV_LOG_ERROR, "Cannot copy data: %d\n", status);
1773             return AVERROR_EXTERNAL;
1774         }
1775
1776         if (sei && !wrote_sei && nal_type == H264_NAL_SEI) {
1777             //Found SEI NAL unit - append.
1778             int wrote_bytes;
1779             int old_sei_length;
1780             int extra_bytes;
1781             uint8_t *new_sei;
1782             old_sei_length = find_sei_end(avctx, dst_box, box_len, &new_sei);
1783             if (old_sei_length < 0)
1784                 return status;
1785
1786             wrote_bytes = write_sei(sei,
1787                                     H264_SEI_TYPE_USER_DATA_REGISTERED,
1788                                     new_sei,
1789                                     remaining_dst_size - old_sei_length);
1790             if (wrote_bytes < 0)
1791                 return wrote_bytes;
1792
1793             if (new_sei + wrote_bytes >= dst_data + remaining_dst_size)
1794                 return AVERROR_BUFFER_TOO_SMALL;
1795
1796             new_sei[wrote_bytes++] = 0x80;
1797             extra_bytes = wrote_bytes - (dst_box + box_len - new_sei);
1798
1799             dst_data += extra_bytes;
1800             remaining_dst_size -= extra_bytes;
1801
1802             wrote_sei = 1;
1803         }
1804
1805         src_offset += curr_src_len;
1806         dst_data += curr_dst_len;
1807
1808         remaining_src_size -= curr_src_len;
1809         remaining_dst_size -= curr_dst_len;
1810     }
1811
1812     return 0;
1813 }
1814
1815 /**
1816  * Returns a sufficient number of bytes to contain the sei data.
1817  * It may be greater than the minimum required.
1818  */
1819 static int get_sei_msg_bytes(const ExtraSEI* sei, int type){
1820     int copied_size;
1821     if (sei->size == 0)
1822         return 0;
1823
1824     copied_size = -copy_emulation_prev(sei->data,
1825                                        sei->size,
1826                                        NULL,
1827                                        0,
1828                                        0);
1829
1830     if ((sei->size % 255) == 0) //may result in an extra byte
1831         copied_size++;
1832
1833     return copied_size + sei->size / 255 + 1 + type / 255 + 1;
1834 }
1835
1836 static int vtenc_cm_to_avpacket(
1837     AVCodecContext    *avctx,
1838     CMSampleBufferRef sample_buffer,
1839     AVPacket          *pkt,
1840     ExtraSEI          *sei)
1841 {
1842     VTEncContext *vtctx = avctx->priv_data;
1843
1844     int     status;
1845     bool    is_key_frame;
1846     bool    add_header;
1847     size_t  length_code_size;
1848     size_t  header_size = 0;
1849     size_t  in_buf_size;
1850     size_t  out_buf_size;
1851     size_t  sei_nalu_size = 0;
1852     int64_t dts_delta;
1853     int64_t time_base_num;
1854     int nalu_count;
1855     CMTime  pts;
1856     CMTime  dts;
1857     CMVideoFormatDescriptionRef vid_fmt;
1858
1859
1860     vtenc_get_frame_info(sample_buffer, &is_key_frame);
1861     status = get_length_code_size(avctx, sample_buffer, &length_code_size);
1862     if (status) return status;
1863
1864     add_header = is_key_frame && !(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER);
1865
1866     if (add_header) {
1867         vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
1868         if (!vid_fmt) {
1869             av_log(avctx, AV_LOG_ERROR, "Cannot get format description.\n");
1870             return AVERROR_EXTERNAL;
1871         }
1872
1873         int status = get_params_size(avctx, vid_fmt, &header_size);
1874         if (status) return status;
1875     }
1876
1877     status = count_nalus(length_code_size, sample_buffer, &nalu_count);
1878     if(status)
1879         return status;
1880
1881     if (sei) {
1882         size_t msg_size = get_sei_msg_bytes(sei,
1883                                             H264_SEI_TYPE_USER_DATA_REGISTERED);
1884
1885         sei_nalu_size = sizeof(start_code) + 1 + msg_size + 1;
