2 * Videotoolbox hardware acceleration
4 * copyright (c) 2012 Sebastien Zwickert
6 * This file is part of FFmpeg.
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.
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.
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
24 #include "videotoolbox.h"
25 #include "libavutil/hwcontext_videotoolbox.h"
26 #include "vt_internal.h"
27 #include "libavutil/avutil.h"
28 #include "libavutil/hwcontext.h"
29 #include "libavutil/pixdesc.h"
30 #include "bytestream.h"
34 #include "mpegvideo.h"
35 #include <TargetConditionals.h>
37 #ifndef kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder
38 # define kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder CFSTR("RequireHardwareAcceleratedVideoDecoder")
40 #ifndef kVTVideoDecoderSpecification_EnableHardwareAcceleratedVideoDecoder
41 # define kVTVideoDecoderSpecification_EnableHardwareAcceleratedVideoDecoder CFSTR("EnableHardwareAcceleratedVideoDecoder")
44 #if !HAVE_KCMVIDEOCODECTYPE_HEVC
45 enum { kCMVideoCodecType_HEVC = 'hvc1' };
48 #define VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING 12
50 typedef struct VTHWFrame {
51 CVPixelBufferRef pixbuf;
52 AVBufferRef *hw_frames_ctx;
55 static void videotoolbox_buffer_release(void *opaque, uint8_t *data)
57 VTHWFrame *ref = (VTHWFrame *)data;
58 av_buffer_unref(&ref->hw_frames_ctx);
59 CVPixelBufferRelease(ref->pixbuf);
64 static int videotoolbox_buffer_copy(VTContext *vtctx,
65 const uint8_t *buffer,
70 tmp = av_fast_realloc(vtctx->bitstream,
71 &vtctx->allocated_size,
75 return AVERROR(ENOMEM);
77 vtctx->bitstream = tmp;
78 memcpy(vtctx->bitstream, buffer, size);
79 vtctx->bitstream_size = size;
84 static int videotoolbox_postproc_frame(void *avctx, AVFrame *frame)
86 VTHWFrame *ref = (VTHWFrame *)frame->buf[0]->data;
89 av_log(avctx, AV_LOG_ERROR, "No frame decoded?\n");
90 av_frame_unref(frame);
91 return AVERROR_EXTERNAL;
94 frame->data[3] = (uint8_t*)ref->pixbuf;
96 if (ref->hw_frames_ctx) {
97 av_buffer_unref(&frame->hw_frames_ctx);
98 frame->hw_frames_ctx = av_buffer_ref(ref->hw_frames_ctx);
99 if (!frame->hw_frames_ctx)
100 return AVERROR(ENOMEM);
106 int ff_videotoolbox_alloc_frame(AVCodecContext *avctx, AVFrame *frame)
108 size_t size = sizeof(VTHWFrame);
109 uint8_t *data = NULL;
110 AVBufferRef *buf = NULL;
111 int ret = ff_attach_decode_data(frame);
112 FrameDecodeData *fdd;
116 data = av_mallocz(size);
118 return AVERROR(ENOMEM);
119 buf = av_buffer_create(data, size, videotoolbox_buffer_release, NULL, 0);
122 return AVERROR(ENOMEM);
126 fdd = (FrameDecodeData*)frame->private_ref->data;
127 fdd->post_process = videotoolbox_postproc_frame;
129 frame->width = avctx->width;
130 frame->height = avctx->height;
131 frame->format = avctx->pix_fmt;
136 #define AV_W8(p, v) *(p) = (v)
138 CFDataRef ff_videotoolbox_avcc_extradata_create(AVCodecContext *avctx)
140 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
141 H264Context *h = avctx->priv_data;
142 CFDataRef data = NULL;
144 int vt_extradata_size = 6 + 2 + h->ps.sps->data_size + 3 + h->ps.pps->data_size;
145 uint8_t *vt_extradata = av_malloc(vt_extradata_size);
151 AV_W8(p + 0, 1); /* version */
152 AV_W8(p + 1, h->ps.sps->data[1]); /* profile */
153 AV_W8(p + 2, h->ps.sps->data[2]); /* profile compat */
154 AV_W8(p + 3, h->ps.sps->data[3]); /* level */
155 AV_W8(p + 4, 0xff); /* 6 bits reserved (111111) + 2 bits nal size length - 3 (11) */
156 AV_W8(p + 5, 0xe1); /* 3 bits reserved (111) + 5 bits number of sps (00001) */
157 AV_WB16(p + 6, h->ps.sps->data_size);
158 memcpy(p + 8, h->ps.sps->data, h->ps.sps->data_size);
159 p += 8 + h->ps.sps->data_size;
160 AV_W8(p + 0, 1); /* number of pps */
161 AV_WB16(p + 1, h->ps.pps->data_size);
162 memcpy(p + 3, h->ps.pps->data, h->ps.pps->data_size);
164 p += 3 + h->ps.pps->data_size;
165 av_assert0(p - vt_extradata == vt_extradata_size);
167 // save sps header (profile/level) used to create decoder session,
168 // so we can detect changes and recreate it.
