libavcodec/videotoolbox.c

   1 /*
   2  * Videotoolbox hardware acceleration
   3  *
   4  * copyright (c) 2012 Sebastien Zwickert
   5  *
   6  * This file is part of FFmpeg.
   7  *
   8  * FFmpeg is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License as published by the Free Software Foundation; either
  11  * version 2.1 of the License, or (at your option) any later version.
  12  *
  13  * FFmpeg is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  16  * Lesser General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU Lesser General Public
  19  * License along with FFmpeg; if not, write to the Free Software
  20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21  */
  22
  23 #include "config.h"
  24 #if CONFIG_VIDEOTOOLBOX
  25 #  include "videotoolbox.h"
  26 #  include "libavutil/hwcontext_videotoolbox.h"
  27 #else
  28 #  include "vda.h"
  29 #endif
  30 #include "vda_vt_internal.h"
  31 #include "libavutil/avutil.h"
  32 #include "libavutil/hwcontext.h"
  33 #include "bytestream.h"
  34 #include "h264dec.h"
  35 #include "mpegvideo.h"
  36 #include <TargetConditionals.h>
  37
  38 #ifndef kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder
  39 #  define kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder CFSTR("RequireHardwareAcceleratedVideoDecoder")
  40 #endif
  41
  42 #define VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING  12
  43
  44 static void videotoolbox_buffer_release(void *opaque, uint8_t *data)
  45 {
  46     CVPixelBufferRef cv_buffer = (CVImageBufferRef)data;
  47     CVPixelBufferRelease(cv_buffer);
  48 }
  49
  50 static int videotoolbox_buffer_copy(VTContext *vtctx,
  51                                     const uint8_t *buffer,
  52                                     uint32_t size)
  53 {
  54     void *tmp;
  55
  56     tmp = av_fast_realloc(vtctx->bitstream,
  57                          &vtctx->allocated_size,
  58                          size);
  59
  60     if (!tmp)
  61         return AVERROR(ENOMEM);
  62
  63     vtctx->bitstream = tmp;
  64     memcpy(vtctx->bitstream, buffer, size);
  65     vtctx->bitstream_size = size;
  66
  67     return 0;
  68 }
  69
  70 int ff_videotoolbox_alloc_frame(AVCodecContext *avctx, AVFrame *frame)
  71 {
  72     frame->width  = avctx->width;
  73     frame->height = avctx->height;
  74     frame->format = avctx->pix_fmt;
  75     frame->buf[0] = av_buffer_alloc(1);
  76
  77     if (!frame->buf[0])
  78         return AVERROR(ENOMEM);
  79
  80     return 0;
  81 }
  82
  83 #define AV_W8(p, v) *(p) = (v)
  84
  85 CFDataRef ff_videotoolbox_avcc_extradata_create(AVCodecContext *avctx)
  86 {
  87     H264Context *h     = avctx->priv_data;
  88     CFDataRef data = NULL;
  89     uint8_t *p;
  90     int vt_extradata_size = 6 + 2 + h->ps.sps->data_size + 3 + h->ps.pps->data_size;
  91     uint8_t *vt_extradata = av_malloc(vt_extradata_size);
  92     if (!vt_extradata)
  93         return NULL;
  94
  95     p = vt_extradata;
  96
  97     AV_W8(p + 0, 1); /* version */
  98     AV_W8(p + 1, h->ps.sps->data[1]); /* profile */
  99     AV_W8(p + 2, h->ps.sps->data[2]); /* profile compat */
 100     AV_W8(p + 3, h->ps.sps->data[3]); /* level */
 101     AV_W8(p + 4, 0xff); /* 6 bits reserved (111111) + 2 bits nal size length - 3 (11) */
 102     AV_W8(p + 5, 0xe1); /* 3 bits reserved (111) + 5 bits number of sps (00001) */
 103     AV_WB16(p + 6, h->ps.sps->data_size);
 104     memcpy(p + 8, h->ps.sps->data, h->ps.sps->data_size);
 105     p += 8 + h->ps.sps->data_size;
 106     AV_W8(p + 0, 1); /* number of pps */
 107     AV_WB16(p + 1, h->ps.pps->data_size);
 108     memcpy(p + 3, h->ps.pps->data, h->ps.pps->data_size);
 109
 110     p += 3 + h->ps.pps->data_size;
 111     av_assert0(p - vt_extradata == vt_extradata_size);
 112
