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) {
 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 (!sample_buffer) {
 579         return;
 580     }
 581
 582     if (!avctx->extradata && (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)) {
 583         int set_status = set_extradata(avctx, sample_buffer);
 584         if (set_status) {
 585             set_async_error(vtctx, set_status);
 586             return;
 587         }
 588     }
 589
 590     vtenc_q_push(vtctx, sample_buffer, sei);
 591 }
 592
 593 static int get_length_code_size(
 594     AVCodecContext    *avctx,
 595     CMSampleBufferRef sample_buffer,
 596     size_t            *size)
 597 {
 598     VTEncContext *vtctx = avctx->priv_data;
 599     CMVideoFormatDescriptionRef vid_fmt;
 600     int isize;
 601     int status;
 602
 603     vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
 604     if (!vid_fmt) {
 605         av_log(avctx, AV_LOG_ERROR, "Error getting buffer format description.\n");
 606         return AVERROR_EXTERNAL;
 607     }
 608
 609     status = vtctx->get_param_set_func(vid_fmt,
 610                                        0,
 611                                        NULL,
 612                                        NULL,
 613                                        NULL,
 614                                        &isize);
 615     if (status) {
 616         av_log(avctx, AV_LOG_ERROR, "Error getting length code size: %d\n", status);
 617         return AVERROR_EXTERNAL;
 618     }
 619
 620     *size = isize;
 621     return 0;
 622 }
 623
 624 /*
 625  * Returns true on success.
 626  *
 627  * If profile_level_val is NULL and this method returns true, don't specify the
 628  * profile/level to the encoder.
 629  */
 630 static bool get_vt_h264_profile_level(AVCodecContext *avctx,
 631                                       CFStringRef    *profile_level_val)
 632 {
 633     VTEncContext *vtctx = avctx->priv_data;
 634     int64_t profile = vtctx->profile;
 635
 636     if (profile == H264_PROF_AUTO && vtctx->level) {
 637         //Need to pick a profile if level is not auto-selected.
 638         profile = vtctx->has_b_frames ? H264_PROF_MAIN : H264_PROF_BASELINE;
 639     }
 640
 641     *profile_level_val = NULL;
 642
 643     switch (profile) {
 644         case H264_PROF_AUTO:
 645             return true;
 646
 647         case H264_PROF_BASELINE:
 648             switch (vtctx->level) {
 649                 case  0: *profile_level_val =
 650                                   compat_keys.kVTProfileLevel_H264_Baseline_AutoLevel; break;
 651                 case 13: *profile_level_val = kVTProfileLevel_H264_Baseline_1_3;       break;
 652                 case 30: *profile_level_val = kVTProfileLevel_H264_Baseline_3_0;       break;
 653                 case 31: *profile_level_val = kVTProfileLevel_H264_Baseline_3_1;       break;
 654                 case 32: *profile_level_val = kVTProfileLevel_H264_Baseline_3_2;       break;
 655                 case 40: *profile_level_val =
 656                                   compat_keys.kVTProfileLevel_H264_Baseline_4_0;       break;
 657                 case 41: *profile_level_val = kVTProfileLevel_H264_Baseline_4_1;       break;
 658                 case 42: *profile_level_val =
 659                                   compat_keys.kVTProfileLevel_H264_Baseline_4_2;       break;
 660                 case 50: *profile_level_val =
 661                                   compat_keys.kVTProfileLevel_H264_Baseline_5_0;       break;
 662                 case 51: *profile_level_val =
 663                                   compat_keys.kVTProfileLevel_H264_Baseline_5_1;       break;
 664                 case 52: *profile_level_val =
 665                                   compat_keys.kVTProfileLevel_H264_Baseline_5_2;       break;
 666             }
 667             break;
 668
 669         case H264_PROF_MAIN:
 670             switch (vtctx->level) {
 671                 case  0: *profile_level_val =
 672                                   compat_keys.kVTProfileLevel_H264_Main_AutoLevel; break;
 673                 case 30: *profile_level_val = kVTProfileLevel_H264_Main_3_0;       break;
 674                 case 31: *profile_level_val = kVTProfileLevel_H264_Main_3_1;       break;
 675                 case 32: *profile_level_val = kVTProfileLevel_H264_Main_3_2;       break;
 676                 case 40: *profile_level_val = kVTProfileLevel_H264_Main_4_0;       break;
 677                 case 41: *profile_level_val = kVTProfileLevel_H264_Main_4_1;       break;
 678                 case 42: *profile_level_val =
 679                                   compat_keys.kVTProfileLevel_H264_Main_4_2;       break;
 680                 case 50: *profile_level_val = kVTProfileLevel_H264_Main_5_0;       break;
 681                 case 51: *profile_level_val =
 682                                   compat_keys.kVTProfileLevel_H264_Main_5_1;       break;
 683                 case 52: *profile_level_val =
 684                                   compat_keys.kVTProfileLevel_H264_Main_5_2;       break;
 685             }
 686             break;
 687
 688         case H264_PROF_HIGH:
 689             switch (vtctx->level) {
 690                 case  0: *profile_level_val =
 691                                   compat_keys.kVTProfileLevel_H264_High_AutoLevel; break;
 692                 case 30: *profile_level_val =
 693                                   compat_keys.kVTProfileLevel_H264_High_3_0;       break;
 694                 case 31: *profile_level_val =
 695                                   compat_keys.kVTProfileLevel_H264_High_3_1;       break;
 696                 case 32: *profile_level_val =
 697                                   compat_keys.kVTProfileLevel_H264_High_3_2;       break;
 698                 case 40: *profile_level_val =
 699                                   compat_keys.kVTProfileLevel_H264_High_4_0;       break;
 700                 case 41: *profile_level_val =
 701                                   compat_keys.kVTProfileLevel_H264_High_4_1;       break;
 702                 case 42: *profile_level_val =
 703                                   compat_keys.kVTProfileLevel_H264_High_4_2;       break;
 704                 case 50: *profile_level_val = kVTProfileLevel_H264_High_5_0;       break;
 705                 case 51: *profile_level_val =
 706                                   compat_keys.kVTProfileLevel_H264_High_5_1;       break;
 707                 case 52: *profile_level_val =
 708                                   compat_keys.kVTProfileLevel_H264_High_5_2;       break;
 709             }
 710             break;
 711     }
 712
 713     if (!*profile_level_val) {
 714         av_log(avctx, AV_LOG_ERROR, "Invalid Profile/Level.\n");
 715         return false;
 716     }
 717
 718     return true;
 719 }
 720
 721 /*
 722  * Returns true on success.
 723  *
 724  * If profile_level_val is NULL and this method returns true, don't specify the
 725  * profile/level to the encoder.
 726  */
 727 static bool get_vt_hevc_profile_level(AVCodecContext *avctx,
 728                                       CFStringRef    *profile_level_val)
 729 {
 730     VTEncContext *vtctx = avctx->priv_data;
 731     int64_t profile = vtctx->profile;
 732
 733     *profile_level_val = NULL;
 734
 735     switch (profile) {
 736         case HEVC_PROF_AUTO:
 737             return true;
 738         case HEVC_PROF_MAIN:
 739             *profile_level_val =
 740                 compat_keys.kVTProfileLevel_HEVC_Main_AutoLevel;
 741             break;
 742         case HEVC_PROF_MAIN10:
 743             *profile_level_val =
 744                 compat_keys.kVTProfileLevel_HEVC_Main10_AutoLevel;
 745             break;
 746     }
 747
 748     if (!*profile_level_val) {
 749         av_log(avctx, AV_LOG_ERROR, "Invalid Profile/Level.\n");
 750         return false;
 751     }
 752
 753     return true;
 754 }
 755
 756 static int get_cv_pixel_format(AVCodecContext* avctx,
 757                                enum AVPixelFormat fmt,
 758                                enum AVColorRange range,
 759                                int* av_pixel_format,
 760                                int* range_guessed)
 761 {
 762     if (range_guessed) *range_guessed = range != AVCOL_RANGE_MPEG &&
 763                                         range != AVCOL_RANGE_JPEG;
 764
 765     //MPEG range is used when no range is set
 766     if (fmt == AV_PIX_FMT_NV12) {
 767         *av_pixel_format = range == AVCOL_RANGE_JPEG ?
 768                                         kCVPixelFormatType_420YpCbCr8BiPlanarFullRange :
 769                                         kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange;
 770     } else if (fmt == AV_PIX_FMT_YUV420P) {
 771         *av_pixel_format = range == AVCOL_RANGE_JPEG ?
 772                                         kCVPixelFormatType_420YpCbCr8PlanarFullRange :
 773                                         kCVPixelFormatType_420YpCbCr8Planar;
 774     } else if (fmt == AV_PIX_FMT_P010LE) {
 775         *av_pixel_format = range == AVCOL_RANGE_JPEG ?