170 memcpy(vtctx->sps, h->ps.sps->data + 1, 3);
172 data = CFDataCreate(kCFAllocatorDefault, vt_extradata, vt_extradata_size);
173 av_free(vt_extradata);
177 CFDataRef ff_videotoolbox_hvcc_extradata_create(AVCodecContext *avctx)
179 HEVCContext *h = avctx->priv_data;
180 int i, num_vps = 0, num_sps = 0, num_pps = 0;
181 const HEVCVPS *vps = h->ps.vps;
182 const HEVCSPS *sps = h->ps.sps;
183 const HEVCPPS *pps = h->ps.pps;
184 PTLCommon ptlc = vps->ptl.general_ptl;
186 uint8_t parallelismType;
187 CFDataRef data = NULL;
189 int vt_extradata_size = 23 + 3 + 3 + 3;
190 uint8_t *vt_extradata;
192 #define COUNT_SIZE_PS(T, t) \
193 for (i = 0; i < HEVC_MAX_##T##PS_COUNT; i++) { \
194 if (h->ps.t##ps_list[i]) { \
195 const HEVC##T##PS *lps = (const HEVC##T##PS *)h->ps.t##ps_list[i]->data; \
196 vt_extradata_size += 2 + lps->data_size; \
205 vt_extradata = av_malloc(vt_extradata_size);
210 /* unsigned int(8) configurationVersion = 1; */
214 * unsigned int(2) general_profile_space;
215 * unsigned int(1) general_tier_flag;
216 * unsigned int(5) general_profile_idc;
218 AV_W8(p + 1, ptlc.profile_space << 6 |
219 ptlc.tier_flag << 5 |
222 /* unsigned int(32) general_profile_compatibility_flags; */
223 memcpy(p + 2, ptlc.profile_compatibility_flag, 4);
225 /* unsigned int(48) general_constraint_indicator_flags; */
226 AV_W8(p + 6, ptlc.progressive_source_flag << 7 |
227 ptlc.interlaced_source_flag << 6 |
228 ptlc.non_packed_constraint_flag << 5 |
229 ptlc.frame_only_constraint_flag << 4);
233 /* unsigned int(8) general_level_idc; */
234 AV_W8(p + 12, ptlc.level_idc);
237 * bit(4) reserved = ‘1111’b;
238 * unsigned int(12) min_spatial_segmentation_idc;
240 AV_W8(p + 13, 0xf0 | (vui.min_spatial_segmentation_idc >> 4));
241 AV_W8(p + 14, vui.min_spatial_segmentation_idc & 0xff);
244 * bit(6) reserved = ‘111111’b;
245 * unsigned int(2) parallelismType;
247 if (!vui.min_spatial_segmentation_idc)
249 else if (pps->entropy_coding_sync_enabled_flag && pps->tiles_enabled_flag)
251 else if (pps->entropy_coding_sync_enabled_flag)
253 else if (pps->tiles_enabled_flag)
257 AV_W8(p + 15, 0xfc | parallelismType);
260 * bit(6) reserved = ‘111111’b;
261 * unsigned int(2) chromaFormat;
263 AV_W8(p + 16, sps->chroma_format_idc | 0xfc);
266 * bit(5) reserved = ‘11111’b;
267 * unsigned int(3) bitDepthLumaMinus8;
269 AV_W8(p + 17, (sps->bit_depth - 8) | 0xfc);
272 * bit(5) reserved = ‘11111’b;
273 * unsigned int(3) bitDepthChromaMinus8;
275 AV_W8(p + 18, (sps->bit_depth_chroma - 8) | 0xfc);
277 /* bit(16) avgFrameRate; */
281 * bit(2) constantFrameRate;
282 * bit(3) numTemporalLayers;
283 * bit(1) temporalIdNested;
284 * unsigned int(2) lengthSizeMinusOne;
286 AV_W8(p + 21, 0 << 6 |
287 sps->max_sub_layers << 3 |
288 sps->temporal_id_nesting_flag << 2 |
291 /* unsigned int(8) numOfArrays; */
296 #define APPEND_PS(T, t) \