 113     data = CFDataCreate(kCFAllocatorDefault, vt_extradata, vt_extradata_size);
 114     av_free(vt_extradata);
 115     return data;
 116 }
 117
 118 int ff_videotoolbox_buffer_create(VTContext *vtctx, AVFrame *frame)
 119 {
 120     av_buffer_unref(&frame->buf[0]);
 121
 122     frame->buf[0] = av_buffer_create((uint8_t*)vtctx->frame,
 123                                      sizeof(vtctx->frame),
 124                                      videotoolbox_buffer_release,
 125                                      NULL,
 126                                      AV_BUFFER_FLAG_READONLY);
 127     if (!frame->buf[0]) {
 128         return AVERROR(ENOMEM);
 129     }
 130
 131     frame->data[3] = (uint8_t*)vtctx->frame;
 132     vtctx->frame = NULL;
 133
 134     return 0;
 135 }
 136
 137 int ff_videotoolbox_h264_start_frame(AVCodecContext *avctx,
 138                                      const uint8_t *buffer,
 139                                      uint32_t size)
 140 {
 141     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 142     H264Context *h  = avctx->priv_data;
 143
 144     vtctx->bitstream_size = 0;
 145
 146     if (h->is_avc == 1) {
 147         return videotoolbox_buffer_copy(vtctx, buffer, size);
 148     }
 149
 150     return 0;
 151 }
 152
 153 int ff_videotoolbox_h264_decode_slice(AVCodecContext *avctx,
 154                                       const uint8_t *buffer,
 155                                       uint32_t size)
 156 {
 157     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 158     H264Context *h  = avctx->priv_data;
 159     void *tmp;
 160
 161     if (h->is_avc == 1)
 162         return 0;
 163
 164     tmp = av_fast_realloc(vtctx->bitstream,
 165                           &vtctx->allocated_size,
 166                           vtctx->bitstream_size+size+4);
 167     if (!tmp)
 168         return AVERROR(ENOMEM);
 169
 170     vtctx->bitstream = tmp;
 171
 172     AV_WB32(vtctx->bitstream + vtctx->bitstream_size, size);
 173     memcpy(vtctx->bitstream + vtctx->bitstream_size + 4, buffer, size);
 174
 175     vtctx->bitstream_size += size + 4;
 176
 177     return 0;
 178 }
 179
 180 int ff_videotoolbox_uninit(AVCodecContext *avctx)
 181 {
 182     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 183     if (vtctx) {
 184         av_freep(&vtctx->bitstream);
 185         if (vtctx->frame)
 186             CVPixelBufferRelease(vtctx->frame);
 187     }
 188
 189     return 0;
 190 }
 191
 192 #if CONFIG_VIDEOTOOLBOX
 193 // Return the AVVideotoolboxContext that matters currently. Where it comes from
 194 // depends on the API used.
 195 static AVVideotoolboxContext *videotoolbox_get_context(AVCodecContext *avctx)
 196 {
 197     // Somewhat tricky because the user can call av_videotoolbox_default_free()
 198     // at any time, even when the codec is closed.
 199     if (avctx->internal && avctx->internal->hwaccel_priv_data) {
 200         VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 201         if (vtctx->vt_ctx)
 202             return vtctx->vt_ctx;
 203     }
 204     return avctx->hwaccel_context;
 205 }
 206
 207 static int videotoolbox_buffer_create(AVCodecContext *avctx, AVFrame *frame)
 208 {
 209     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 210     CVPixelBufferRef pixbuf = (CVPixelBufferRef)vtctx->frame;
 211     OSType pixel_format = CVPixelBufferGetPixelFormatType(pixbuf);
 212     enum AVPixelFormat sw_format = av_map_videotoolbox_format_to_pixfmt(pixel_format);
 213     int width = CVPixelBufferGetWidth(pixbuf);
 214     int height = CVPixelBufferGetHeight(pixbuf);
 215     AVHWFramesContext *cached_frames;
 216     int ret;
 217
 218     ret = ff_videotoolbox_buffer_create(vtctx, frame);
 219     if (ret < 0)
 220         return ret;
 221
 222     // Old API code path.