 776                                         kCVPixelFormatType_420YpCbCr10BiPlanarFullRange :
 777                                         kCVPixelFormatType_420YpCbCr10BiPlanarVideoRange;
 778         *av_pixel_format = kCVPixelFormatType_420YpCbCr10BiPlanarVideoRange;
 779     } else {
 780         return AVERROR(EINVAL);
 781     }
 782
 783     return 0;
 784 }
 785
 786 static void add_color_attr(AVCodecContext *avctx, CFMutableDictionaryRef dict) {
 787     VTEncContext *vtctx = avctx->priv_data;
 788
 789     if (vtctx->color_primaries) {
 790         CFDictionarySetValue(dict,
 791                              kCVImageBufferColorPrimariesKey,
 792                              vtctx->color_primaries);
 793     }
 794
 795     if (vtctx->transfer_function) {
 796         CFDictionarySetValue(dict,
 797                              kCVImageBufferTransferFunctionKey,
 798                              vtctx->transfer_function);
 799     }
 800
 801     if (vtctx->ycbcr_matrix) {
 802         CFDictionarySetValue(dict,
 803                              kCVImageBufferYCbCrMatrixKey,
 804                              vtctx->ycbcr_matrix);
 805     }
 806 }
 807
 808 static int create_cv_pixel_buffer_info(AVCodecContext* avctx,
 809                                        CFMutableDictionaryRef* dict)
 810 {
 811     CFNumberRef cv_color_format_num = NULL;
 812     CFNumberRef width_num = NULL;
 813     CFNumberRef height_num = NULL;
 814     CFMutableDictionaryRef pixel_buffer_info = NULL;
 815     int cv_color_format;
 816     int status = get_cv_pixel_format(avctx,
 817                                      avctx->pix_fmt,
 818                                      avctx->color_range,
 819                                      &cv_color_format,
 820                                      NULL);
 821     if (status) return status;
 822
 823     pixel_buffer_info = CFDictionaryCreateMutable(
 824                             kCFAllocatorDefault,
 825                             20,
 826                             &kCFCopyStringDictionaryKeyCallBacks,
 827                             &kCFTypeDictionaryValueCallBacks);
 828
 829     if (!pixel_buffer_info) goto pbinfo_nomem;
 830
 831     cv_color_format_num = CFNumberCreate(kCFAllocatorDefault,
 832                                          kCFNumberSInt32Type,
 833                                          &cv_color_format);
 834     if (!cv_color_format_num) goto pbinfo_nomem;
 835
 836     CFDictionarySetValue(pixel_buffer_info,
 837                          kCVPixelBufferPixelFormatTypeKey,
 838                          cv_color_format_num);
 839     vt_release_num(&cv_color_format_num);
 840
 841     width_num = CFNumberCreate(kCFAllocatorDefault,
 842                                kCFNumberSInt32Type,
 843                                &avctx->width);
 844     if (!width_num) return AVERROR(ENOMEM);
 845
 846     CFDictionarySetValue(pixel_buffer_info,
 847                          kCVPixelBufferWidthKey,
 848                          width_num);
 849     vt_release_num(&width_num);
 850
 851     height_num = CFNumberCreate(kCFAllocatorDefault,
 852                                 kCFNumberSInt32Type,
 853                                 &avctx->height);
 854     if (!height_num) goto pbinfo_nomem;
 855
 856     CFDictionarySetValue(pixel_buffer_info,
 857                          kCVPixelBufferHeightKey,
 858                          height_num);
 859     vt_release_num(&height_num);
 860
 861     add_color_attr(avctx, pixel_buffer_info);
 862
 863     *dict = pixel_buffer_info;
 864     return 0;
 865
 866 pbinfo_nomem:
 867     vt_release_num(&cv_color_format_num);
 868     vt_release_num(&width_num);
 869     vt_release_num(&height_num);
 870     if (pixel_buffer_info) CFRelease(pixel_buffer_info);
 871
 872     return AVERROR(ENOMEM);
 873 }
 874
 875 static int get_cv_color_primaries(AVCodecContext *avctx,
 876                                   CFStringRef *primaries)
 877 {
 878     enum AVColorPrimaries pri = avctx->color_primaries;
 879     switch (pri) {
 880         case AVCOL_PRI_UNSPECIFIED:
 881             *primaries = NULL;
 882             break;
 883
 884         case AVCOL_PRI_BT709:
 885             *primaries = kCVImageBufferColorPrimaries_ITU_R_709_2;
 886             break;
 887
 888         case AVCOL_PRI_BT2020:
 889             *primaries = compat_keys.kCVImageBufferColorPrimaries_ITU_R_2020;
 890             break;
 891
 892         default:
 893             av_log(avctx, AV_LOG_ERROR, "Color primaries %s is not supported.\n", av_color_primaries_name(pri));
 894             *primaries = NULL;
 895             return -1;
 896     }
 897
 898     return 0;
 899 }
 900
 901 static int get_cv_transfer_function(AVCodecContext *avctx,
 902                                     CFStringRef *transfer_fnc,
 903                                     CFNumberRef *gamma_level)
 904 {
 905     enum AVColorTransferCharacteristic trc = avctx->color_trc;
 906     Float32 gamma;
 907     *gamma_level = NULL;
 908
 909     switch (trc) {
 910         case AVCOL_TRC_UNSPECIFIED:
 911             *transfer_fnc = NULL;
 912             break;
 913
 914         case AVCOL_TRC_BT709:
 915             *transfer_fnc = kCVImageBufferTransferFunction_ITU_R_709_2;
 916             break;
 917
 918         case AVCOL_TRC_SMPTE240M:
 919             *transfer_fnc = kCVImageBufferTransferFunction_SMPTE_240M_1995;
 920             break;
 921
 922 #if HAVE_KCVIMAGEBUFFERTRANSFERFUNCTION_SMPTE_ST_2084_PQ
 923         case AVCOL_TRC_SMPTE2084:
 924             *transfer_fnc = kCVImageBufferTransferFunction_SMPTE_ST_2084_PQ;
 925             break;
 926 #endif
 927 #if HAVE_KCVIMAGEBUFFERTRANSFERFUNCTION_LINEAR
 928         case AVCOL_TRC_LINEAR:
 929             *transfer_fnc = kCVImageBufferTransferFunction_Linear;
 930             break;
 931 #endif
 932 #if HAVE_KCVIMAGEBUFFERTRANSFERFUNCTION_ITU_R_2100_HLG
 933         case AVCOL_TRC_ARIB_STD_B67:
 934             *transfer_fnc = kCVImageBufferTransferFunction_ITU_R_2100_HLG;
 935             break;
 936 #endif
 937
 938         case AVCOL_TRC_GAMMA22:
 939             gamma = 2.2;
 940             *transfer_fnc = kCVImageBufferTransferFunction_UseGamma;
 941             *gamma_level = CFNumberCreate(NULL, kCFNumberFloat32Type, &gamma);
 942             break;
 943
 944         case AVCOL_TRC_GAMMA28:
 945             gamma = 2.8;
 946             *transfer_fnc = kCVImageBufferTransferFunction_UseGamma;
 947             *gamma_level = CFNumberCreate(NULL, kCFNumberFloat32Type, &gamma);
 948             break;
 949
 950         case AVCOL_TRC_BT2020_10:
 951         case AVCOL_TRC_BT2020_12:
 952             *transfer_fnc = compat_keys.kCVImageBufferTransferFunction_ITU_R_2020;
 953             break;
 954
 955         default:
 956             *transfer_fnc = NULL;
 957             av_log(avctx, AV_LOG_ERROR, "Transfer function %s is not supported.\n", av_color_transfer_name(trc));
 958             return -1;
 959     }
 960
 961     return 0;
 962 }
 963
 964 static int get_cv_ycbcr_matrix(AVCodecContext *avctx, CFStringRef *matrix) {
 965     switch(avctx->colorspace) {
 966         case AVCOL_SPC_BT709:
 967             *matrix = kCVImageBufferYCbCrMatrix_ITU_R_709_2;
 968             break;
 969
 970         case AVCOL_SPC_UNSPECIFIED:
 971             *matrix = NULL;
 972             break;
 973
 974         case AVCOL_SPC_BT470BG:
 975         case AVCOL_SPC_SMPTE170M:
 976             *matrix = kCVImageBufferYCbCrMatrix_ITU_R_601_4;
 977             break;
 978
 979         case AVCOL_SPC_SMPTE240M:
 980             *matrix = kCVImageBufferYCbCrMatrix_SMPTE_240M_1995;
 981             break;
 982
 983         case AVCOL_SPC_BT2020_NCL:
 984             *matrix = compat_keys.kCVImageBufferYCbCrMatrix_ITU_R_2020;
 985             break;
 986
 987         default:
 988             av_log(avctx, AV_LOG_ERROR, "Color space %s is not supported.\n", av_color_space_name(avctx->colorspace));
 989             return -1;
 990     }
 991
 992     return 0;
 993 }
 994
 995 static int vtenc_create_encoder(AVCodecContext   *avctx,
 996                                 CMVideoCodecType codec_type,
 997                                 CFStringRef      profile_level,
 998                                 CFNumberRef      gamma_level,
 999                                 CFDictionaryRef  enc_info,
1000                                 CFDictionaryRef  pixel_buffer_info,
1001                                 VTCompressionSessionRef *session)
1002 {
1003     VTEncContext *vtctx = avctx->priv_data;
1004     SInt32       bit_rate = avctx->bit_rate;
1005     SInt32       max_rate = avctx->rc_max_rate;
1006     CFNumberRef  bit_rate_num;
1007     CFNumberRef  bytes_per_second;
1008     CFNumberRef  one_second;
1009     CFArrayRef   data_rate_limits;
1010     int64_t      bytes_per_second_value = 0;
1011     int64_t      one_second_value = 0;
1012     void         *nums[2];
1013
1014     int status = VTCompressionSessionCreate(kCFAllocatorDefault,
1015                                             avctx->width,
1016                                             avctx->height,
1017                                             codec_type,
1018                                             enc_info,
1019                                             pixel_buffer_info,
1020                                             kCFAllocatorDefault,
1021                                             vtenc_output_callback,
1022                                             avctx,
1023                                             session);
1024
1025     if (status || !vtctx->session) {