1886     }
1887
1888     in_buf_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
1889     out_buf_size = header_size +
1890                    in_buf_size +
1891                    sei_nalu_size +
1892                    nalu_count * ((int)sizeof(start_code) - (int)length_code_size);
1893
1894     status = ff_alloc_packet2(avctx, pkt, out_buf_size, out_buf_size);
1895     if (status < 0)
1896         return status;
1897
1898     if (add_header) {
1899         status = copy_param_sets(avctx, vid_fmt, pkt->data, out_buf_size);
1900         if(status) return status;
1901     }
1902
1903     status = copy_replace_length_codes(
1904         avctx,
1905         length_code_size,
1906         sample_buffer,
1907         sei,
1908         pkt->data + header_size,
1909         pkt->size - header_size
1910     );
1911
1912     if (status) {
1913         av_log(avctx, AV_LOG_ERROR, "Error copying packet data: %d\n", status);
1914         return status;
1915     }
1916
1917     if (is_key_frame) {
1918         pkt->flags |= AV_PKT_FLAG_KEY;
1919     }
1920
1921     pts = CMSampleBufferGetPresentationTimeStamp(sample_buffer);
1922     dts = CMSampleBufferGetDecodeTimeStamp      (sample_buffer);
1923
1924     if (CMTIME_IS_INVALID(dts)) {
1925         if (!vtctx->has_b_frames) {
1926             dts = pts;
1927         } else {
1928             av_log(avctx, AV_LOG_ERROR, "DTS is invalid.\n");
1929             return AVERROR_EXTERNAL;
1930         }
1931     }
1932
1933     dts_delta = vtctx->dts_delta >= 0 ? vtctx->dts_delta : 0;
1934     time_base_num = avctx->time_base.num;
1935     pkt->pts = pts.value / time_base_num;
1936     pkt->dts = dts.value / time_base_num - dts_delta;
1937     pkt->size = out_buf_size;
1938
1939     return 0;
1940 }
1941
1942 /*
1943  * contiguous_buf_size is 0 if not contiguous, and the size of the buffer
1944  * containing all planes if so.
1945  */
1946 static int get_cv_pixel_info(
1947     AVCodecContext *avctx,
1948     const AVFrame  *frame,
1949     int            *color,
1950     int            *plane_count,
1951     size_t         *widths,
1952     size_t         *heights,
1953     size_t         *strides,
1954     size_t         *contiguous_buf_size)
1955 {
1956     VTEncContext *vtctx = avctx->priv_data;
1957     int av_format       = frame->format;
1958     int av_color_range  = frame->color_range;
1959     int i;
1960     int range_guessed;
1961     int status;
1962
1963     status = get_cv_pixel_format(avctx, av_format, av_color_range, color, &range_guessed);
1964     if (status) {
1965         av_log(avctx,
1966             AV_LOG_ERROR,
1967             "Could not get pixel format for color format '%s' range '%s'.\n",
1968             av_get_pix_fmt_name(av_format),
1969             av_color_range > AVCOL_RANGE_UNSPECIFIED &&
1970             av_color_range < AVCOL_RANGE_NB ?
1971                av_color_range_name(av_color_range) :
1972                "Unknown");
1973
1974         return AVERROR(EINVAL);
1975     }
1976
1977     if (range_guessed) {
1978         if (!vtctx->warned_color_range) {
1979             vtctx->warned_color_range = true;
1980             av_log(avctx,
1981                    AV_LOG_WARNING,
1982                    "Color range not set for %s. Using MPEG range.\n",
1983                    av_get_pix_fmt_name(av_format));
1984         }
1985     }
1986
1987     switch (av_format) {
1988     case AV_PIX_FMT_NV12:
1989         *plane_count = 2;
1990
1991         widths [0] = avctx->width;
1992         heights[0] = avctx->height;
1993         strides[0] = frame ? frame->linesize[0] : avctx->width;
1994
1995         widths [1] = (avctx->width  + 1) / 2;
1996         heights[1] = (avctx->height + 1) / 2;
1997         strides[1] = frame ? frame->linesize[1] : (avctx->width + 1) & -2;
1998         break;
1999
2000     case AV_PIX_FMT_YUV420P:
2001         *plane_count = 3;
2002
2003         widths [0] = avctx->width;