 223     if (!vtctx->cached_hw_frames_ctx)
 224         return 0;
 225
 226     cached_frames = (AVHWFramesContext*)vtctx->cached_hw_frames_ctx->data;
 227
 228     if (cached_frames->sw_format != sw_format ||
 229         cached_frames->width != width ||
 230         cached_frames->height != height) {
 231         AVBufferRef *hw_frames_ctx = av_hwframe_ctx_alloc(cached_frames->device_ref);
 232         AVHWFramesContext *hw_frames;
 233         if (!hw_frames_ctx)
 234             return AVERROR(ENOMEM);
 235
 236         hw_frames = (AVHWFramesContext*)hw_frames_ctx->data;
 237         hw_frames->format = cached_frames->format;
 238         hw_frames->sw_format = sw_format;
 239         hw_frames->width = width;
 240         hw_frames->height = height;
 241
 242         ret = av_hwframe_ctx_init(hw_frames_ctx);
 243         if (ret < 0) {
 244             av_buffer_unref(&hw_frames_ctx);
 245             return ret;
 246         }
 247
 248         av_buffer_unref(&vtctx->cached_hw_frames_ctx);
 249         vtctx->cached_hw_frames_ctx = hw_frames_ctx;
 250     }
 251
 252     av_assert0(!frame->hw_frames_ctx);
 253     frame->hw_frames_ctx = av_buffer_ref(vtctx->cached_hw_frames_ctx);
 254     if (!frame->hw_frames_ctx)
 255         return AVERROR(ENOMEM);
 256
 257     return 0;
 258 }
 259
 260 static void videotoolbox_write_mp4_descr_length(PutByteContext *pb, int length)
 261 {
 262     int i;
 263     uint8_t b;
 264
 265     for (i = 3; i >= 0; i--) {
 266         b = (length >> (i * 7)) & 0x7F;
 267         if (i != 0)
 268             b |= 0x80;
 269
 270         bytestream2_put_byteu(pb, b);
 271     }
 272 }
 273
 274 static CFDataRef videotoolbox_esds_extradata_create(AVCodecContext *avctx)
 275 {
 276     CFDataRef data;
 277     uint8_t *rw_extradata;
 278     PutByteContext pb;
 279     int full_size = 3 + 5 + 13 + 5 + avctx->extradata_size + 3;
 280     // ES_DescrTag data + DecoderConfigDescrTag + data + DecSpecificInfoTag + size + SLConfigDescriptor
 281     int config_size = 13 + 5 + avctx->extradata_size;
 282     int s;
 283
 284     if (!(rw_extradata = av_mallocz(full_size + VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING)))
 285         return NULL;
 286
 287     bytestream2_init_writer(&pb, rw_extradata, full_size + VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING);
 288     bytestream2_put_byteu(&pb, 0);        // version
 289     bytestream2_put_ne24(&pb, 0);         // flags
 290
 291     // elementary stream descriptor
 292     bytestream2_put_byteu(&pb, 0x03);     // ES_DescrTag
 293     videotoolbox_write_mp4_descr_length(&pb, full_size);
 294     bytestream2_put_ne16(&pb, 0);         // esid
 295     bytestream2_put_byteu(&pb, 0);        // stream priority (0-32)
 296
 297     // decoder configuration descriptor
 298     bytestream2_put_byteu(&pb, 0x04);     // DecoderConfigDescrTag
 299     videotoolbox_write_mp4_descr_length(&pb, config_size);
 300     bytestream2_put_byteu(&pb, 32);       // object type indication. 32 = AV_CODEC_ID_MPEG4
 301     bytestream2_put_byteu(&pb, 0x11);     // stream type
 302     bytestream2_put_ne24(&pb, 0);         // buffer size
 303     bytestream2_put_ne32(&pb, 0);         // max bitrate
 304     bytestream2_put_ne32(&pb, 0);         // avg bitrate
 305
 306     // decoder specific descriptor
 307     bytestream2_put_byteu(&pb, 0x05);     ///< DecSpecificInfoTag
 308     videotoolbox_write_mp4_descr_length(&pb, avctx->extradata_size);
 309
 310     bytestream2_put_buffer(&pb, avctx->extradata, avctx->extradata_size);
 311
 312     // SLConfigDescriptor
 313     bytestream2_put_byteu(&pb, 0x06);     // SLConfigDescrTag
 314     bytestream2_put_byteu(&pb, 0x01);     // length
 315     bytestream2_put_byteu(&pb, 0x02);     //
 316
 317     s = bytestream2_size_p(&pb);
 318
 319     data = CFDataCreate(kCFAllocatorDefault, rw_extradata, s);
 320
 321     av_freep(&rw_extradata);
 322     return data;
 323 }
 324
 325 static CMSampleBufferRef videotoolbox_sample_buffer_create(CMFormatDescriptionRef fmt_desc,
 326                                                            void *buffer,
 327                                                            int size)
 328 {
 329     OSStatus status;
 330     CMBlockBufferRef  block_buf;
 331     CMSampleBufferRef sample_buf;
 332
 333     block_buf  = NULL;