1026         av_log(avctx, AV_LOG_ERROR, "Error: cannot create compression session: %d\n", status);
1027
1028 #if !TARGET_OS_IPHONE
1029         if (!vtctx->allow_sw) {
1030             av_log(avctx, AV_LOG_ERROR, "Try -allow_sw 1. The hardware encoder may be busy, or not supported.\n");
1031         }
1032 #endif
1033
1034         return AVERROR_EXTERNAL;
1035     }
1036
1037     bit_rate_num = CFNumberCreate(kCFAllocatorDefault,
1038                                   kCFNumberSInt32Type,
1039                                   &bit_rate);
1040     if (!bit_rate_num) return AVERROR(ENOMEM);
1041
1042     status = VTSessionSetProperty(vtctx->session,
1043                                   kVTCompressionPropertyKey_AverageBitRate,
1044                                   bit_rate_num);
1045     CFRelease(bit_rate_num);
1046
1047     if (status) {
1048         av_log(avctx, AV_LOG_ERROR, "Error setting bitrate property: %d\n", status);
1049         return AVERROR_EXTERNAL;
1050     }
1051
1052     if (vtctx->codec_id == AV_CODEC_ID_H264 && max_rate > 0) {
1053         // kVTCompressionPropertyKey_DataRateLimits is not available for HEVC
1054         bytes_per_second_value = max_rate >> 3;
1055         bytes_per_second = CFNumberCreate(kCFAllocatorDefault,
1056                                           kCFNumberSInt64Type,
1057                                           &bytes_per_second_value);
1058         if (!bytes_per_second) {
1059             return AVERROR(ENOMEM);
1060         }
1061         one_second_value = 1;
1062         one_second = CFNumberCreate(kCFAllocatorDefault,
1063                                     kCFNumberSInt64Type,
1064                                     &one_second_value);
1065         if (!one_second) {
1066             CFRelease(bytes_per_second);
1067             return AVERROR(ENOMEM);
1068         }
1069         nums[0] = (void *)bytes_per_second;
1070         nums[1] = (void *)one_second;
1071         data_rate_limits = CFArrayCreate(kCFAllocatorDefault,
1072                                          (const void **)nums,
1073                                          2,
1074                                          &kCFTypeArrayCallBacks);
1075
1076         if (!data_rate_limits) {
1077             CFRelease(bytes_per_second);
1078             CFRelease(one_second);
1079             return AVERROR(ENOMEM);
1080         }
1081         status = VTSessionSetProperty(vtctx->session,
1082                                       kVTCompressionPropertyKey_DataRateLimits,
1083                                       data_rate_limits);
1084
1085         CFRelease(bytes_per_second);
1086         CFRelease(one_second);
1087         CFRelease(data_rate_limits);
1088
1089         if (status) {
1090             av_log(avctx, AV_LOG_ERROR, "Error setting max bitrate property: %d\n", status);
1091             return AVERROR_EXTERNAL;
1092         }
1093     }
1094
1095     if (vtctx->codec_id == AV_CODEC_ID_H264) {
1096         // kVTCompressionPropertyKey_ProfileLevel is not available for HEVC
1097         if (profile_level) {
1098             status = VTSessionSetProperty(vtctx->session,
1099                                         kVTCompressionPropertyKey_ProfileLevel,
1100                                         profile_level);
1101             if (status) {
1102                 av_log(avctx, AV_LOG_ERROR, "Error setting profile/level property: %d\n", status);
1103             }
1104         }
1105     }
1106
1107     if (avctx->gop_size > 0) {
1108         CFNumberRef interval = CFNumberCreate(kCFAllocatorDefault,
1109                                               kCFNumberIntType,
1110                                               &avctx->gop_size);
1111         if (!interval) {
1112             return AVERROR(ENOMEM);
1113         }
1114
1115         status = VTSessionSetProperty(vtctx->session,
1116                                       kVTCompressionPropertyKey_MaxKeyFrameInterval,
1117                                       interval);
1118         CFRelease(interval);
1119
1120         if (status) {
1121             av_log(avctx, AV_LOG_ERROR, "Error setting 'max key-frame interval' property: %d\n", status);
1122             return AVERROR_EXTERNAL;
1123         }
1124     }
1125
1126     if (vtctx->frames_before) {
1127         status = VTSessionSetProperty(vtctx->session,
1128                                       kVTCompressionPropertyKey_MoreFramesBeforeStart,
1129                                       kCFBooleanTrue);
1130
1131         if (status == kVTPropertyNotSupportedErr) {
1132             av_log(avctx, AV_LOG_WARNING, "frames_before property is not supported on this device. Ignoring.\n");
1133         } else if (status) {
1134             av_log(avctx, AV_LOG_ERROR, "Error setting frames_before property: %d\n", status);
1135         }
1136     }
1137
1138     if (vtctx->frames_after) {
1139         status = VTSessionSetProperty(vtctx->session,
1140                                       kVTCompressionPropertyKey_MoreFramesAfterEnd,
1141                                       kCFBooleanTrue);
1142
1143         if (status == kVTPropertyNotSupportedErr) {
1144             av_log(avctx, AV_LOG_WARNING, "frames_after property is not supported on this device. Ignoring.\n");
1145         } else if (status) {
1146             av_log(avctx, AV_LOG_ERROR, "Error setting frames_after property: %d\n", status);
1147         }
1148     }
1149
1150     if (avctx->sample_aspect_ratio.num != 0) {
1151         CFNumberRef num;
1152         CFNumberRef den;
1153         CFMutableDictionaryRef par;
1154         AVRational *avpar = &avctx->sample_aspect_ratio;
1155
1156         av_reduce(&avpar->num, &avpar->den,
1157                    avpar->num,  avpar->den,
1158                   0xFFFFFFFF);
1159
1160         num = CFNumberCreate(kCFAllocatorDefault,
1161                              kCFNumberIntType,
1162                              &avpar->num);
1163
1164         den = CFNumberCreate(kCFAllocatorDefault,
1165                              kCFNumberIntType,
1166                              &avpar->den);
1167
1168
1169
1170         par = CFDictionaryCreateMutable(kCFAllocatorDefault,
1171                                         2,
1172                                         &kCFCopyStringDictionaryKeyCallBacks,
1173                                         &kCFTypeDictionaryValueCallBacks);
1174
1175         if (!par || !num || !den) {
1176             if (par) CFRelease(par);
1177             if (num) CFRelease(num);
1178             if (den) CFRelease(den);
1179
1180             return AVERROR(ENOMEM);
1181         }
1182
1183         CFDictionarySetValue(
1184             par,
1185             kCMFormatDescriptionKey_PixelAspectRatioHorizontalSpacing,
1186             num);
1187
1188         CFDictionarySetValue(
1189             par,
1190             kCMFormatDescriptionKey_PixelAspectRatioVerticalSpacing,
1191             den);
1192
1193         status = VTSessionSetProperty(vtctx->session,
1194                                       kVTCompressionPropertyKey_PixelAspectRatio,
1195                                       par);
1196
1197         CFRelease(par);
1198         CFRelease(num);
1199         CFRelease(den);
1200
1201         if (status) {
1202             av_log(avctx,
1203                    AV_LOG_ERROR,
1204                    "Error setting pixel aspect ratio to %d:%d: %d.\n",
1205                    avctx->sample_aspect_ratio.num,
1206                    avctx->sample_aspect_ratio.den,
1207                    status);
1208
1209             return AVERROR_EXTERNAL;
1210         }
1211     }
1212
1213
1214     if (vtctx->transfer_function) {
1215         status = VTSessionSetProperty(vtctx->session,
1216                                       kVTCompressionPropertyKey_TransferFunction,
1217                                       vtctx->transfer_function);
1218
1219         if (status) {
1220             av_log(avctx, AV_LOG_WARNING, "Could not set transfer function: %d\n", status);
1221         }
1222     }
1223
1224
1225     if (vtctx->ycbcr_matrix) {
1226         status = VTSessionSetProperty(vtctx->session,
1227                                       kVTCompressionPropertyKey_YCbCrMatrix,
1228                                       vtctx->ycbcr_matrix);
1229
1230         if (status) {
1231             av_log(avctx, AV_LOG_WARNING, "Could not set ycbcr matrix: %d\n", status);
1232         }
1233     }
1234
1235
1236     if (vtctx->color_primaries) {
1237         status = VTSessionSetProperty(vtctx->session,
1238                                       kVTCompressionPropertyKey_ColorPrimaries,
1239                                       vtctx->color_primaries);
1240
1241         if (status) {
1242             av_log(avctx, AV_LOG_WARNING, "Could not set color primaries: %d\n", status);
1243         }
1244     }
1245
1246     if (gamma_level) {
1247         status = VTSessionSetProperty(vtctx->session,
1248                                       kCVImageBufferGammaLevelKey,
1249                                       gamma_level);
1250
1251         if (status) {
1252             av_log(avctx, AV_LOG_WARNING, "Could not set gamma level: %d\n", status);
1253         }
1254     }
1255
1256     if (!vtctx->has_b_frames) {
1257         status = VTSessionSetProperty(vtctx->session,
1258                                       kVTCompressionPropertyKey_AllowFrameReordering,
1259                                       kCFBooleanFalse);
1260
1261         if (status) {
1262             av_log(avctx, AV_LOG_ERROR, "Error setting 'allow frame reordering' property: %d\n", status);
1263             return AVERROR_EXTERNAL;