298 * bit(1) array_completeness; \
299 * unsigned int(1) reserved = 0; \
300 * unsigned int(6) NAL_unit_type; \
303 HEVC_NAL_##T##PS & 0x3f); \
304 /* unsigned int(16) numNalus; */ \
305 AV_WB16(p + 1, num_##t##ps); \
307 for (i = 0; i < HEVC_MAX_##T##PS_COUNT; i++) { \
308 if (h->ps.t##ps_list[i]) { \
309 const HEVC##T##PS *lps = (const HEVC##T##PS *)h->ps.t##ps_list[i]->data; \
310 /* unsigned int(16) nalUnitLength; */ \
311 AV_WB16(p, lps->data_size); \
312 /* bit(8*nalUnitLength) nalUnit; */ \
313 memcpy(p + 2, lps->data, lps->data_size); \
314 p += 2 + lps->data_size; \
322 av_assert0(p - vt_extradata == vt_extradata_size);
324 data = CFDataCreate(kCFAllocatorDefault, vt_extradata, vt_extradata_size);
325 av_free(vt_extradata);
329 int ff_videotoolbox_h264_start_frame(AVCodecContext *avctx,
330 const uint8_t *buffer,
333 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
334 H264Context *h = avctx->priv_data;
336 if (h->is_avc == 1) {
337 return videotoolbox_buffer_copy(vtctx, buffer, size);
343 static int videotoolbox_h264_decode_params(AVCodecContext *avctx,
345 const uint8_t *buffer,
348 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
349 H264Context *h = avctx->priv_data;
351 // save sps header (profile/level) used to create decoder session
353 memcpy(vtctx->sps, h->ps.sps->data + 1, 3);
355 if (type == H264_NAL_SPS) {
356 if (size > 4 && memcmp(vtctx->sps, buffer + 1, 3) != 0) {
357 vtctx->reconfig_needed = true;
358 memcpy(vtctx->sps, buffer + 1, 3);
362 // pass-through SPS/PPS changes to the decoder
363 return ff_videotoolbox_h264_decode_slice(avctx, buffer, size);
366 static int videotoolbox_common_decode_slice(AVCodecContext *avctx,
367 const uint8_t *buffer,
370 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
373 tmp = av_fast_realloc(vtctx->bitstream,
374 &vtctx->allocated_size,
375 vtctx->bitstream_size+size+4);
377 return AVERROR(ENOMEM);
379 vtctx->bitstream = tmp;
381 AV_WB32(vtctx->bitstream + vtctx->bitstream_size, size);
382 memcpy(vtctx->bitstream + vtctx->bitstream_size + 4, buffer, size);
384 vtctx->bitstream_size += size + 4;
389 int ff_videotoolbox_h264_decode_slice(AVCodecContext *avctx,
390 const uint8_t *buffer,
393 H264Context *h = avctx->priv_data;
398 return videotoolbox_common_decode_slice(avctx, buffer, size);
401 int ff_videotoolbox_uninit(AVCodecContext *avctx)
403 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
405 av_freep(&vtctx->bitstream);
407 CVPixelBufferRelease(vtctx->frame);
413 #if CONFIG_VIDEOTOOLBOX
414 // Return the AVVideotoolboxContext that matters currently. Where it comes from
415 // depends on the API used.
416 static AVVideotoolboxContext *videotoolbox_get_context(AVCodecContext *avctx)
418 // Somewhat tricky because the user can call av_videotoolbox_default_free()
419 // at any time, even when the codec is closed.