2004         heights[0] = avctx->height;
2005         strides[0] = frame ? frame->linesize[0] : avctx->width;
2006
2007         widths [1] = (avctx->width  + 1) / 2;
2008         heights[1] = (avctx->height + 1) / 2;
2009         strides[1] = frame ? frame->linesize[1] : (avctx->width + 1) / 2;
2010
2011         widths [2] = (avctx->width  + 1) / 2;
2012         heights[2] = (avctx->height + 1) / 2;
2013         strides[2] = frame ? frame->linesize[2] : (avctx->width + 1) / 2;
2014         break;
2015
2016     case AV_PIX_FMT_P010LE:
2017         *plane_count = 2;
2018         widths[0] = avctx->width;
2019         heights[0] = avctx->height;
2020         strides[0] = frame ? frame->linesize[0] : (avctx->width * 2 + 63) & -64;
2021
2022         widths[1] = (avctx->width + 1) / 2;
2023         heights[1] = (avctx->height + 1) / 2;
2024         strides[1] = frame ? frame->linesize[1] : ((avctx->width + 1) / 2 + 63) & -64;
2025         break;
2026
2027     default:
2028         av_log(
2029                avctx,
2030                AV_LOG_ERROR,
2031                "Could not get frame format info for color %d range %d.\n",
2032                av_format,
2033                av_color_range);
2034
2035         return AVERROR(EINVAL);
2036     }
2037
2038     *contiguous_buf_size = 0;
2039     for (i = 0; i < *plane_count; i++) {
2040         if (i < *plane_count - 1 &&
2041             frame->data[i] + strides[i] * heights[i] != frame->data[i + 1]) {
2042             *contiguous_buf_size = 0;
2043             break;
2044         }
2045
2046         *contiguous_buf_size += strides[i] * heights[i];
2047     }
2048
2049     return 0;
2050 }
2051
2052 //Not used on OSX - frame is never copied.
2053 static int copy_avframe_to_pixel_buffer(AVCodecContext   *avctx,
2054                                         const AVFrame    *frame,
2055                                         CVPixelBufferRef cv_img,
2056                                         const size_t     *plane_strides,
2057                                         const size_t     *plane_rows)
2058 {
2059     int i, j;
2060     size_t plane_count;
2061     int status;
2062     int rows;
2063     int src_stride;
2064     int dst_stride;
2065     uint8_t *src_addr;
2066     uint8_t *dst_addr;
2067     size_t copy_bytes;
2068
2069     status = CVPixelBufferLockBaseAddress(cv_img, 0);
2070     if (status) {
2071         av_log(
2072             avctx,
2073             AV_LOG_ERROR,
2074             "Error: Could not lock base address of CVPixelBuffer: %d.\n",
2075             status
2076         );
2077     }
2078
2079     if (CVPixelBufferIsPlanar(cv_img)) {
2080         plane_count = CVPixelBufferGetPlaneCount(cv_img);
2081         for (i = 0; frame->data[i]; i++) {
2082             if (i == plane_count) {
2083                 CVPixelBufferUnlockBaseAddress(cv_img, 0);
2084                 av_log(avctx,
2085                     AV_LOG_ERROR,
2086                     "Error: different number of planes in AVFrame and CVPixelBuffer.\n"
2087                 );
2088
2089                 return AVERROR_EXTERNAL;
2090             }
2091
2092             dst_addr = (uint8_t*)CVPixelBufferGetBaseAddressOfPlane(cv_img, i);
2093             src_addr = (uint8_t*)frame->data[i];
2094             dst_stride = CVPixelBufferGetBytesPerRowOfPlane(cv_img, i);
2095             src_stride = plane_strides[i];
2096             rows = plane_rows[i];
2097
2098             if (dst_stride == src_stride) {
2099                 memcpy(dst_addr, src_addr, src_stride * rows);
2100             } else {
2101                 copy_bytes = dst_stride < src_stride ? dst_stride : src_stride;
2102
2103                 for (j = 0; j < rows; j++) {
2104                     memcpy(dst_addr + j * dst_stride, src_addr + j * src_stride, copy_bytes);
2105                 }
2106             }
2107         }
2108     } else {
2109         if (frame->data[1]) {