1264         }
1265     }
1266
1267     if (vtctx->entropy != VT_ENTROPY_NOT_SET) {
1268         CFStringRef entropy = vtctx->entropy == VT_CABAC ?
1269                                 compat_keys.kVTH264EntropyMode_CABAC:
1270                                 compat_keys.kVTH264EntropyMode_CAVLC;
1271
1272         status = VTSessionSetProperty(vtctx->session,
1273                                       compat_keys.kVTCompressionPropertyKey_H264EntropyMode,
1274                                       entropy);
1275
1276         if (status) {
1277             av_log(avctx, AV_LOG_ERROR, "Error setting entropy property: %d\n", status);
1278         }
1279     }
1280
1281     if (vtctx->realtime) {
1282         status = VTSessionSetProperty(vtctx->session,
1283                                       compat_keys.kVTCompressionPropertyKey_RealTime,
1284                                       kCFBooleanTrue);
1285
1286         if (status) {
1287             av_log(avctx, AV_LOG_ERROR, "Error setting realtime property: %d\n", status);
1288         }
1289     }
1290
1291     status = VTCompressionSessionPrepareToEncodeFrames(vtctx->session);
1292     if (status) {
1293         av_log(avctx, AV_LOG_ERROR, "Error: cannot prepare encoder: %d\n", status);
1294         return AVERROR_EXTERNAL;
1295     }
1296
1297     return 0;
1298 }
1299
1300 static int vtenc_configure_encoder(AVCodecContext *avctx)
1301 {
1302     CFMutableDictionaryRef enc_info;
1303     CFMutableDictionaryRef pixel_buffer_info;
1304     CMVideoCodecType       codec_type;
1305     VTEncContext           *vtctx = avctx->priv_data;
1306     CFStringRef            profile_level;
1307     CFNumberRef            gamma_level = NULL;
1308     int                    status;
1309
1310     codec_type = get_cm_codec_type(avctx->codec_id);
1311     if (!codec_type) {
1312         av_log(avctx, AV_LOG_ERROR, "Error: no mapping for AVCodecID %d\n", avctx->codec_id);
1313         return AVERROR(EINVAL);
1314     }
1315
1316     vtctx->codec_id = avctx->codec_id;
1317
1318     if (vtctx->codec_id == AV_CODEC_ID_H264) {
1319         vtctx->get_param_set_func = CMVideoFormatDescriptionGetH264ParameterSetAtIndex;
1320
1321         vtctx->has_b_frames = avctx->max_b_frames > 0;
1322         if(vtctx->has_b_frames && vtctx->profile == H264_PROF_BASELINE){
1323             av_log(avctx, AV_LOG_WARNING, "Cannot use B-frames with baseline profile. Output will not contain B-frames.\n");
1324             vtctx->has_b_frames = false;
1325         }
1326
1327         if (vtctx->entropy == VT_CABAC && vtctx->profile == H264_PROF_BASELINE) {
1328             av_log(avctx, AV_LOG_WARNING, "CABAC entropy requires 'main' or 'high' profile, but baseline was requested. Encode will not use CABAC entropy.\n");
1329             vtctx->entropy = VT_ENTROPY_NOT_SET;
1330         }
1331
1332         if (!get_vt_h264_profile_level(avctx, &profile_level)) return AVERROR(EINVAL);
1333     } else {
1334         vtctx->get_param_set_func = compat_keys.CMVideoFormatDescriptionGetHEVCParameterSetAtIndex;
1335         if (!vtctx->get_param_set_func) return AVERROR(EINVAL);
1336         if (!get_vt_hevc_profile_level(avctx, &profile_level)) return AVERROR(EINVAL);
1337     }
1338
1339     enc_info = CFDictionaryCreateMutable(
1340         kCFAllocatorDefault,
1341         20,
1342         &kCFCopyStringDictionaryKeyCallBacks,
1343         &kCFTypeDictionaryValueCallBacks
1344     );
1345
1346     if (!enc_info) return AVERROR(ENOMEM);
1347
1348 #if !TARGET_OS_IPHONE
1349     if(vtctx->require_sw) {
1350         CFDictionarySetValue(enc_info,
1351                              compat_keys.kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder,
1352                              kCFBooleanFalse);
1353     } else if (!vtctx->allow_sw) {
1354         CFDictionarySetValue(enc_info,
1355                              compat_keys.kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder,
1356                              kCFBooleanTrue);
1357     } else {
1358         CFDictionarySetValue(enc_info,
1359                              compat_keys.kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder,
1360                              kCFBooleanTrue);
1361     }
1362 #endif
1363
1364     if (avctx->pix_fmt != AV_PIX_FMT_VIDEOTOOLBOX) {
1365         status = create_cv_pixel_buffer_info(avctx, &pixel_buffer_info);
1366         if (status)
1367             goto init_cleanup;
1368     } else {
1369         pixel_buffer_info = NULL;
1370     }
1371
1372     vtctx->dts_delta = vtctx->has_b_frames ? -1 : 0;
1373
1374     get_cv_transfer_function(avctx, &vtctx->transfer_function, &gamma_level);
1375     get_cv_ycbcr_matrix(avctx, &vtctx->ycbcr_matrix);
1376     get_cv_color_primaries(avctx, &vtctx->color_primaries);
1377
1378
1379     if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
1380         status = vtenc_populate_extradata(avctx,
1381                                           codec_type,
1382                                           profile_level,
1383                                           gamma_level,
1384                                           enc_info,
1385                                           pixel_buffer_info);
1386         if (status)
1387             goto init_cleanup;
1388     }
1389
1390     status = vtenc_create_encoder(avctx,
1391                                   codec_type,
1392                                   profile_level,
1393                                   gamma_level,
1394                                   enc_info,
1395                                   pixel_buffer_info,
1396                                   &vtctx->session);
1397
1398 init_cleanup:
1399     if (gamma_level)
1400         CFRelease(gamma_level);
1401
1402     if (pixel_buffer_info)
1403         CFRelease(pixel_buffer_info);
1404
1405     CFRelease(enc_info);
1406
1407     return status;
1408 }
1409
1410 static av_cold int vtenc_init(AVCodecContext *avctx)
1411 {
1412     VTEncContext    *vtctx = avctx->priv_data;
1413     CFBooleanRef    has_b_frames_cfbool;
1414     int             status;
1415
1416     pthread_once(&once_ctrl, loadVTEncSymbols);
1417
1418     pthread_mutex_init(&vtctx->lock, NULL);
1419     pthread_cond_init(&vtctx->cv_sample_sent, NULL);
1420
1421     vtctx->session = NULL;
1422     status = vtenc_configure_encoder(avctx);
1423     if (status) return status;
1424
1425     status = VTSessionCopyProperty(vtctx->session,
1426                                    kVTCompressionPropertyKey_AllowFrameReordering,
1427                                    kCFAllocatorDefault,
1428                                    &has_b_frames_cfbool);
1429
1430     if (!status && has_b_frames_cfbool) {
1431         //Some devices don't output B-frames for main profile, even if requested.
1432         vtctx->has_b_frames = CFBooleanGetValue(has_b_frames_cfbool);
1433         CFRelease(has_b_frames_cfbool);
1434     }
1435     avctx->has_b_frames = vtctx->has_b_frames;
1436
1437     return 0;
1438 }
1439
1440 static void vtenc_get_frame_info(CMSampleBufferRef buffer, bool *is_key_frame)
1441 {
1442     CFArrayRef      attachments;
1443     CFDictionaryRef attachment;
1444     CFBooleanRef    not_sync;
1445     CFIndex         len;
1446
1447     attachments = CMSampleBufferGetSampleAttachmentsArray(buffer, false);
1448     len = !attachments ? 0 : CFArrayGetCount(attachments);
1449
1450     if (!len) {
1451         *is_key_frame = true;
1452         return;
1453     }
1454
1455     attachment = CFArrayGetValueAtIndex(attachments, 0);
1456
1457     if (CFDictionaryGetValueIfPresent(attachment,
1458                                       kCMSampleAttachmentKey_NotSync,
1459                                       (const void **)&not_sync))
1460     {
1461         *is_key_frame = !CFBooleanGetValue(not_sync);
1462     } else {
1463         *is_key_frame = true;
1464     }
1465 }
1466
1467 static int is_post_sei_nal_type(int nal_type){
1468     return nal_type != H264_NAL_SEI &&
1469            nal_type != H264_NAL_SPS &&
1470            nal_type != H264_NAL_PPS &&
1471            nal_type != H264_NAL_AUD;
1472 }
1473
1474 /*
1475  * Finds the sei message start/size of type find_sei_type.
1476  * If more than one of that type exists, the last one is returned.
1477  */
1478 static int find_sei_end(AVCodecContext *avctx,
1479                         uint8_t        *nal_data,
1480                         size_t          nal_size,
1481                         uint8_t       **sei_end)
1482 {
1483     int nal_type;
1484     size_t sei_payload_size = 0;
1485     int sei_payload_type = 0;
1486     *sei_end = NULL;
1487     uint8_t *nal_start = nal_data;
1488
1489     if (!nal_size)
1490         return 0;
1491
1492     nal_type = *nal_data & 0x1F;
1493     if (nal_type != H264_NAL_SEI)
1494         return 0;
1495
1496     nal_data++;
1497     nal_size--;
1498
1499     if (nal_data[nal_size - 1] == 0x80)
1500         nal_size--;
1501
1502     while (nal_size > 0 && *nal_data > 0) {
1503         do{
1504             sei_payload_type += *nal_data;
1505             nal_data++;
1506             nal_size--;
1507         } while (nal_size > 0 && *nal_data == 0xFF);
1508
1509         if (!nal_size) {
1510             av_log(avctx, AV_LOG_ERROR, "Unexpected end of SEI NAL Unit parsing type.\n");
1511             return AVERROR_INVALIDDATA;
1512         }
1513
1514         do{
1515             sei_payload_size += *nal_data;
1516             nal_data++;
1517             nal_size--;
1518         } while (nal_size > 0 && *nal_data == 0xFF);
1519
1520         if (nal_size < sei_payload_size) {
1521             av_log(avctx, AV_LOG_ERROR, "Unexpected end of SEI NAL Unit parsing size.\n");
1522             return AVERROR_INVALIDDATA;
1523         }
1524
1525         nal_data += sei_payload_size;
1526         nal_size -= sei_payload_size;
1527     }
1528
1529     *sei_end = nal_data;
1530
1531     return nal_data - nal_start + 1;
1532 }
1533
1534 /**
1535  * Copies the data inserting emulation prevention bytes as needed.
1536  * Existing data in the destination can be taken into account by providing
1537  * dst with a dst_offset > 0.
1538  *
1539  * @return The number of bytes copied on success. On failure, the negative of
1540  *         the number of bytes needed to copy src is returned.