420 if (avctx->internal && avctx->internal->hwaccel_priv_data) {
421 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
423 return vtctx->vt_ctx;
425 return avctx->hwaccel_context;
428 static int videotoolbox_buffer_create(AVCodecContext *avctx, AVFrame *frame)
430 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
431 CVPixelBufferRef pixbuf = (CVPixelBufferRef)vtctx->frame;
432 OSType pixel_format = CVPixelBufferGetPixelFormatType(pixbuf);
433 enum AVPixelFormat sw_format = av_map_videotoolbox_format_to_pixfmt(pixel_format);
434 int width = CVPixelBufferGetWidth(pixbuf);
435 int height = CVPixelBufferGetHeight(pixbuf);
436 AVHWFramesContext *cached_frames;
440 if (!frame->buf[0] || frame->data[3]) {
441 av_log(avctx, AV_LOG_ERROR, "videotoolbox: invalid state\n");
442 av_frame_unref(frame);
443 return AVERROR_EXTERNAL;
446 ref = (VTHWFrame *)frame->buf[0]->data;
449 CVPixelBufferRelease(ref->pixbuf);
450 ref->pixbuf = vtctx->frame;
453 // Old API code path.
454 if (!vtctx->cached_hw_frames_ctx)
457 cached_frames = (AVHWFramesContext*)vtctx->cached_hw_frames_ctx->data;
459 if (cached_frames->sw_format != sw_format ||
460 cached_frames->width != width ||
461 cached_frames->height != height) {
462 AVBufferRef *hw_frames_ctx = av_hwframe_ctx_alloc(cached_frames->device_ref);
463 AVHWFramesContext *hw_frames;
465 return AVERROR(ENOMEM);
467 hw_frames = (AVHWFramesContext*)hw_frames_ctx->data;
468 hw_frames->format = cached_frames->format;
469 hw_frames->sw_format = sw_format;
470 hw_frames->width = width;
471 hw_frames->height = height;
473 ret = av_hwframe_ctx_init(hw_frames_ctx);
475 av_buffer_unref(&hw_frames_ctx);
479 av_buffer_unref(&vtctx->cached_hw_frames_ctx);
480 vtctx->cached_hw_frames_ctx = hw_frames_ctx;
483 av_buffer_unref(&ref->hw_frames_ctx);
484 ref->hw_frames_ctx = av_buffer_ref(vtctx->cached_hw_frames_ctx);
485 if (!ref->hw_frames_ctx)
486 return AVERROR(ENOMEM);
491 static void videotoolbox_write_mp4_descr_length(PutByteContext *pb, int length)
496 for (i = 3; i >= 0; i--) {
497 b = (length >> (i * 7)) & 0x7F;
501 bytestream2_put_byteu(pb, b);
505 static CFDataRef videotoolbox_esds_extradata_create(AVCodecContext *avctx)
508 uint8_t *rw_extradata;
510 int full_size = 3 + 5 + 13 + 5 + avctx->extradata_size + 3;
511 // ES_DescrTag data + DecoderConfigDescrTag + data + DecSpecificInfoTag + size + SLConfigDescriptor
512 int config_size = 13 + 5 + avctx->extradata_size;
515 if (!(rw_extradata = av_mallocz(full_size + VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING)))
518 bytestream2_init_writer(&pb, rw_extradata, full_size + VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING);
519 bytestream2_put_byteu(&pb, 0); // version
520 bytestream2_put_ne24(&pb, 0); // flags
522 // elementary stream descriptor
523 bytestream2_put_byteu(&pb, 0x03); // ES_DescrTag
524 videotoolbox_write_mp4_descr_length(&pb, full_size);
525 bytestream2_put_ne16(&pb, 0); // esid
526 bytestream2_put_byteu(&pb, 0); // stream priority (0-32)
528 // decoder configuration descriptor
529 bytestream2_put_byteu(&pb, 0x04); // DecoderConfigDescrTag
530 videotoolbox_write_mp4_descr_length(&pb, config_size);
531 bytestream2_put_byteu(&pb, 32); // object type indication. 32 = AV_CODEC_ID_MPEG4
532 bytestream2_put_byteu(&pb, 0x11); // stream type
533 bytestream2_put_ne24(&pb, 0); // buffer size
534 bytestream2_put_ne32(&pb, 0); // max bitrate
535 bytestream2_put_ne32(&pb, 0); // avg bitrate
537 // decoder specific descriptor
538 bytestream2_put_byteu(&pb, 0x05); ///< DecSpecificInfoTag
539 videotoolbox_write_mp4_descr_length(&pb, avctx->extradata_size);
541 bytestream2_put_buffer(&pb, avctx->extradata, avctx->extradata_size);
543 // SLConfigDescriptor
544 bytestream2_put_byteu(&pb, 0x06); // SLConfigDescrTag
545 bytestream2_put_byteu(&pb, 0x01); // length