 334     sample_buf = NULL;
 335
 336     status = CMBlockBufferCreateWithMemoryBlock(kCFAllocatorDefault,// structureAllocator
 337                                                 buffer,             // memoryBlock
 338                                                 size,               // blockLength
 339                                                 kCFAllocatorNull,   // blockAllocator
 340                                                 NULL,               // customBlockSource
 341                                                 0,                  // offsetToData
 342                                                 size,               // dataLength
 343                                                 0,                  // flags
 344                                                 &block_buf);
 345
 346     if (!status) {
 347         status = CMSampleBufferCreate(kCFAllocatorDefault,  // allocator
 348                                       block_buf,            // dataBuffer
 349                                       TRUE,                 // dataReady
 350                                       0,                    // makeDataReadyCallback
 351                                       0,                    // makeDataReadyRefcon
 352                                       fmt_desc,             // formatDescription
 353                                       1,                    // numSamples
 354                                       0,                    // numSampleTimingEntries
 355                                       NULL,                 // sampleTimingArray
 356                                       0,                    // numSampleSizeEntries
 357                                       NULL,                 // sampleSizeArray
 358                                       &sample_buf);
 359     }
 360
 361     if (block_buf)
 362         CFRelease(block_buf);
 363
 364     return sample_buf;
 365 }
 366
 367 static void videotoolbox_decoder_callback(void *opaque,
 368                                           void *sourceFrameRefCon,
 369                                           OSStatus status,
 370                                           VTDecodeInfoFlags flags,
 371                                           CVImageBufferRef image_buffer,
 372                                           CMTime pts,
 373                                           CMTime duration)
 374 {
 375     AVCodecContext *avctx = opaque;
 376     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 377
 378     if (vtctx->frame) {
 379         CVPixelBufferRelease(vtctx->frame);
 380         vtctx->frame = NULL;
 381     }
 382
 383     if (!image_buffer) {
 384         av_log(NULL, AV_LOG_DEBUG, "vt decoder cb: output image buffer is null\n");
 385         return;
 386     }
 387
 388     vtctx->frame = CVPixelBufferRetain(image_buffer);
 389 }
 390
 391 static OSStatus videotoolbox_session_decode_frame(AVCodecContext *avctx)
 392 {
 393     OSStatus status;
 394     CMSampleBufferRef sample_buf;
 395     AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
 396     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 397
 398     sample_buf = videotoolbox_sample_buffer_create(videotoolbox->cm_fmt_desc,
 399                                                    vtctx->bitstream,
 400                                                    vtctx->bitstream_size);
 401
 402     if (!sample_buf)
 403         return -1;
 404
 405     status = VTDecompressionSessionDecodeFrame(videotoolbox->session,
 406                                                sample_buf,
 407                                                0,       // decodeFlags
 408                                                NULL,    // sourceFrameRefCon
 409                                                0);      // infoFlagsOut
 410     if (status == noErr)
 411         status = VTDecompressionSessionWaitForAsynchronousFrames(videotoolbox->session);
 412
 413     CFRelease(sample_buf);
 414
 415     return status;
 416 }
 417
 418 static int videotoolbox_common_end_frame(AVCodecContext *avctx, AVFrame *frame)
 419 {
 420     int status;
 421     AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
 422     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 423
 424     if (!videotoolbox->session || !vtctx->bitstream)
 425         return AVERROR_INVALIDDATA;