1541  */
1542 static int copy_emulation_prev(const uint8_t *src,
1543                                size_t         src_size,
1544                                uint8_t       *dst,
1545                                ssize_t        dst_offset,
1546                                size_t         dst_size)
1547 {
1548     int zeros = 0;
1549     int wrote_bytes;
1550     uint8_t* dst_start;
1551     uint8_t* dst_end = dst + dst_size;
1552     const uint8_t* src_end = src + src_size;
1553     int start_at = dst_offset > 2 ? dst_offset - 2 : 0;
1554     int i;
1555     for (i = start_at; i < dst_offset && i < dst_size; i++) {
1556         if (!dst[i])
1557             zeros++;
1558         else
1559             zeros = 0;
1560     }
1561
1562     dst += dst_offset;
1563     dst_start = dst;
1564     for (; src < src_end; src++, dst++) {
1565         if (zeros == 2) {
1566             int insert_ep3_byte = *src <= 3;
1567             if (insert_ep3_byte) {
1568                 if (dst < dst_end)
1569                     *dst = 3;
1570                 dst++;
1571             }
1572
1573             zeros = 0;
1574         }
1575
1576         if (dst < dst_end)
1577             *dst = *src;
1578
1579         if (!*src)
1580             zeros++;
1581         else
1582             zeros = 0;
1583     }
1584
1585     wrote_bytes = dst - dst_start;
1586
1587     if (dst > dst_end)
1588         return -wrote_bytes;
1589
1590     return wrote_bytes;
1591 }
1592
1593 static int write_sei(const ExtraSEI *sei,
1594                      int             sei_type,
1595                      uint8_t        *dst,
1596                      size_t          dst_size)
1597 {
1598     uint8_t *sei_start = dst;
1599     size_t remaining_sei_size = sei->size;
1600     size_t remaining_dst_size = dst_size;
1601     int header_bytes;
1602     int bytes_written;
1603     ssize_t offset;
1604
1605     if (!remaining_dst_size)
1606         return AVERROR_BUFFER_TOO_SMALL;
1607
1608     while (sei_type && remaining_dst_size != 0) {
1609         int sei_byte = sei_type > 255 ? 255 : sei_type;
1610         *dst = sei_byte;
1611
1612         sei_type -= sei_byte;
1613         dst++;
1614         remaining_dst_size--;
1615     }
1616
1617     if (!dst_size)
1618         return AVERROR_BUFFER_TOO_SMALL;
1619
1620     while (remaining_sei_size && remaining_dst_size != 0) {
1621         int size_byte = remaining_sei_size > 255 ? 255 : remaining_sei_size;
1622         *dst = size_byte;
1623
1624         remaining_sei_size -= size_byte;
1625         dst++;
1626         remaining_dst_size--;
1627     }
1628
1629     if (remaining_dst_size < sei->size)
1630         return AVERROR_BUFFER_TOO_SMALL;
1631
1632     header_bytes = dst - sei_start;
1633
1634     offset = header_bytes;
1635     bytes_written = copy_emulation_prev(sei->data,
1636                                         sei->size,
1637                                         sei_start,
1638                                         offset,
1639                                         dst_size);
1640     if (bytes_written < 0)
1641         return AVERROR_BUFFER_TOO_SMALL;
1642
1643     bytes_written += header_bytes;
1644     return bytes_written;
1645 }
1646
1647 /**
1648  * Copies NAL units and replaces length codes with
1649  * H.264 Annex B start codes. On failure, the contents of
1650  * dst_data may have been modified.
1651  *
1652  * @param length_code_size Byte length of each length code
1653  * @param sample_buffer NAL units prefixed with length codes.
1654  * @param sei Optional A53 closed captions SEI data.
1655  * @param dst_data Must be zeroed before calling this function.
1656  *                 Contains the copied NAL units prefixed with
1657  *                 start codes when the function returns
1658  *                 successfully.
1659  * @param dst_size Length of dst_data
1660  * @return 0 on success
1661  *         AVERROR_INVALIDDATA if length_code_size is invalid
1662  *         AVERROR_BUFFER_TOO_SMALL if dst_data is too small
1663  *         or if a length_code in src_data specifies data beyond
1664  *         the end of its buffer.
1665  */
1666 static int copy_replace_length_codes(
1667     AVCodecContext *avctx,
1668     size_t        length_code_size,
1669     CMSampleBufferRef sample_buffer,
1670     ExtraSEI      *sei,
1671     uint8_t       *dst_data,
1672     size_t        dst_size)
1673 {
1674     size_t src_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
1675     size_t remaining_src_size = src_size;
1676     size_t remaining_dst_size = dst_size;
1677     size_t src_offset = 0;
1678     int wrote_sei = 0;
1679     int status;
1680     uint8_t size_buf[4];
1681     uint8_t nal_type;
1682     CMBlockBufferRef block = CMSampleBufferGetDataBuffer(sample_buffer);
1683
1684     if (length_code_size > 4) {
1685         return AVERROR_INVALIDDATA;
1686     }
1687
1688     while (remaining_src_size > 0) {
1689         size_t curr_src_len;
1690         size_t curr_dst_len;
1691         size_t box_len = 0;
1692         size_t i;
1693
1694         uint8_t       *dst_box;
1695
1696         status = CMBlockBufferCopyDataBytes(block,
1697                                             src_offset,
1698                                             length_code_size,
1699                                             size_buf);
1700         if (status) {
1701             av_log(avctx, AV_LOG_ERROR, "Cannot copy length: %d\n", status);
1702             return AVERROR_EXTERNAL;
1703         }
1704
1705         status = CMBlockBufferCopyDataBytes(block,
1706                                             src_offset + length_code_size,
1707                                             1,
1708                                             &nal_type);
1709
1710         if (status) {
1711             av_log(avctx, AV_LOG_ERROR, "Cannot copy type: %d\n", status);
1712             return AVERROR_EXTERNAL;
1713         }
1714
1715         nal_type &= 0x1F;
1716
1717         for (i = 0; i < length_code_size; i++) {
1718             box_len <<= 8;
1719             box_len |= size_buf[i];
1720         }
1721
1722         if (sei && !wrote_sei && is_post_sei_nal_type(nal_type)) {
1723             //No SEI NAL unit - insert.
1724             int wrote_bytes;
1725
1726             memcpy(dst_data, start_code, sizeof(start_code));
1727             dst_data += sizeof(start_code);
1728             remaining_dst_size -= sizeof(start_code);
1729
1730             *dst_data = H264_NAL_SEI;
1731             dst_data++;
1732             remaining_dst_size--;
1733
1734             wrote_bytes = write_sei(sei,
1735                                     H264_SEI_TYPE_USER_DATA_REGISTERED,
1736                                     dst_data,
1737                                     remaining_dst_size);
1738
1739             if (wrote_bytes < 0)
1740                 return wrote_bytes;
1741
1742             remaining_dst_size -= wrote_bytes;
1743             dst_data += wrote_bytes;
1744
1745             if (remaining_dst_size <= 0)
1746                 return AVERROR_BUFFER_TOO_SMALL;
1747
1748             *dst_data = 0x80;
1749
1750             dst_data++;
1751             remaining_dst_size--;
1752
1753             wrote_sei = 1;
1754         }
1755
1756         curr_src_len = box_len + length_code_size;
1757         curr_dst_len = box_len + sizeof(start_code);
1758
1759         if (remaining_src_size < curr_src_len) {
1760             return AVERROR_BUFFER_TOO_SMALL;
1761         }
1762
1763         if (remaining_dst_size < curr_dst_len) {
1764             return AVERROR_BUFFER_TOO_SMALL;
1765         }
1766
1767         dst_box = dst_data + sizeof(start_code);
1768
1769         memcpy(dst_data, start_code, sizeof(start_code));
1770         status = CMBlockBufferCopyDataBytes(block,
1771                                             src_offset + length_code_size,
1772                                             box_len,
1773                                             dst_box);
1774
1775         if (status) {
1776             av_log(avctx, AV_LOG_ERROR, "Cannot copy data: %d\n", status);
1777             return AVERROR_EXTERNAL;
1778         }
1779
1780         if (sei && !wrote_sei && nal_type == H264_NAL_SEI) {
1781             //Found SEI NAL unit - append.
1782             int wrote_bytes;
1783             int old_sei_length;
1784             int extra_bytes;
1785             uint8_t *new_sei;
1786             old_sei_length = find_sei_end(avctx, dst_box, box_len, &new_sei);
1787             if (old_sei_length < 0)
1788                 return status;
1789
1790             wrote_bytes = write_sei(sei,
1791                                     H264_SEI_TYPE_USER_DATA_REGISTERED,
1792                                     new_sei,
1793                                     remaining_dst_size - old_sei_length);
1794             if (wrote_bytes < 0)
1795                 return wrote_bytes;
1796
1797             if (new_sei + wrote_bytes >= dst_data + remaining_dst_size)
1798                 return AVERROR_BUFFER_TOO_SMALL;
1799
1800             new_sei[wrote_bytes++] = 0x80;
1801             extra_bytes = wrote_bytes - (dst_box + box_len - new_sei);
1802
1803             dst_data += extra_bytes;
1804             remaining_dst_size -= extra_bytes;
1805
1806             wrote_sei = 1;
1807         }
1808
1809         src_offset += curr_src_len;
1810         dst_data += curr_dst_len;
1811
1812         remaining_src_size -= curr_src_len;
1813         remaining_dst_size -= curr_dst_len;
1814     }
1815
1816     return 0;
1817 }
1818
1819 /**
1820  * Returns a sufficient number of bytes to contain the sei data.
1821  * It may be greater than the minimum required.
1822  */
1823 static int get_sei_msg_bytes(const ExtraSEI* sei, int type){
1824     int copied_size;
1825     if (sei->size == 0)
1826         return 0;
1827
1828     copied_size = -copy_emulation_prev(sei->data,
1829                                        sei->size,
1830                                        NULL,
1831                                        0,
1832                                        0);
1833
1834     if ((sei->size % 255) == 0) //may result in an extra byte
1835         copied_size++;
1836
1837     return copied_size + sei->size / 255 + 1 + type / 255 + 1;
1838 }
1839
1840 static int vtenc_cm_to_avpacket(
1841     AVCodecContext    *avctx,
1842     CMSampleBufferRef sample_buffer,
1843     AVPacket          *pkt,
1844     ExtraSEI          *sei)
1845 {
1846     VTEncContext *vtctx = avctx->priv_data;
1847
1848     int     status;
1849     bool    is_key_frame;
1850     bool    add_header;
1851     size_t  length_code_size;
1852     size_t  header_size = 0;
1853     size_t  in_buf_size;
1854     size_t  out_buf_size;
1855     size_t  sei_nalu_size = 0;
1856     int64_t dts_delta;
1857     int64_t time_base_num;
1858     int nalu_count;
1859     CMTime  pts;
1860     CMTime  dts;
1861     CMVideoFormatDescriptionRef vid_fmt;
1862
1863
1864     vtenc_get_frame_info(sample_buffer, &is_key_frame);
1865     status = get_length_code_size(avctx, sample_buffer, &length_code_size);
1866     if (status) return status;
1867
1868     add_header = is_key_frame && !(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER);
1869
1870     if (add_header) {
1871         vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
1872         if (!vid_fmt) {
1873             av_log(avctx, AV_LOG_ERROR, "Cannot get format description.\n");
1874             return AVERROR_EXTERNAL;
1875         }
1876
1877         int status = get_params_size(avctx, vid_fmt, &header_size);
1878         if (status) return status;
1879     }
1880
1881     status = count_nalus(length_code_size, sample_buffer, &nalu_count);
1882     if(status)
1883         return status;
1884
1885     if (sei) {
1886         size_t msg_size = get_sei_msg_bytes(sei,
1887                                             H264_SEI_TYPE_USER_DATA_REGISTERED);
1888
1889         sei_nalu_size = sizeof(start_code) + 1 + msg_size + 1;
1890     }
1891
1892     in_buf_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
1893     out_buf_size = header_size +
1894                    in_buf_size +
1895                    sei_nalu_size +
1896                    nalu_count * ((int)sizeof(start_code) - (int)length_code_size);
1897
1898     status = ff_alloc_packet2(avctx, pkt, out_buf_size, out_buf_size);
1899     if (status < 0)
1900         return status;
1901
1902     if (add_header) {
1903         status = copy_param_sets(avctx, vid_fmt, pkt->data, out_buf_size);
1904         if(status) return status;
1905     }
1906
1907     status = copy_replace_length_codes(
1908         avctx,
1909         length_code_size,
1910         sample_buffer,
1911         sei,
1912         pkt->data + header_size,
1913         pkt->size - header_size
1914     );
1915
1916     if (status) {
1917         av_log(avctx, AV_LOG_ERROR, "Error copying packet data: %d\n", status);
1918         return status;
1919     }
1920
1921     if (is_key_frame) {
1922         pkt->flags |= AV_PKT_FLAG_KEY;
1923     }
1924
1925     pts = CMSampleBufferGetPresentationTimeStamp(sample_buffer);
1926     dts = CMSampleBufferGetDecodeTimeStamp      (sample_buffer);
1927
1928     if (CMTIME_IS_INVALID(dts)) {
1929         if (!vtctx->has_b_frames) {
1930             dts = pts;
1931         } else {
1932             av_log(avctx, AV_LOG_ERROR, "DTS is invalid.\n");
1933             return AVERROR_EXTERNAL;
1934         }
1935     }
1936
1937     dts_delta = vtctx->dts_delta >= 0 ? vtctx->dts_delta : 0;
1938     time_base_num = avctx->time_base.num;
1939     pkt->pts = pts.value / time_base_num;
1940     pkt->dts = dts.value / time_base_num - dts_delta;
1941     pkt->size = out_buf_size;
1942
1943     return 0;
1944 }
1945
1946 /*
1947  * contiguous_buf_size is 0 if not contiguous, and the size of the buffer
1948  * containing all planes if so.