546 bytestream2_put_byteu(&pb, 0x02); //
548 s = bytestream2_size_p(&pb);
550 data = CFDataCreate(kCFAllocatorDefault, rw_extradata, s);
552 av_freep(&rw_extradata);
556 static CMSampleBufferRef videotoolbox_sample_buffer_create(CMFormatDescriptionRef fmt_desc,
561 CMBlockBufferRef block_buf;
562 CMSampleBufferRef sample_buf;
567 status = CMBlockBufferCreateWithMemoryBlock(kCFAllocatorDefault,// structureAllocator
568 buffer, // memoryBlock
570 kCFAllocatorNull, // blockAllocator
571 NULL, // customBlockSource
578 status = CMSampleBufferCreate(kCFAllocatorDefault, // allocator
579 block_buf, // dataBuffer
581 0, // makeDataReadyCallback
582 0, // makeDataReadyRefcon
583 fmt_desc, // formatDescription
585 0, // numSampleTimingEntries
586 NULL, // sampleTimingArray
587 0, // numSampleSizeEntries
588 NULL, // sampleSizeArray
593 CFRelease(block_buf);
598 static void videotoolbox_decoder_callback(void *opaque,
599 void *sourceFrameRefCon,
601 VTDecodeInfoFlags flags,
602 CVImageBufferRef image_buffer,
606 AVCodecContext *avctx = opaque;
607 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
610 CVPixelBufferRelease(vtctx->frame);
615 av_log(NULL, AV_LOG_DEBUG, "vt decoder cb: output image buffer is null\n");
619 vtctx->frame = CVPixelBufferRetain(image_buffer);
622 static OSStatus videotoolbox_session_decode_frame(AVCodecContext *avctx)
625 CMSampleBufferRef sample_buf;
626 AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
627 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
629 sample_buf = videotoolbox_sample_buffer_create(videotoolbox->cm_fmt_desc,
631 vtctx->bitstream_size);
636 status = VTDecompressionSessionDecodeFrame(videotoolbox->session,
639 NULL, // sourceFrameRefCon
642 status = VTDecompressionSessionWaitForAsynchronousFrames(videotoolbox->session);
644 CFRelease(sample_buf);
649 static CMVideoFormatDescriptionRef videotoolbox_format_desc_create(CMVideoCodecType codec_type,
650 CFDictionaryRef decoder_spec,
654 CMFormatDescriptionRef cm_fmt_desc;
657 status = CMVideoFormatDescriptionCreate(kCFAllocatorDefault,
661 decoder_spec, // Dictionary of extension
670 static CFDictionaryRef videotoolbox_buffer_attributes_create(int width,
674 CFMutableDictionaryRef buffer_attributes;
675 CFMutableDictionaryRef io_surface_properties;
676 CFNumberRef cv_pix_fmt;
680 w = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &width);
681 h = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &height);
682 cv_pix_fmt = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &pix_fmt);
684 buffer_attributes = CFDictionaryCreateMutable(kCFAllocatorDefault,
686 &kCFTypeDictionaryKeyCallBacks,
687 &kCFTypeDictionaryValueCallBacks);
688 io_surface_properties = CFDictionaryCreateMutable(kCFAllocatorDefault,
690 &kCFTypeDictionaryKeyCallBacks,
691 &kCFTypeDictionaryValueCallBacks);
694 CFDictionarySetValue(buffer_attributes, kCVPixelBufferPixelFormatTypeKey, cv_pix_fmt);
695 CFDictionarySetValue(buffer_attributes, kCVPixelBufferIOSurfacePropertiesKey, io_surface_properties);
696 CFDictionarySetValue(buffer_attributes, kCVPixelBufferWidthKey, w);
697 CFDictionarySetValue(buffer_attributes, kCVPixelBufferHeightKey, h);
699 CFDictionarySetValue(buffer_attributes, kCVPixelBufferOpenGLESCompatibilityKey, kCFBooleanTrue);
701 CFDictionarySetValue(buffer_attributes, kCVPixelBufferIOSurfaceOpenGLTextureCompatibilityKey, kCFBooleanTrue);
704 CFRelease(io_surface_properties);
705 CFRelease(cv_pix_fmt);
709 return buffer_attributes;
712 static CFDictionaryRef videotoolbox_decoder_config_create(CMVideoCodecType codec_type,
713 AVCodecContext *avctx)
715 CFMutableDictionaryRef config_info = CFDictionaryCreateMutable(kCFAllocatorDefault,
717 &kCFTypeDictionaryKeyCallBacks,
718 &kCFTypeDictionaryValueCallBacks);
720 CFDictionarySetValue(config_info,
721 codec_type == kCMVideoCodecType_HEVC ?