2110             CVPixelBufferUnlockBaseAddress(cv_img, 0);
2111             av_log(avctx,
2112                 AV_LOG_ERROR,
2113                 "Error: different number of planes in AVFrame and non-planar CVPixelBuffer.\n"
2114             );
2115
2116             return AVERROR_EXTERNAL;
2117         }
2118
2119         dst_addr = (uint8_t*)CVPixelBufferGetBaseAddress(cv_img);
2120         src_addr = (uint8_t*)frame->data[0];
2121         dst_stride = CVPixelBufferGetBytesPerRow(cv_img);
2122         src_stride = plane_strides[0];
2123         rows = plane_rows[0];
2124
2125         if (dst_stride == src_stride) {
2126             memcpy(dst_addr, src_addr, src_stride * rows);
2127         } else {
2128             copy_bytes = dst_stride < src_stride ? dst_stride : src_stride;
2129
2130             for (j = 0; j < rows; j++) {
2131                 memcpy(dst_addr + j * dst_stride, src_addr + j * src_stride, copy_bytes);
2132             }
2133         }
2134     }
2135
2136     status = CVPixelBufferUnlockBaseAddress(cv_img, 0);
2137     if (status) {
2138         av_log(avctx, AV_LOG_ERROR, "Error: Could not unlock CVPixelBuffer base address: %d.\n", status);
2139         return AVERROR_EXTERNAL;
2140     }
2141
2142     return 0;
2143 }
2144
2145 static int create_cv_pixel_buffer(AVCodecContext   *avctx,
2146                                   const AVFrame    *frame,
2147                                   CVPixelBufferRef *cv_img)
2148 {
2149     int plane_count;
2150     int color;
2151     size_t widths [AV_NUM_DATA_POINTERS];
2152     size_t heights[AV_NUM_DATA_POINTERS];
2153     size_t strides[AV_NUM_DATA_POINTERS];
2154     int status;
2155     size_t contiguous_buf_size;
2156     CVPixelBufferPoolRef pix_buf_pool;
2157     VTEncContext* vtctx = avctx->priv_data;
2158
2159     if (avctx->pix_fmt == AV_PIX_FMT_VIDEOTOOLBOX) {
2160         av_assert0(frame->format == AV_PIX_FMT_VIDEOTOOLBOX);
2161
2162         *cv_img = (CVPixelBufferRef)frame->data[3];
2163         av_assert0(*cv_img);
2164
2165         CFRetain(*cv_img);
2166         return 0;
2167     }
2168
2169     memset(widths,  0, sizeof(widths));
2170     memset(heights, 0, sizeof(heights));
2171     memset(strides, 0, sizeof(strides));
2172
2173     status = get_cv_pixel_info(
2174         avctx,
2175         frame,
2176         &color,
2177         &plane_count,
2178         widths,
2179         heights,
2180         strides,
2181         &contiguous_buf_size
2182     );
2183
2184     if (status) {
2185         av_log(
2186             avctx,
2187             AV_LOG_ERROR,
2188             "Error: Cannot convert format %d color_range %d: %d\n",
2189             frame->format,
2190             frame->color_range,
2191             status
2192         );
2193
2194         return AVERROR_EXTERNAL;
2195     }
2196
2197     pix_buf_pool = VTCompressionSessionGetPixelBufferPool(vtctx->session);
2198     if (!pix_buf_pool) {
2199         /* On iOS, the VT session is invalidated when the APP switches from
2200          * foreground to background and vice versa. Fetch the actual error code
2201          * of the VT session to detect that case and restart the VT session
2202          * accordingly. */
2203         OSStatus vtstatus;
2204
2205         vtstatus = VTCompressionSessionPrepareToEncodeFrames(vtctx->session);
2206         if (vtstatus == kVTInvalidSessionErr) {
2207             CFRelease(vtctx->session);
2208             vtctx->session = NULL;
2209             status = vtenc_configure_encoder(avctx);
2210             if (status == 0)
2211                 pix_buf_pool = VTCompressionSessionGetPixelBufferPool(vtctx->session);
2212         }
2213         if (!pix_buf_pool) {
2214             av_log(avctx, AV_LOG_ERROR, "Could not get pixel buffer pool.\n");
2215             return AVERROR_EXTERNAL;
2216         }
2217         else
2218             av_log(avctx, AV_LOG_WARNING, "VT session restarted because of a "
2219                    "kVTInvalidSessionErr error.\n");