 426
 427     status = videotoolbox_session_decode_frame(avctx);
 428
 429     if (status) {
 430         av_log(avctx, AV_LOG_ERROR, "Failed to decode frame (%d)\n", status);
 431         return AVERROR_UNKNOWN;
 432     }
 433
 434     if (!vtctx->frame)
 435         return AVERROR_UNKNOWN;
 436
 437     return videotoolbox_buffer_create(avctx, frame);
 438 }
 439
 440 static int videotoolbox_h264_end_frame(AVCodecContext *avctx)
 441 {
 442     H264Context *h = avctx->priv_data;
 443     AVFrame *frame = h->cur_pic_ptr->f;
 444
 445     return videotoolbox_common_end_frame(avctx, frame);
 446 }
 447
 448 static int videotoolbox_mpeg_start_frame(AVCodecContext *avctx,
 449                                          const uint8_t *buffer,
 450                                          uint32_t size)
 451 {
 452     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 453
 454     return videotoolbox_buffer_copy(vtctx, buffer, size);
 455 }
 456
 457 static int videotoolbox_mpeg_decode_slice(AVCodecContext *avctx,
 458                                           const uint8_t *buffer,
 459                                           uint32_t size)
 460 {
 461     return 0;
 462 }
 463
 464 static int videotoolbox_mpeg_end_frame(AVCodecContext *avctx)
 465 {
 466     MpegEncContext *s = avctx->priv_data;
 467     AVFrame *frame = s->current_picture_ptr->f;
 468
 469     return videotoolbox_common_end_frame(avctx, frame);
 470 }
 471
 472 static CFDictionaryRef videotoolbox_decoder_config_create(CMVideoCodecType codec_type,
 473                                                           AVCodecContext *avctx)
 474 {
 475     CFMutableDictionaryRef config_info = CFDictionaryCreateMutable(kCFAllocatorDefault,
 476                                                                    0,
 477                                                                    &kCFTypeDictionaryKeyCallBacks,
 478                                                                    &kCFTypeDictionaryValueCallBacks);
 479
 480     CFDictionarySetValue(config_info,
 481                          kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder,
 482                          kCFBooleanTrue);
 483
 484     if (avctx->extradata_size) {
 485         CFMutableDictionaryRef avc_info;
 486         CFDataRef data = NULL;
 487
 488         avc_info = CFDictionaryCreateMutable(kCFAllocatorDefault,
 489                                              1,
 490                                              &kCFTypeDictionaryKeyCallBacks,
 491                                              &kCFTypeDictionaryValueCallBacks);
 492
 493         switch (codec_type) {
 494         case kCMVideoCodecType_MPEG4Video :
 495             data = videotoolbox_esds_extradata_create(avctx);
 496             if (data)
 497                 CFDictionarySetValue(avc_info, CFSTR("esds"), data);
 498             break;
 499         case kCMVideoCodecType_H264 :
 500             data = ff_videotoolbox_avcc_extradata_create(avctx);
 501             if (data)
 502                 CFDictionarySetValue(avc_info, CFSTR("avcC"), data);
 503             break;
 504         default:
 505             break;
 506         }
 507
 508         CFDictionarySetValue(config_info,
 509                 kCMFormatDescriptionExtension_SampleDescriptionExtensionAtoms,
 510                 avc_info);
 511
 512         if (data)
 513             CFRelease(data);
 514
 515         CFRelease(avc_info);
 516     }
 517     return config_info;
 518 }
 519
 520 static CFDictionaryRef videotoolbox_buffer_attributes_create(int width,
 521                                                              int height,
 522                                                              OSType pix_fmt)
 523 {
 524     CFMutableDictionaryRef buffer_attributes;
 525     CFMutableDictionaryRef io_surface_properties;
 526     CFNumberRef cv_pix_fmt;
 527     CFNumberRef w;
 528     CFNumberRef h;
 529
 530     w = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &width);
 531     h = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &height);
 532     cv_pix_fmt = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &pix_fmt);
 533
 534     buffer_attributes = CFDictionaryCreateMutable(kCFAllocatorDefault,
 535                                                   4,