722 kVTVideoDecoderSpecification_EnableHardwareAcceleratedVideoDecoder :
723 kVTVideoDecoderSpecification_RequireHardwareAcceleratedVideoDecoder,
726 CFMutableDictionaryRef avc_info;
727 CFDataRef data = NULL;
729 avc_info = CFDictionaryCreateMutable(kCFAllocatorDefault,
731 &kCFTypeDictionaryKeyCallBacks,
732 &kCFTypeDictionaryValueCallBacks);
734 switch (codec_type) {
735 case kCMVideoCodecType_MPEG4Video :
736 if (avctx->extradata_size)
737 data = videotoolbox_esds_extradata_create(avctx);
739 CFDictionarySetValue(avc_info, CFSTR("esds"), data);
741 case kCMVideoCodecType_H264 :
742 data = ff_videotoolbox_avcc_extradata_create(avctx);
744 CFDictionarySetValue(avc_info, CFSTR("avcC"), data);
746 case kCMVideoCodecType_HEVC :
747 data = ff_videotoolbox_hvcc_extradata_create(avctx);
749 CFDictionarySetValue(avc_info, CFSTR("hvcC"), data);
755 CFDictionarySetValue(config_info,
756 kCMFormatDescriptionExtension_SampleDescriptionExtensionAtoms,
766 static int videotoolbox_start(AVCodecContext *avctx)
768 AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
770 VTDecompressionOutputCallbackRecord decoder_cb;
771 CFDictionaryRef decoder_spec;
772 CFDictionaryRef buf_attr;
775 av_log(avctx, AV_LOG_ERROR, "hwaccel context is not set\n");
779 switch( avctx->codec_id ) {
780 case AV_CODEC_ID_H263 :
781 videotoolbox->cm_codec_type = kCMVideoCodecType_H263;
783 case AV_CODEC_ID_H264 :
784 videotoolbox->cm_codec_type = kCMVideoCodecType_H264;
786 case AV_CODEC_ID_HEVC :
787 videotoolbox->cm_codec_type = kCMVideoCodecType_HEVC;
789 case AV_CODEC_ID_MPEG1VIDEO :
790 videotoolbox->cm_codec_type = kCMVideoCodecType_MPEG1Video;
792 case AV_CODEC_ID_MPEG2VIDEO :
793 videotoolbox->cm_codec_type = kCMVideoCodecType_MPEG2Video;
795 case AV_CODEC_ID_MPEG4 :
796 videotoolbox->cm_codec_type = kCMVideoCodecType_MPEG4Video;
802 decoder_spec = videotoolbox_decoder_config_create(videotoolbox->cm_codec_type, avctx);
805 av_log(avctx, AV_LOG_ERROR, "decoder specification creation failed\n");
809 videotoolbox->cm_fmt_desc = videotoolbox_format_desc_create(videotoolbox->cm_codec_type,
813 if (!videotoolbox->cm_fmt_desc) {
815 CFRelease(decoder_spec);
817 av_log(avctx, AV_LOG_ERROR, "format description creation failed\n");
821 buf_attr = videotoolbox_buffer_attributes_create(avctx->width,
823 videotoolbox->cv_pix_fmt_type);
825 decoder_cb.decompressionOutputCallback = videotoolbox_decoder_callback;
826 decoder_cb.decompressionOutputRefCon = avctx;
828 status = VTDecompressionSessionCreate(NULL, // allocator
829 videotoolbox->cm_fmt_desc, // videoFormatDescription
830 decoder_spec, // videoDecoderSpecification
831 buf_attr, // destinationImageBufferAttributes
832 &decoder_cb, // outputCallback
833 &videotoolbox->session); // decompressionSessionOut
836 CFRelease(decoder_spec);
841 case kVTVideoDecoderNotAvailableNowErr:
842 av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox session not available.\n");
843 return AVERROR(ENOSYS);
844 case kVTVideoDecoderUnsupportedDataFormatErr:
845 av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox does not support this format.\n");
846 return AVERROR(ENOSYS);
847 case kVTCouldNotFindVideoDecoderErr:
848 av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox decoder for this format not found.\n");
849 return AVERROR(ENOSYS);
850 case kVTVideoDecoderMalfunctionErr:
851 av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox malfunction.\n");
852 return AVERROR(EINVAL);
853 case kVTVideoDecoderBadDataErr:
854 av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox reported invalid data.\n");
855 return AVERROR_INVALIDDATA;