2220     }
2221
2222     status = CVPixelBufferPoolCreatePixelBuffer(NULL,
2223                                                 pix_buf_pool,
2224                                                 cv_img);
2225
2226
2227     if (status) {
2228         av_log(avctx, AV_LOG_ERROR, "Could not create pixel buffer from pool: %d.\n", status);
2229         return AVERROR_EXTERNAL;
2230     }
2231
2232     status = copy_avframe_to_pixel_buffer(avctx, frame, *cv_img, strides, heights);
2233     if (status) {
2234         CFRelease(*cv_img);
2235         *cv_img = NULL;
2236         return status;
2237     }
2238
2239     return 0;
2240 }
2241
2242 static int create_encoder_dict_h264(const AVFrame *frame,
2243                                     CFDictionaryRef* dict_out)
2244 {
2245     CFDictionaryRef dict = NULL;
2246     if (frame->pict_type == AV_PICTURE_TYPE_I) {
2247         const void *keys[] = { kVTEncodeFrameOptionKey_ForceKeyFrame };
2248         const void *vals[] = { kCFBooleanTrue };
2249
2250         dict = CFDictionaryCreate(NULL, keys, vals, 1, NULL, NULL);
2251         if(!dict) return AVERROR(ENOMEM);
2252     }
2253
2254     *dict_out = dict;
2255     return 0;
2256 }
2257
2258 static int vtenc_send_frame(AVCodecContext *avctx,
2259                             VTEncContext   *vtctx,
2260                             const AVFrame  *frame)
2261 {
2262     CMTime time;
2263     CFDictionaryRef frame_dict;
2264     CVPixelBufferRef cv_img = NULL;
2265     AVFrameSideData *side_data = NULL;
2266     ExtraSEI *sei = NULL;
2267     int status = create_cv_pixel_buffer(avctx, frame, &cv_img);
2268
2269     if (status) return status;
2270
2271     status = create_encoder_dict_h264(frame, &frame_dict);
2272     if (status) {
2273         CFRelease(cv_img);
2274         return status;
2275     }
2276
2277     side_data = av_frame_get_side_data(frame, AV_FRAME_DATA_A53_CC);
2278     if (vtctx->a53_cc && side_data && side_data->size) {
2279         sei = av_mallocz(sizeof(*sei));
2280         if (!sei) {
2281             av_log(avctx, AV_LOG_ERROR, "Not enough memory for closed captions, skipping\n");
2282         } else {
2283             int ret = ff_alloc_a53_sei(frame, 0, &sei->data, &sei->size);
2284             if (ret < 0) {
2285                 av_log(avctx, AV_LOG_ERROR, "Not enough memory for closed captions, skipping\n");
2286                 av_free(sei);
2287                 sei = NULL;
2288             }
2289         }
2290     }
2291
2292     time = CMTimeMake(frame->pts * avctx->time_base.num, avctx->time_base.den);
2293     status = VTCompressionSessionEncodeFrame(
2294         vtctx->session,
2295         cv_img,
2296         time,
2297         kCMTimeInvalid,
2298         frame_dict,
2299         sei,
2300         NULL
2301     );
2302
2303     if (frame_dict) CFRelease(frame_dict);
2304     CFRelease(cv_img);
2305
2306     if (status) {
2307         av_log(avctx, AV_LOG_ERROR, "Error: cannot encode frame: %d\n", status);
2308         return AVERROR_EXTERNAL;
2309     }
2310
2311     return 0;
2312 }
2313
2314 static av_cold int vtenc_frame(
2315     AVCodecContext *avctx,
2316     AVPacket       *pkt,
2317     const AVFrame  *frame,
2318     int            *got_packet)
2319 {
2320     VTEncContext *vtctx = avctx->priv_data;
2321     bool get_frame;
2322     int status;
2323     CMSampleBufferRef buf = NULL;
2324     ExtraSEI *sei = NULL;
2325
2326     if (frame) {
2327         status = vtenc_send_frame(avctx, vtctx, frame);
2328
2329         if (status) {
2330             status = AVERROR_EXTERNAL;
2331             goto end_nopkt;
2332         }
2333
2334         if (vtctx->frame_ct_in == 0) {
2335             vtctx->first_pts = frame->pts;
2336         } else if(vtctx->frame_ct_in == 1 && vtctx->has_b_frames) {
2337             vtctx->dts_delta = frame->pts - vtctx->first_pts;
2338         }
2339
2340         vtctx->frame_ct_in++;