 536                                                   &kCFTypeDictionaryKeyCallBacks,
 537                                                   &kCFTypeDictionaryValueCallBacks);
 538     io_surface_properties = CFDictionaryCreateMutable(kCFAllocatorDefault,
 539                                                       0,
 540                                                       &kCFTypeDictionaryKeyCallBacks,
 541                                                       &kCFTypeDictionaryValueCallBacks);
 542
 543     if (pix_fmt)
 544         CFDictionarySetValue(buffer_attributes, kCVPixelBufferPixelFormatTypeKey, cv_pix_fmt);
 545     CFDictionarySetValue(buffer_attributes, kCVPixelBufferIOSurfacePropertiesKey, io_surface_properties);
 546     CFDictionarySetValue(buffer_attributes, kCVPixelBufferWidthKey, w);
 547     CFDictionarySetValue(buffer_attributes, kCVPixelBufferHeightKey, h);
 548 #if TARGET_OS_IPHONE
 549     CFDictionarySetValue(buffer_attributes, kCVPixelBufferOpenGLESCompatibilityKey, kCFBooleanTrue);
 550 #else
 551     CFDictionarySetValue(buffer_attributes, kCVPixelBufferIOSurfaceOpenGLTextureCompatibilityKey, kCFBooleanTrue);
 552 #endif
 553
 554     CFRelease(io_surface_properties);
 555     CFRelease(cv_pix_fmt);
 556     CFRelease(w);
 557     CFRelease(h);
 558
 559     return buffer_attributes;
 560 }
 561
 562 static CMVideoFormatDescriptionRef videotoolbox_format_desc_create(CMVideoCodecType codec_type,
 563                                                                    CFDictionaryRef decoder_spec,
 564                                                                    int width,
 565                                                                    int height)
 566 {
 567     CMFormatDescriptionRef cm_fmt_desc;
 568     OSStatus status;
 569
 570     status = CMVideoFormatDescriptionCreate(kCFAllocatorDefault,
 571                                             codec_type,
 572                                             width,
 573                                             height,
 574                                             decoder_spec, // Dictionary of extension
 575                                             &cm_fmt_desc);
 576
 577     if (status)
 578         return NULL;
 579
 580     return cm_fmt_desc;
 581 }
 582
 583 static int videotoolbox_default_init(AVCodecContext *avctx)
 584 {
 585     AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
 586     OSStatus status;
 587     VTDecompressionOutputCallbackRecord decoder_cb;
 588     CFDictionaryRef decoder_spec;
 589     CFDictionaryRef buf_attr;
 590
 591     if (!videotoolbox) {
 592         av_log(avctx, AV_LOG_ERROR, "hwaccel context is not set\n");
 593         return -1;
 594     }
 595
 596     switch( avctx->codec_id ) {
 597     case AV_CODEC_ID_H263 :
 598         videotoolbox->cm_codec_type = kCMVideoCodecType_H263;
 599         break;
 600     case AV_CODEC_ID_H264 :
 601         videotoolbox->cm_codec_type = kCMVideoCodecType_H264;
 602         break;
 603     case AV_CODEC_ID_MPEG1VIDEO :
 604         videotoolbox->cm_codec_type = kCMVideoCodecType_MPEG1Video;
 605         break;
 606     case AV_CODEC_ID_MPEG2VIDEO :
 607         videotoolbox->cm_codec_type = kCMVideoCodecType_MPEG2Video;
 608         break;
 609     case AV_CODEC_ID_MPEG4 :
 610         videotoolbox->cm_codec_type = kCMVideoCodecType_MPEG4Video;
 611         break;
 612     default :
 613         break;
 614     }
 615
 616     decoder_spec = videotoolbox_decoder_config_create(videotoolbox->cm_codec_type, avctx);
 617
 618     videotoolbox->cm_fmt_desc = videotoolbox_format_desc_create(videotoolbox->cm_codec_type,
 619                                                                 decoder_spec,
 620                                                                 avctx->width,
 621                                                                 avctx->height);
 622     if (!videotoolbox->cm_fmt_desc) {
 623         if (decoder_spec)
 624             CFRelease(decoder_spec);
 625
 626         av_log(avctx, AV_LOG_ERROR, "format description creation failed\n");
 627         return -1;
 628     }
 629
 630     buf_attr = videotoolbox_buffer_attributes_create(avctx->width,
 631                                                      avctx->height,
 632                                                      videotoolbox->cv_pix_fmt_type);
 633