1949  */
1950 static int get_cv_pixel_info(
1951     AVCodecContext *avctx,
1952     const AVFrame  *frame,
1953     int            *color,
1954     int            *plane_count,
1955     size_t         *widths,
1956     size_t         *heights,
1957     size_t         *strides,
1958     size_t         *contiguous_buf_size)
1959 {
1960     VTEncContext *vtctx = avctx->priv_data;
1961     int av_format       = frame->format;
1962     int av_color_range  = frame->color_range;
1963     int i;
1964     int range_guessed;
1965     int status;
1966
1967     status = get_cv_pixel_format(avctx, av_format, av_color_range, color, &range_guessed);
1968     if (status) {
1969         av_log(avctx,
1970             AV_LOG_ERROR,
1971             "Could not get pixel format for color format '%s' range '%s'.\n",
1972             av_get_pix_fmt_name(av_format),
1973             av_color_range > AVCOL_RANGE_UNSPECIFIED &&
1974             av_color_range < AVCOL_RANGE_NB ?
1975                av_color_range_name(av_color_range) :
1976                "Unknown");
1977
1978         return AVERROR(EINVAL);
1979     }
1980
1981     if (range_guessed) {
1982         if (!vtctx->warned_color_range) {
1983             vtctx->warned_color_range = true;
1984             av_log(avctx,
1985                    AV_LOG_WARNING,
1986                    "Color range not set for %s. Using MPEG range.\n",
1987                    av_get_pix_fmt_name(av_format));
1988         }
1989     }
1990
1991     switch (av_format) {
1992     case AV_PIX_FMT_NV12:
1993         *plane_count = 2;
1994
1995         widths [0] = avctx->width;
1996         heights[0] = avctx->height;
1997         strides[0] = frame ? frame->linesize[0] : avctx->width;
1998
1999         widths [1] = (avctx->width  + 1) / 2;
2000         heights[1] = (avctx->height + 1) / 2;
2001         strides[1] = frame ? frame->linesize[1] : (avctx->width + 1) & -2;
2002         break;
2003
2004     case AV_PIX_FMT_YUV420P:
2005         *plane_count = 3;
2006
2007         widths [0] = avctx->width;
2008         heights[0] = avctx->height;
2009         strides[0] = frame ? frame->linesize[0] : avctx->width;
2010
2011         widths [1] = (avctx->width  + 1) / 2;
2012         heights[1] = (avctx->height + 1) / 2;
2013         strides[1] = frame ? frame->linesize[1] : (avctx->width + 1) / 2;
2014
2015         widths [2] = (avctx->width  + 1) / 2;
2016         heights[2] = (avctx->height + 1) / 2;
2017         strides[2] = frame ? frame->linesize[2] : (avctx->width + 1) / 2;
2018         break;
2019
2020     case AV_PIX_FMT_P010LE:
2021         *plane_count = 2;
2022         widths[0] = avctx->width;
2023         heights[0] = avctx->height;
2024         strides[0] = frame ? frame->linesize[0] : (avctx->width * 2 + 63) & -64;
2025
2026         widths[1] = (avctx->width + 1) / 2;
2027         heights[1] = (avctx->height + 1) / 2;
2028         strides[1] = frame ? frame->linesize[1] : ((avctx->width + 1) / 2 + 63) & -64;
2029         break;
2030
2031     default:
2032         av_log(
2033                avctx,
2034                AV_LOG_ERROR,
2035                "Could not get frame format info for color %d range %d.\n",
2036                av_format,
2037                av_color_range);
2038
2039         return AVERROR(EINVAL);
2040     }
2041
2042     *contiguous_buf_size = 0;
2043     for (i = 0; i < *plane_count; i++) {
2044         if (i < *plane_count - 1 &&
2045             frame->data[i] + strides[i] * heights[i] != frame->data[i + 1]) {
2046             *contiguous_buf_size = 0;
2047             break;
2048         }
2049
2050         *contiguous_buf_size += strides[i] * heights[i];
2051     }
2052
2053     return 0;
2054 }
2055
2056 //Not used on OSX - frame is never copied.
2057 static int copy_avframe_to_pixel_buffer(AVCodecContext   *avctx,
2058                                         const AVFrame    *frame,
2059                                         CVPixelBufferRef cv_img,
2060                                         const size_t     *plane_strides,
2061                                         const size_t     *plane_rows)
2062 {
2063     int i, j;
2064     size_t plane_count;
2065     int status;
2066     int rows;
2067     int src_stride;
2068     int dst_stride;
2069     uint8_t *src_addr;
2070     uint8_t *dst_addr;
2071     size_t copy_bytes;
2072
2073     status = CVPixelBufferLockBaseAddress(cv_img, 0);
2074     if (status) {
2075         av_log(
2076             avctx,
2077             AV_LOG_ERROR,
2078             "Error: Could not lock base address of CVPixelBuffer: %d.\n",
2079             status
2080         );
2081     }
2082
2083     if (CVPixelBufferIsPlanar(cv_img)) {
2084         plane_count = CVPixelBufferGetPlaneCount(cv_img);
2085         for (i = 0; frame->data[i]; i++) {
2086             if (i == plane_count) {
2087                 CVPixelBufferUnlockBaseAddress(cv_img, 0);
2088                 av_log(avctx,
2089                     AV_LOG_ERROR,
2090                     "Error: different number of planes in AVFrame and CVPixelBuffer.\n"
2091                 );
2092
2093                 return AVERROR_EXTERNAL;
2094             }
2095
2096             dst_addr = (uint8_t*)CVPixelBufferGetBaseAddressOfPlane(cv_img, i);
2097             src_addr = (uint8_t*)frame->data[i];
2098             dst_stride = CVPixelBufferGetBytesPerRowOfPlane(cv_img, i);
2099             src_stride = plane_strides[i];
2100             rows = plane_rows[i];
2101
2102             if (dst_stride == src_stride) {
2103                 memcpy(dst_addr, src_addr, src_stride * rows);
2104             } else {
2105                 copy_bytes = dst_stride < src_stride ? dst_stride : src_stride;
2106
2107                 for (j = 0; j < rows; j++) {
2108                     memcpy(dst_addr + j * dst_stride, src_addr + j * src_stride, copy_bytes);
2109                 }
2110             }
2111         }
2112     } else {
2113         if (frame->data[1]) {
2114             CVPixelBufferUnlockBaseAddress(cv_img, 0);
2115             av_log(avctx,
2116                 AV_LOG_ERROR,
2117                 "Error: different number of planes in AVFrame and non-planar CVPixelBuffer.\n"
2118             );
2119
2120             return AVERROR_EXTERNAL;
2121         }
2122
2123         dst_addr = (uint8_t*)CVPixelBufferGetBaseAddress(cv_img);
2124         src_addr = (uint8_t*)frame->data[0];
2125         dst_stride = CVPixelBufferGetBytesPerRow(cv_img);
2126         src_stride = plane_strides[0];
2127         rows = plane_rows[0];
2128
2129         if (dst_stride == src_stride) {
2130             memcpy(dst_addr, src_addr, src_stride * rows);
2131         } else {
2132             copy_bytes = dst_stride < src_stride ? dst_stride : src_stride;
2133
2134             for (j = 0; j < rows; j++) {
2135                 memcpy(dst_addr + j * dst_stride, src_addr + j * src_stride, copy_bytes);
2136             }
2137         }
2138     }
2139
2140     status = CVPixelBufferUnlockBaseAddress(cv_img, 0);
2141     if (status) {
2142         av_log(avctx, AV_LOG_ERROR, "Error: Could not unlock CVPixelBuffer base address: %d.\n", status);
2143         return AVERROR_EXTERNAL;
2144     }
2145
2146     return 0;
2147 }
2148
2149 static int create_cv_pixel_buffer(AVCodecContext   *avctx,
2150                                   const AVFrame    *frame,
2151                                   CVPixelBufferRef *cv_img)
2152 {
2153     int plane_count;
2154     int color;
2155     size_t widths [AV_NUM_DATA_POINTERS];
2156     size_t heights[AV_NUM_DATA_POINTERS];
2157     size_t strides[AV_NUM_DATA_POINTERS];
2158     int status;
2159     size_t contiguous_buf_size;
2160     CVPixelBufferPoolRef pix_buf_pool;
2161     VTEncContext* vtctx = avctx->priv_data;
2162
2163     if (avctx->pix_fmt == AV_PIX_FMT_VIDEOTOOLBOX) {
2164         av_assert0(frame->format == AV_PIX_FMT_VIDEOTOOLBOX);
2165
2166         *cv_img = (CVPixelBufferRef)frame->data[3];
2167         av_assert0(*cv_img);
2168
2169         CFRetain(*cv_img);
2170         return 0;
2171     }
2172
2173     memset(widths,  0, sizeof(widths));
2174     memset(heights, 0, sizeof(heights));
2175     memset(strides, 0, sizeof(strides));
2176
2177     status = get_cv_pixel_info(
2178         avctx,
2179         frame,
2180         &color,
2181         &plane_count,
2182         widths,
2183         heights,
2184         strides,
2185         &contiguous_buf_size
2186     );
2187
2188     if (status) {
2189         av_log(
2190             avctx,
2191             AV_LOG_ERROR,
2192             "Error: Cannot convert format %d color_range %d: %d\n",
2193             frame->format,
2194             frame->color_range,
2195             status
2196         );