2341     } else if(!vtctx->flushing) {
2342         vtctx->flushing = true;
2343
2344         status = VTCompressionSessionCompleteFrames(vtctx->session,
2345                                                     kCMTimeIndefinite);
2346
2347         if (status) {
2348             av_log(avctx, AV_LOG_ERROR, "Error flushing frames: %d\n", status);
2349             status = AVERROR_EXTERNAL;
2350             goto end_nopkt;
2351         }
2352     }
2353
2354     *got_packet = 0;
2355     get_frame = vtctx->dts_delta >= 0 || !frame;
2356     if (!get_frame) {
2357         status = 0;
2358         goto end_nopkt;
2359     }
2360
2361     status = vtenc_q_pop(vtctx, !frame, &buf, &sei);
2362     if (status) goto end_nopkt;
2363     if (!buf)   goto end_nopkt;
2364
2365     status = vtenc_cm_to_avpacket(avctx, buf, pkt, sei);
2366     if (sei) {
2367         if (sei->data) av_free(sei->data);
2368         av_free(sei);
2369     }
2370     CFRelease(buf);
2371     if (status) goto end_nopkt;
2372
2373     *got_packet = 1;
2374     return 0;
2375
2376 end_nopkt:
2377     av_packet_unref(pkt);
2378     return status;
2379 }
2380
2381 static int vtenc_populate_extradata(AVCodecContext   *avctx,
2382                                     CMVideoCodecType codec_type,
2383                                     CFStringRef      profile_level,
2384                                     CFNumberRef      gamma_level,
2385                                     CFDictionaryRef  enc_info,
2386                                     CFDictionaryRef  pixel_buffer_info)
2387 {
2388     VTEncContext *vtctx = avctx->priv_data;
2389     int status;
2390     CVPixelBufferPoolRef pool = NULL;
2391     CVPixelBufferRef pix_buf = NULL;
2392     CMTime time;
2393     CMSampleBufferRef buf = NULL;
2394
2395     status = vtenc_create_encoder(avctx,
2396                                   codec_type,
2397                                   profile_level,
2398                                   gamma_level,
2399                                   enc_info,
2400                                   pixel_buffer_info,
2401                                   &vtctx->session);
2402     if (status)
2403         goto pe_cleanup;
2404
2405     pool = VTCompressionSessionGetPixelBufferPool(vtctx->session);
2406     if(!pool){
2407         av_log(avctx, AV_LOG_ERROR, "Error getting pixel buffer pool.\n");
2408         goto pe_cleanup;
2409     }
2410
2411     status = CVPixelBufferPoolCreatePixelBuffer(NULL,
2412                                                 pool,
2413                                                 &pix_buf);
2414
2415     if(status != kCVReturnSuccess){
2416         av_log(avctx, AV_LOG_ERROR, "Error creating frame from pool: %d\n", status);
2417         goto pe_cleanup;
2418     }
2419
2420     time = CMTimeMake(0, avctx->time_base.den);
2421     status = VTCompressionSessionEncodeFrame(vtctx->session,
2422                                              pix_buf,
2423                                              time,
2424                                              kCMTimeInvalid,
2425                                              NULL,
2426                                              NULL,
2427                                              NULL);
2428
2429     if (status) {
2430         av_log(avctx,
2431                AV_LOG_ERROR,
2432                "Error sending frame for extradata: %d\n",
2433                status);
2434
2435         goto pe_cleanup;
2436     }
2437
2438     //Populates extradata - output frames are flushed and param sets are available.
2439     status = VTCompressionSessionCompleteFrames(vtctx->session,
2440                                                 kCMTimeIndefinite);
2441
2442     if (status)
2443         goto pe_cleanup;
2444
2445     status = vtenc_q_pop(vtctx, 0, &buf, NULL);
2446     if (status) {
2447         av_log(avctx, AV_LOG_ERROR, "popping: %d\n", status);
2448         goto pe_cleanup;