 634     decoder_cb.decompressionOutputCallback = videotoolbox_decoder_callback;
 635     decoder_cb.decompressionOutputRefCon   = avctx;
 636
 637     status = VTDecompressionSessionCreate(NULL,                      // allocator
 638                                           videotoolbox->cm_fmt_desc, // videoFormatDescription
 639                                           decoder_spec,              // videoDecoderSpecification
 640                                           buf_attr,                  // destinationImageBufferAttributes
 641                                           &decoder_cb,               // outputCallback
 642                                           &videotoolbox->session);   // decompressionSessionOut
 643
 644     if (decoder_spec)
 645         CFRelease(decoder_spec);
 646     if (buf_attr)
 647         CFRelease(buf_attr);
 648
 649     switch (status) {
 650     case kVTVideoDecoderNotAvailableNowErr:
 651     case kVTVideoDecoderUnsupportedDataFormatErr:
 652         return AVERROR(ENOSYS);
 653     case kVTVideoDecoderMalfunctionErr:
 654         return AVERROR(EINVAL);
 655     case kVTVideoDecoderBadDataErr :
 656         return AVERROR_INVALIDDATA;
 657     case 0:
 658         return 0;
 659     default:
 660         return AVERROR_UNKNOWN;
 661     }
 662 }
 663
 664 static void videotoolbox_default_free(AVCodecContext *avctx)
 665 {
 666     AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
 667
 668     if (videotoolbox) {
 669         if (videotoolbox->cm_fmt_desc)
 670             CFRelease(videotoolbox->cm_fmt_desc);
 671
 672         if (videotoolbox->session) {
 673             VTDecompressionSessionInvalidate(videotoolbox->session);
 674             CFRelease(videotoolbox->session);
 675         }
 676     }
 677 }
 678
 679 static int videotoolbox_uninit(AVCodecContext *avctx)
 680 {
 681     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 682     if (!vtctx)
 683         return 0;
 684
 685     ff_videotoolbox_uninit(avctx);
 686
 687     if (vtctx->vt_ctx)
 688         videotoolbox_default_free(avctx);
 689
 690     av_buffer_unref(&vtctx->cached_hw_frames_ctx);
 691     av_freep(&vtctx->vt_ctx);
 692
 693     return 0;
 694 }
 695
 696 static int videotoolbox_common_init(AVCodecContext *avctx)
 697 {
 698     VTContext *vtctx = avctx->internal->hwaccel_priv_data;
 699     AVHWFramesContext *hw_frames;
 700     int err;
 701
 702     // Old API - do nothing.
 703     if (avctx->hwaccel_context)
 704         return 0;
 705
 706     if (!avctx->hw_frames_ctx && !avctx->hw_device_ctx) {
 707         av_log(avctx, AV_LOG_ERROR,
 708                "Either hw_frames_ctx or hw_device_ctx must be set.\n");
 709         return AVERROR(EINVAL);
 710     }
 711
 712     vtctx->vt_ctx = av_videotoolbox_alloc_context();
 713     if (!vtctx->vt_ctx) {
 714         err = AVERROR(ENOMEM);
 715         goto fail;
 716     }
 717
 718     if (avctx->hw_frames_ctx) {
 719         hw_frames = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
 720     } else {
 721         avctx->hw_frames_ctx = av_hwframe_ctx_alloc(avctx->hw_device_ctx);
 722         if (!avctx->hw_frames_ctx) {
 723             err = AVERROR(ENOMEM);
 724             goto fail;
 725         }
 726
 727         hw_frames = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
 728         hw_frames->format = AV_PIX_FMT_VIDEOTOOLBOX;
 729         hw_frames->sw_format = AV_PIX_FMT_NV12; // same as av_videotoolbox_alloc_context()
 730         hw_frames->width = avctx->width;
 731         hw_frames->height = avctx->height;
 732
 733         err = av_hwframe_ctx_init(avctx->hw_frames_ctx);
 734         if (err < 0) {
 735             av_buffer_unref(&avctx->hw_frames_ctx);
 736             goto fail;
 737         }
 738     }
 739
 740     vtctx->cached_hw_frames_ctx = av_buffer_ref(avctx->hw_frames_ctx);
 741     if (!vtctx->cached_hw_frames_ctx) {
 742         err = AVERROR(ENOMEM);
 743         goto fail;
 744     }
 745
 746     vtctx->vt_ctx->cv_pix_fmt_type =
 747         av_map_videotoolbox_format_from_pixfmt(hw_frames->sw_format);
 748     if (!vtctx->vt_ctx->cv_pix_fmt_type) {
 749         av_log(avctx, AV_LOG_ERROR, "Unknown sw_format.\n");
 750         err = AVERROR(EINVAL);
 751         goto fail;
 752     }
 753
 754     err = videotoolbox_default_init(avctx);
 755     if (err < 0)