2197
2198         return AVERROR_EXTERNAL;
2199     }
2200
2201     pix_buf_pool = VTCompressionSessionGetPixelBufferPool(vtctx->session);
2202     if (!pix_buf_pool) {
2203         /* On iOS, the VT session is invalidated when the APP switches from
2204          * foreground to background and vice versa. Fetch the actual error code
2205          * of the VT session to detect that case and restart the VT session
2206          * accordingly. */
2207         OSStatus vtstatus;
2208
2209         vtstatus = VTCompressionSessionPrepareToEncodeFrames(vtctx->session);
2210         if (vtstatus == kVTInvalidSessionErr) {
2211             CFRelease(vtctx->session);
2212             vtctx->session = NULL;
2213             status = vtenc_configure_encoder(avctx);
2214             if (status == 0)
2215                 pix_buf_pool = VTCompressionSessionGetPixelBufferPool(vtctx->session);
2216         }
2217         if (!pix_buf_pool) {
2218             av_log(avctx, AV_LOG_ERROR, "Could not get pixel buffer pool.\n");
2219             return AVERROR_EXTERNAL;
2220         }
2221         else
2222             av_log(avctx, AV_LOG_WARNING, "VT session restarted because of a "
2223                    "kVTInvalidSessionErr error.\n");
2224     }
2225
2226     status = CVPixelBufferPoolCreatePixelBuffer(NULL,
2227                                                 pix_buf_pool,
2228                                                 cv_img);
2229
2230
2231     if (status) {
2232         av_log(avctx, AV_LOG_ERROR, "Could not create pixel buffer from pool: %d.\n", status);
2233         return AVERROR_EXTERNAL;
2234     }
2235
2236     status = copy_avframe_to_pixel_buffer(avctx, frame, *cv_img, strides, heights);
2237     if (status) {
2238         CFRelease(*cv_img);
2239         *cv_img = NULL;
2240         return status;
2241     }
2242
2243     return 0;
2244 }
2245
2246 static int create_encoder_dict_h264(const AVFrame *frame,
2247                                     CFDictionaryRef* dict_out)
2248 {
2249     CFDictionaryRef dict = NULL;
2250     if (frame->pict_type == AV_PICTURE_TYPE_I) {
2251         const void *keys[] = { kVTEncodeFrameOptionKey_ForceKeyFrame };
2252         const void *vals[] = { kCFBooleanTrue };
2253
2254         dict = CFDictionaryCreate(NULL, keys, vals, 1, NULL, NULL);
2255         if(!dict) return AVERROR(ENOMEM);
2256     }
2257
2258     *dict_out = dict;
2259     return 0;
2260 }
2261
2262 static int vtenc_send_frame(AVCodecContext *avctx,
2263                             VTEncContext   *vtctx,
2264                             const AVFrame  *frame)
2265 {
2266     CMTime time;
2267     CFDictionaryRef frame_dict;
2268     CVPixelBufferRef cv_img = NULL;
2269     AVFrameSideData *side_data = NULL;
2270     ExtraSEI *sei = NULL;
2271     int status = create_cv_pixel_buffer(avctx, frame, &cv_img);
2272
2273     if (status) return status;
2274
2275     status = create_encoder_dict_h264(frame, &frame_dict);
2276     if (status) {
2277         CFRelease(cv_img);
2278         return status;
2279     }
2280
2281     side_data = av_frame_get_side_data(frame, AV_FRAME_DATA_A53_CC);
2282     if (vtctx->a53_cc && side_data && side_data->size) {
2283         sei = av_mallocz(sizeof(*sei));
2284         if (!sei) {
2285             av_log(avctx, AV_LOG_ERROR, "Not enough memory for closed captions, skipping\n");
2286         } else {
2287             int ret = ff_alloc_a53_sei(frame, 0, &sei->data, &sei->size);
2288             if (ret < 0) {
2289                 av_log(avctx, AV_LOG_ERROR, "Not enough memory for closed captions, skipping\n");
2290                 av_free(sei);
2291                 sei = NULL;
2292             }
2293         }
2294     }
2295
2296     time = CMTimeMake(frame->pts * avctx->time_base.num, avctx->time_base.den);
2297     status = VTCompressionSessionEncodeFrame(
2298         vtctx->session,
2299         cv_img,
2300         time,
2301         kCMTimeInvalid,
2302         frame_dict,
2303         sei,
2304         NULL
2305     );
2306
2307     if (frame_dict) CFRelease(frame_dict);
2308     CFRelease(cv_img);
2309
2310     if (status) {
2311         av_log(avctx, AV_LOG_ERROR, "Error: cannot encode frame: %d\n", status);
2312         return AVERROR_EXTERNAL;
2313     }
2314
2315     return 0;
2316 }
2317
2318 static av_cold int vtenc_frame(
2319     AVCodecContext *avctx,
2320     AVPacket       *pkt,
2321     const AVFrame  *frame,
2322     int            *got_packet)
2323 {
2324     VTEncContext *vtctx = avctx->priv_data;
2325     bool get_frame;
2326     int status;
2327     CMSampleBufferRef buf = NULL;
2328     ExtraSEI *sei = NULL;
2329
2330     if (frame) {
2331         status = vtenc_send_frame(avctx, vtctx, frame);
2332
2333         if (status) {
2334             status = AVERROR_EXTERNAL;
2335             goto end_nopkt;
2336         }
2337
2338         if (vtctx->frame_ct_in == 0) {
2339             vtctx->first_pts = frame->pts;
2340         } else if(vtctx->frame_ct_in == 1 && vtctx->has_b_frames) {
2341             vtctx->dts_delta = frame->pts - vtctx->first_pts;
2342         }
2343
2344         vtctx->frame_ct_in++;
2345     } else if(!vtctx->flushing) {
2346         vtctx->flushing = true;
2347
2348         status = VTCompressionSessionCompleteFrames(vtctx->session,
2349                                                     kCMTimeIndefinite);
2350
2351         if (status) {
2352             av_log(avctx, AV_LOG_ERROR, "Error flushing frames: %d\n", status);
2353             status = AVERROR_EXTERNAL;
2354             goto end_nopkt;
2355         }
2356     }
2357
2358     *got_packet = 0;
2359     get_frame = vtctx->dts_delta >= 0 || !frame;
2360     if (!get_frame) {
2361         status = 0;
2362         goto end_nopkt;
2363     }
2364
2365     status = vtenc_q_pop(vtctx, !frame, &buf, &sei);
2366     if (status) goto end_nopkt;
2367     if (!buf)   goto end_nopkt;
2368
2369     status = vtenc_cm_to_avpacket(avctx, buf, pkt, sei);
2370     if (sei) {
2371         if (sei->data) av_free(sei->data);
2372         av_free(sei);
2373     }
2374     CFRelease(buf);
2375     if (status) goto end_nopkt;
2376
2377     *got_packet = 1;
2378     return 0;
2379
2380 end_nopkt:
2381     av_packet_unref(pkt);
2382     return status;
2383 }
2384
2385 static int vtenc_populate_extradata(AVCodecContext   *avctx,
2386                                     CMVideoCodecType codec_type,
2387                                     CFStringRef      profile_level,
2388                                     CFNumberRef      gamma_level,
2389                                     CFDictionaryRef  enc_info,
2390                                     CFDictionaryRef  pixel_buffer_info)
2391 {
2392     VTEncContext *vtctx = avctx->priv_data;
2393     int status;
2394     CVPixelBufferPoolRef pool = NULL;
2395     CVPixelBufferRef pix_buf = NULL;
2396     CMTime time;
2397     CMSampleBufferRef buf = NULL;
2398
2399     status = vtenc_create_encoder(avctx,
2400                                   codec_type,
2401                                   profile_level,
2402                                   gamma_level,
2403                                   enc_info,
2404                                   pixel_buffer_info,
2405                                   &vtctx->session);
2406     if (status)
2407         goto pe_cleanup;
2408
2409     pool = VTCompressionSessionGetPixelBufferPool(vtctx->session);
2410     if(!pool){
2411         av_log(avctx, AV_LOG_ERROR, "Error getting pixel buffer pool.\n");
2412         goto pe_cleanup;
2413     }
2414
2415     status = CVPixelBufferPoolCreatePixelBuffer(NULL,
2416                                                 pool,
2417                                                 &pix_buf);
2418
2419     if(status != kCVReturnSuccess){
2420         av_log(avctx, AV_LOG_ERROR, "Error creating frame from pool: %d\n", status);
2421         goto pe_cleanup;
2422     }
2423
2424     time = CMTimeMake(0, avctx->time_base.den);
2425     status = VTCompressionSessionEncodeFrame(vtctx->session,
2426                                              pix_buf,
2427                                              time,
2428                                              kCMTimeInvalid,
2429                                              NULL,
2430                                              NULL,
2431                                              NULL);
2432
2433     if (status) {
2434         av_log(avctx,
2435                AV_LOG_ERROR,
2436                "Error sending frame for extradata: %d\n",
2437                status);
2438
2439         goto pe_cleanup;
2440     }
2441
2442     //Populates extradata - output frames are flushed and param sets are available.