2449     }
2450
2451     CFRelease(buf);
2452
2453
2454
2455 pe_cleanup:
2456     if(vtctx->session)
2457         CFRelease(vtctx->session);
2458
2459     vtctx->session = NULL;
2460     vtctx->frame_ct_out = 0;
2461
2462     av_assert0(status != 0 || (avctx->extradata && avctx->extradata_size > 0));
2463
2464     return status;
2465 }
2466
2467 static av_cold int vtenc_close(AVCodecContext *avctx)
2468 {
2469     VTEncContext *vtctx = avctx->priv_data;
2470
2471     pthread_cond_destroy(&vtctx->cv_sample_sent);
2472     pthread_mutex_destroy(&vtctx->lock);
2473
2474     if(!vtctx->session) return 0;
2475
2476     VTCompressionSessionCompleteFrames(vtctx->session,
2477                                        kCMTimeIndefinite);
2478     clear_frame_queue(vtctx);
2479     CFRelease(vtctx->session);
2480     vtctx->session = NULL;
2481
2482     if (vtctx->color_primaries) {
2483         CFRelease(vtctx->color_primaries);
2484         vtctx->color_primaries = NULL;
2485     }
2486
2487     if (vtctx->transfer_function) {
2488         CFRelease(vtctx->transfer_function);
2489         vtctx->transfer_function = NULL;
2490     }
2491
2492     if (vtctx->ycbcr_matrix) {
2493         CFRelease(vtctx->ycbcr_matrix);
2494         vtctx->ycbcr_matrix = NULL;
2495     }
2496
2497     return 0;
2498 }
2499
2500 static const enum AVPixelFormat avc_pix_fmts[] = {
2501     AV_PIX_FMT_VIDEOTOOLBOX,
2502     AV_PIX_FMT_NV12,
2503     AV_PIX_FMT_YUV420P,
2504     AV_PIX_FMT_NONE
2505 };
2506
2507 static const enum AVPixelFormat hevc_pix_fmts[] = {
2508     AV_PIX_FMT_VIDEOTOOLBOX,
2509     AV_PIX_FMT_NV12,
2510     AV_PIX_FMT_YUV420P,
2511     AV_PIX_FMT_P010LE,
2512     AV_PIX_FMT_NONE
2513 };
2514
2515 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
2516 #define COMMON_OPTIONS \
2517     { "allow_sw", "Allow software encoding", OFFSET(allow_sw), AV_OPT_TYPE_BOOL, \
2518         { .i64 = 0 }, 0, 1, VE }, \
2519     { "require_sw", "Require software encoding", OFFSET(require_sw), AV_OPT_TYPE_BOOL, \
2520         { .i64 = 0 }, 0, 1, VE }, \
2521     { "realtime", "Hint that encoding should happen in real-time if not faster (e.g. capturing from camera).", \
2522         OFFSET(realtime), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, \
2523     { "frames_before", "Other frames will come before the frames in this session. This helps smooth concatenation issues.", \
2524         OFFSET(frames_before), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, \
2525     { "frames_after", "Other frames will come after the frames in this session. This helps smooth concatenation issues.", \
2526         OFFSET(frames_after), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
2527
2528 #define OFFSET(x) offsetof(VTEncContext, x)
2529 static const AVOption h264_options[] = {
2530     { "profile", "Profile", OFFSET(profile), AV_OPT_TYPE_INT, { .i64 = H264_PROF_AUTO }, H264_PROF_AUTO, H264_PROF_COUNT, VE, "profile" },
2531     { "baseline", "Baseline Profile", 0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_BASELINE }, INT_MIN, INT_MAX, VE, "profile" },
2532     { "main",     "Main Profile",     0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_MAIN     }, INT_MIN, INT_MAX, VE, "profile" },
2533     { "high",     "High Profile",     0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_HIGH     }, INT_MIN, INT_MAX, VE, "profile" },
2534
2535     { "level", "Level", OFFSET(level), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 52, VE, "level" },
2536     { "1.3", "Level 1.3, only available with Baseline Profile", 0, AV_OPT_TYPE_CONST, { .i64 = 13 }, INT_MIN, INT_MAX, VE, "level" },
2537     { "3.0", "Level 3.0", 0, AV_OPT_TYPE_CONST, { .i64 = 30 }, INT_MIN, INT_MAX, VE, "level" },
2538     { "3.1", "Level 3.1", 0, AV_OPT_TYPE_CONST, { .i64 = 31 }, INT_MIN, INT_MAX, VE, "level" },
2539     { "3.2", "Level 3.2", 0, AV_OPT_TYPE_CONST, { .i64 = 32 }, INT_MIN, INT_MAX, VE, "level" },