 756         goto fail;
 757
 758     return 0;
 759
 760 fail:
 761     videotoolbox_uninit(avctx);
 762     return err;
 763 }
 764
 765 AVHWAccel ff_h263_videotoolbox_hwaccel = {
 766     .name           = "h263_videotoolbox",
 767     .type           = AVMEDIA_TYPE_VIDEO,
 768     .id             = AV_CODEC_ID_H263,
 769     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
 770     .alloc_frame    = ff_videotoolbox_alloc_frame,
 771     .start_frame    = videotoolbox_mpeg_start_frame,
 772     .decode_slice   = videotoolbox_mpeg_decode_slice,
 773     .end_frame      = videotoolbox_mpeg_end_frame,
 774     .init           = videotoolbox_common_init,
 775     .uninit         = videotoolbox_uninit,
 776     .priv_data_size = sizeof(VTContext),
 777 };
 778
 779 AVHWAccel ff_h264_videotoolbox_hwaccel = {
 780     .name           = "h264_videotoolbox",
 781     .type           = AVMEDIA_TYPE_VIDEO,
 782     .id             = AV_CODEC_ID_H264,
 783     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
 784     .alloc_frame    = ff_videotoolbox_alloc_frame,
 785     .start_frame    = ff_videotoolbox_h264_start_frame,
 786     .decode_slice   = ff_videotoolbox_h264_decode_slice,
 787     .end_frame      = videotoolbox_h264_end_frame,
 788     .init           = videotoolbox_common_init,
 789     .uninit         = videotoolbox_uninit,
 790     .priv_data_size = sizeof(VTContext),
 791 };
 792
 793 AVHWAccel ff_mpeg1_videotoolbox_hwaccel = {
 794     .name           = "mpeg1_videotoolbox",
 795     .type           = AVMEDIA_TYPE_VIDEO,
 796     .id             = AV_CODEC_ID_MPEG1VIDEO,
 797     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
 798     .alloc_frame    = ff_videotoolbox_alloc_frame,
 799     .start_frame    = videotoolbox_mpeg_start_frame,
 800     .decode_slice   = videotoolbox_mpeg_decode_slice,
 801     .end_frame      = videotoolbox_mpeg_end_frame,
 802     .init           = videotoolbox_common_init,
 803     .uninit         = videotoolbox_uninit,
 804     .priv_data_size = sizeof(VTContext),
 805 };
 806
 807 AVHWAccel ff_mpeg2_videotoolbox_hwaccel = {
 808     .name           = "mpeg2_videotoolbox",
 809     .type           = AVMEDIA_TYPE_VIDEO,
 810     .id             = AV_CODEC_ID_MPEG2VIDEO,
 811     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
 812     .alloc_frame    = ff_videotoolbox_alloc_frame,
 813     .start_frame    = videotoolbox_mpeg_start_frame,
 814     .decode_slice   = videotoolbox_mpeg_decode_slice,
 815     .end_frame      = videotoolbox_mpeg_end_frame,
 816     .init           = videotoolbox_common_init,
 817     .uninit         = videotoolbox_uninit,
 818     .priv_data_size = sizeof(VTContext),
 819 };
 820
 821 AVHWAccel ff_mpeg4_videotoolbox_hwaccel = {
 822     .name           = "mpeg4_videotoolbox",
 823     .type           = AVMEDIA_TYPE_VIDEO,
 824     .id             = AV_CODEC_ID_MPEG4,
 825     .pix_fmt        = AV_PIX_FMT_VIDEOTOOLBOX,
 826     .alloc_frame    = ff_videotoolbox_alloc_frame,
 827     .start_frame    = videotoolbox_mpeg_start_frame,
 828     .decode_slice   = videotoolbox_mpeg_decode_slice,
 829     .end_frame      = videotoolbox_mpeg_end_frame,
 830     .init           = videotoolbox_common_init,
 831     .uninit         = videotoolbox_uninit,
 832     .priv_data_size = sizeof(VTContext),
 833 };
 834
 835 AVVideotoolboxContext *av_videotoolbox_alloc_context(void)
 836 {
 837     AVVideotoolboxContext *ret = av_mallocz(sizeof(*ret));
 838
 839     if (ret) {
 840         ret->output_callback = videotoolbox_decoder_callback;
 841         ret->cv_pix_fmt_type = kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange;
 842     }
 843
 844     return ret;
 845 }
 846
 847 int av_videotoolbox_default_init(AVCodecContext *avctx)
 848 {
 849     return av_videotoolbox_default_init2(avctx, NULL);
 850 }
 851
 852 int av_videotoolbox_default_init2(AVCodecContext *avctx, AVVideotoolboxContext *vtctx)
 853 {
 854     avctx->hwaccel_context = vtctx ?: av_videotoolbox_alloc_context();
 855     if (!avctx->hwaccel_context)
 856         return AVERROR(ENOMEM);
 857     return videotoolbox_default_init(avctx);
 858 }
 859
 860 void av_videotoolbox_default_free(AVCodecContext *avctx)
 861 {
 862
 863     videotoolbox_default_free(avctx);
 864     av_freep(&avctx->hwaccel_context);
 865 }
 866 #endif /* CONFIG_VIDEOTOOLBOX */