2443     status = VTCompressionSessionCompleteFrames(vtctx->session,
2444                                                 kCMTimeIndefinite);
2445
2446     if (status)
2447         goto pe_cleanup;
2448
2449     status = vtenc_q_pop(vtctx, 0, &buf, NULL);
2450     if (status) {
2451         av_log(avctx, AV_LOG_ERROR, "popping: %d\n", status);
2452         goto pe_cleanup;
2453     }
2454
2455     CFRelease(buf);
2456
2457
2458
2459 pe_cleanup:
2460     if(vtctx->session)
2461         CFRelease(vtctx->session);
2462
2463     vtctx->session = NULL;
2464     vtctx->frame_ct_out = 0;
2465
2466     av_assert0(status != 0 || (avctx->extradata && avctx->extradata_size > 0));
2467
2468     return status;
2469 }
2470
2471 static av_cold int vtenc_close(AVCodecContext *avctx)
2472 {
2473     VTEncContext *vtctx = avctx->priv_data;
2474
2475     pthread_cond_destroy(&vtctx->cv_sample_sent);
2476     pthread_mutex_destroy(&vtctx->lock);
2477
2478     if(!vtctx->session) return 0;
2479
2480     VTCompressionSessionCompleteFrames(vtctx->session,
2481                                        kCMTimeIndefinite);
2482     clear_frame_queue(vtctx);
2483     CFRelease(vtctx->session);
2484     vtctx->session = NULL;
2485
2486     if (vtctx->color_primaries) {
2487         CFRelease(vtctx->color_primaries);
2488         vtctx->color_primaries = NULL;
2489     }
2490
2491     if (vtctx->transfer_function) {
2492         CFRelease(vtctx->transfer_function);
2493         vtctx->transfer_function = NULL;
2494     }
2495
2496     if (vtctx->ycbcr_matrix) {
2497         CFRelease(vtctx->ycbcr_matrix);
2498         vtctx->ycbcr_matrix = NULL;
2499     }
2500
2501     return 0;
2502 }
2503
2504 static const enum AVPixelFormat avc_pix_fmts[] = {
2505     AV_PIX_FMT_VIDEOTOOLBOX,
2506     AV_PIX_FMT_NV12,
2507     AV_PIX_FMT_YUV420P,
2508     AV_PIX_FMT_NONE
2509 };
2510
2511 static const enum AVPixelFormat hevc_pix_fmts[] = {
2512     AV_PIX_FMT_VIDEOTOOLBOX,
2513     AV_PIX_FMT_NV12,
2514     AV_PIX_FMT_YUV420P,
2515     AV_PIX_FMT_P010LE,
2516     AV_PIX_FMT_NONE
2517 };
2518
2519 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
2520 #define COMMON_OPTIONS \
2521     { "allow_sw", "Allow software encoding", OFFSET(allow_sw), AV_OPT_TYPE_BOOL, \
2522         { .i64 = 0 }, 0, 1, VE }, \
2523     { "require_sw", "Require software encoding", OFFSET(require_sw), AV_OPT_TYPE_BOOL, \
2524         { .i64 = 0 }, 0, 1, VE }, \
2525     { "realtime", "Hint that encoding should happen in real-time if not faster (e.g. capturing from camera).", \
2526         OFFSET(realtime), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, \
2527     { "frames_before", "Other frames will come before the frames in this session. This helps smooth concatenation issues.", \
2528         OFFSET(frames_before), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, \
2529     { "frames_after", "Other frames will come after the frames in this session. This helps smooth concatenation issues.", \
2530         OFFSET(frames_after), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
2531
2532 #define OFFSET(x) offsetof(VTEncContext, x)
2533 static const AVOption h264_options[] = {
2534     { "profile", "Profile", OFFSET(profile), AV_OPT_TYPE_INT, { .i64 = H264_PROF_AUTO }, H264_PROF_AUTO, H264_PROF_COUNT, VE, "profile" },
2535     { "baseline", "Baseline Profile", 0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_BASELINE }, INT_MIN, INT_MAX, VE, "profile" },
2536     { "main",     "Main Profile",     0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_MAIN     }, INT_MIN, INT_MAX, VE, "profile" },
2537     { "high",     "High Profile",     0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_HIGH     }, INT_MIN, INT_MAX, VE, "profile" },
2538
2539     { "level", "Level", OFFSET(level), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 52, VE, "level" },
2540     { "1.3", "Level 1.3, only available with Baseline Profile", 0, AV_OPT_TYPE_CONST, { .i64 = 13 }, INT_MIN, INT_MAX, VE, "level" },
2541     { "3.0", "Level 3.0", 0, AV_OPT_TYPE_CONST, { .i64 = 30 }, INT_MIN, INT_MAX, VE, "level" },
2542     { "3.1", "Level 3.1", 0, AV_OPT_TYPE_CONST, { .i64 = 31 }, INT_MIN, INT_MAX, VE, "level" },
2543     { "3.2", "Level 3.2", 0, AV_OPT_TYPE_CONST, { .i64 = 32 }, INT_MIN, INT_MAX, VE, "level" },
2544     { "4.0", "Level 4.0", 0, AV_OPT_TYPE_CONST, { .i64 = 40 }, INT_MIN, INT_MAX, VE, "level" },
2545     { "4.1", "Level 4.1", 0, AV_OPT_TYPE_CONST, { .i64 = 41 }, INT_MIN, INT_MAX, VE, "level" },
2546     { "4.2", "Level 4.2", 0, AV_OPT_TYPE_CONST, { .i64 = 42 }, INT_MIN, INT_MAX, VE, "level" },
2547     { "5.0", "Level 5.0", 0, AV_OPT_TYPE_CONST, { .i64 = 50 }, INT_MIN, INT_MAX, VE, "level" },
2548     { "5.1", "Level 5.1", 0, AV_OPT_TYPE_CONST, { .i64 = 51 }, INT_MIN, INT_MAX, VE, "level" },
2549     { "5.2", "Level 5.2", 0, AV_OPT_TYPE_CONST, { .i64 = 52 }, INT_MIN, INT_MAX, VE, "level" },
2550
2551     { "coder", "Entropy coding", OFFSET(entropy), AV_OPT_TYPE_INT, { .i64 = VT_ENTROPY_NOT_SET }, VT_ENTROPY_NOT_SET, VT_CABAC, VE, "coder" },
2552     { "cavlc", "CAVLC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CAVLC }, INT_MIN, INT_MAX, VE, "coder" },
2553     { "vlc",   "CAVLC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CAVLC }, INT_MIN, INT_MAX, VE, "coder" },
2554     { "cabac", "CABAC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CABAC }, INT_MIN, INT_MAX, VE, "coder" },
2555     { "ac",    "CABAC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CABAC }, INT_MIN, INT_MAX, VE, "coder" },
2556
2557     { "a53cc", "Use A53 Closed Captions (if available)", OFFSET(a53_cc), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, VE },
2558
2559     COMMON_OPTIONS
2560     { NULL },
2561 };
2562
2563 static const AVClass h264_videotoolbox_class = {
2564     .class_name = "h264_videotoolbox",
2565     .item_name  = av_default_item_name,
2566     .option     = h264_options,
2567     .version    = LIBAVUTIL_VERSION_INT,
2568 };
2569
2570 AVCodec ff_h264_videotoolbox_encoder = {
2571     .name             = "h264_videotoolbox",
2572     .long_name        = NULL_IF_CONFIG_SMALL("VideoToolbox H.264 Encoder"),
2573     .type             = AVMEDIA_TYPE_VIDEO,
2574     .id               = AV_CODEC_ID_H264,
2575     .priv_data_size   = sizeof(VTEncContext),
2576     .pix_fmts         = avc_pix_fmts,
2577     .init             = vtenc_init,
2578     .encode2          = vtenc_frame,
2579     .close            = vtenc_close,
2580     .capabilities     = AV_CODEC_CAP_DELAY,
2581     .priv_class       = &h264_videotoolbox_class,
2582     .caps_internal    = FF_CODEC_CAP_INIT_THREADSAFE |
2583                         FF_CODEC_CAP_INIT_CLEANUP,
2584 };
2585
2586 static const AVOption hevc_options[] = {
2587     { "profile", "Profile", OFFSET(profile), AV_OPT_TYPE_INT, { .i64 = HEVC_PROF_AUTO }, HEVC_PROF_AUTO, HEVC_PROF_COUNT, VE, "profile" },
2588     { "main",     "Main Profile",     0, AV_OPT_TYPE_CONST, { .i64 = HEVC_PROF_MAIN   }, INT_MIN, INT_MAX, VE, "profile" },
2589     { "main10",   "Main10 Profile",   0, AV_OPT_TYPE_CONST, { .i64 = HEVC_PROF_MAIN10 }, INT_MIN, INT_MAX, VE, "profile" },
2590
2591     COMMON_OPTIONS
2592     { NULL },
2593 };
2594
2595 static const AVClass hevc_videotoolbox_class = {
2596     .class_name = "hevc_videotoolbox",
2597     .item_name  = av_default_item_name,
2598     .option     = hevc_options,
2599     .version    = LIBAVUTIL_VERSION_INT,
2600 };
2601
2602 AVCodec ff_hevc_videotoolbox_encoder = {
2603     .name             = "hevc_videotoolbox",
2604     .long_name        = NULL_IF_CONFIG_SMALL("VideoToolbox H.265 Encoder"),
2605     .type             = AVMEDIA_TYPE_VIDEO,
2606     .id               = AV_CODEC_ID_HEVC,
2607     .priv_data_size   = sizeof(VTEncContext),
2608     .pix_fmts         = hevc_pix_fmts,
2609     .init             = vtenc_init,
2610     .encode2          = vtenc_frame,
2611     .close            = vtenc_close,
2612     .capabilities     = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_HARDWARE,
2613     .priv_class       = &hevc_videotoolbox_class,
2614     .caps_internal    = FF_CODEC_CAP_INIT_THREADSAFE |
2615                         FF_CODEC_CAP_INIT_CLEANUP,
2616     .wrapper_name     = "videotoolbox",
2617 };