2540     { "4.0", "Level 4.0", 0, AV_OPT_TYPE_CONST, { .i64 = 40 }, INT_MIN, INT_MAX, VE, "level" },
2541     { "4.1", "Level 4.1", 0, AV_OPT_TYPE_CONST, { .i64 = 41 }, INT_MIN, INT_MAX, VE, "level" },
2542     { "4.2", "Level 4.2", 0, AV_OPT_TYPE_CONST, { .i64 = 42 }, INT_MIN, INT_MAX, VE, "level" },
2543     { "5.0", "Level 5.0", 0, AV_OPT_TYPE_CONST, { .i64 = 50 }, INT_MIN, INT_MAX, VE, "level" },
2544     { "5.1", "Level 5.1", 0, AV_OPT_TYPE_CONST, { .i64 = 51 }, INT_MIN, INT_MAX, VE, "level" },
2545     { "5.2", "Level 5.2", 0, AV_OPT_TYPE_CONST, { .i64 = 52 }, INT_MIN, INT_MAX, VE, "level" },
2546
2547     { "coder", "Entropy coding", OFFSET(entropy), AV_OPT_TYPE_INT, { .i64 = VT_ENTROPY_NOT_SET }, VT_ENTROPY_NOT_SET, VT_CABAC, VE, "coder" },
2548     { "cavlc", "CAVLC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CAVLC }, INT_MIN, INT_MAX, VE, "coder" },
2549     { "vlc",   "CAVLC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CAVLC }, INT_MIN, INT_MAX, VE, "coder" },
2550     { "cabac", "CABAC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CABAC }, INT_MIN, INT_MAX, VE, "coder" },
2551     { "ac",    "CABAC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CABAC }, INT_MIN, INT_MAX, VE, "coder" },
2552
2553     { "a53cc", "Use A53 Closed Captions (if available)", OFFSET(a53_cc), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, VE },
2554
2555     COMMON_OPTIONS
2556     { NULL },
2557 };
2558
2559 static const AVClass h264_videotoolbox_class = {
2560     .class_name = "h264_videotoolbox",
2561     .item_name  = av_default_item_name,
2562     .option     = h264_options,
2563     .version    = LIBAVUTIL_VERSION_INT,
2564 };
2565
2566 AVCodec ff_h264_videotoolbox_encoder = {
2567     .name             = "h264_videotoolbox",
2568     .long_name        = NULL_IF_CONFIG_SMALL("VideoToolbox H.264 Encoder"),
2569     .type             = AVMEDIA_TYPE_VIDEO,
2570     .id               = AV_CODEC_ID_H264,
2571     .priv_data_size   = sizeof(VTEncContext),
2572     .pix_fmts         = avc_pix_fmts,
2573     .init             = vtenc_init,
2574     .encode2          = vtenc_frame,
2575     .close            = vtenc_close,
2576     .capabilities     = AV_CODEC_CAP_DELAY,
2577     .priv_class       = &h264_videotoolbox_class,
2578     .caps_internal    = FF_CODEC_CAP_INIT_THREADSAFE |
2579                         FF_CODEC_CAP_INIT_CLEANUP,
2580 };
2581
2582 static const AVOption hevc_options[] = {
2583     { "profile", "Profile", OFFSET(profile), AV_OPT_TYPE_INT, { .i64 = HEVC_PROF_AUTO }, HEVC_PROF_AUTO, HEVC_PROF_COUNT, VE, "profile" },
2584     { "main",     "Main Profile",     0, AV_OPT_TYPE_CONST, { .i64 = HEVC_PROF_MAIN   }, INT_MIN, INT_MAX, VE, "profile" },
2585     { "main10",   "Main10 Profile",   0, AV_OPT_TYPE_CONST, { .i64 = HEVC_PROF_MAIN10 }, INT_MIN, INT_MAX, VE, "profile" },
2586
2587     COMMON_OPTIONS
2588     { NULL },
2589 };
2590
2591 static const AVClass hevc_videotoolbox_class = {
2592     .class_name = "hevc_videotoolbox",
2593     .item_name  = av_default_item_name,
2594     .option     = hevc_options,
2595     .version    = LIBAVUTIL_VERSION_INT,
2596 };
2597
2598 AVCodec ff_hevc_videotoolbox_encoder = {
2599     .name             = "hevc_videotoolbox",
2600     .long_name        = NULL_IF_CONFIG_SMALL("VideoToolbox H.265 Encoder"),
2601     .type             = AVMEDIA_TYPE_VIDEO,
2602     .id               = AV_CODEC_ID_HEVC,
2603     .priv_data_size   = sizeof(VTEncContext),
2604     .pix_fmts         = hevc_pix_fmts,
2605     .init             = vtenc_init,
2606     .encode2          = vtenc_frame,
2607     .close            = vtenc_close,
2608     .capabilities     = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_HARDWARE,
2609     .priv_class       = &hevc_videotoolbox_class,
2610     .caps_internal    = FF_CODEC_CAP_INIT_THREADSAFE |
2611                         FF_CODEC_CAP_INIT_CLEANUP,
2612     .wrapper_name     = "videotoolbox",
2613 };