859 av_log(avctx, AV_LOG_VERBOSE, "Unknown VideoToolbox session creation error %d\n", (int)status);
860 return AVERROR_UNKNOWN;
864 static void videotoolbox_stop(AVCodecContext *avctx)
866 AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
870 if (videotoolbox->cm_fmt_desc) {
871 CFRelease(videotoolbox->cm_fmt_desc);
872 videotoolbox->cm_fmt_desc = NULL;
875 if (videotoolbox->session) {
876 VTDecompressionSessionInvalidate(videotoolbox->session);
877 CFRelease(videotoolbox->session);
878 videotoolbox->session = NULL;
882 static const char *videotoolbox_error_string(OSStatus status)
885 case kVTVideoDecoderBadDataErr:
887 case kVTVideoDecoderMalfunctionErr:
888 return "decoder malfunction";
889 case kVTInvalidSessionErr:
890 return "invalid session";
895 static int videotoolbox_common_end_frame(AVCodecContext *avctx, AVFrame *frame)
898 AVVideotoolboxContext *videotoolbox = videotoolbox_get_context(avctx);
899 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
901 frame->crop_right = 0;
902 frame->crop_left = 0;
904 frame->crop_bottom = 0;
906 if (vtctx->reconfig_needed == true) {
907 vtctx->reconfig_needed = false;
908 av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox decoder needs reconfig, restarting..\n");
909 videotoolbox_stop(avctx);
910 if (videotoolbox_start(avctx) != 0) {
911 return AVERROR_EXTERNAL;
915 if (!videotoolbox->session || !vtctx->bitstream || !vtctx->bitstream_size)
916 return AVERROR_INVALIDDATA;
918 status = videotoolbox_session_decode_frame(avctx);
919 if (status != noErr) {
920 if (status == kVTVideoDecoderMalfunctionErr || status == kVTInvalidSessionErr)
921 vtctx->reconfig_needed = true;
922 av_log(avctx, AV_LOG_ERROR, "Failed to decode frame (%s, %d)\n", videotoolbox_error_string(status), (int)status);
923 return AVERROR_UNKNOWN;
927 vtctx->reconfig_needed = true;
928 return AVERROR_UNKNOWN;
931 return videotoolbox_buffer_create(avctx, frame);
934 static int videotoolbox_h264_end_frame(AVCodecContext *avctx)
936 H264Context *h = avctx->priv_data;
937 AVFrame *frame = h->cur_pic_ptr->f;
938 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
939 int ret = videotoolbox_common_end_frame(avctx, frame);
940 vtctx->bitstream_size = 0;
944 static int videotoolbox_hevc_start_frame(AVCodecContext *avctx,
945 const uint8_t *buffer,
951 static int videotoolbox_hevc_decode_slice(AVCodecContext *avctx,
952 const uint8_t *buffer,
955 return videotoolbox_common_decode_slice(avctx, buffer, size);
959 static int videotoolbox_hevc_decode_params(AVCodecContext *avctx,
961 const uint8_t *buffer,
964 return videotoolbox_common_decode_slice(avctx, buffer, size);
967 static int videotoolbox_hevc_end_frame(AVCodecContext *avctx)
969 HEVCContext *h = avctx->priv_data;
970 AVFrame *frame = h->ref->frame;
971 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
973 h->output_frame->crop_right = 0;
974 h->output_frame->crop_left = 0;
975 h->output_frame->crop_top = 0;
976 h->output_frame->crop_bottom = 0;
978 int ret = videotoolbox_common_end_frame(avctx, frame);
979 vtctx->bitstream_size = 0;
983 static int videotoolbox_mpeg_start_frame(AVCodecContext *avctx,
984 const uint8_t *buffer,
987 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
989 return videotoolbox_buffer_copy(vtctx, buffer, size);
992 static int videotoolbox_mpeg_decode_slice(AVCodecContext *avctx,
993 const uint8_t *buffer,
999 static int videotoolbox_mpeg_end_frame(AVCodecContext *avctx)
1001 MpegEncContext *s = avctx->priv_data;
1002 AVFrame *frame = s->current_picture_ptr->f;
1004 return videotoolbox_common_end_frame(avctx, frame);
1007 static int videotoolbox_uninit(AVCodecContext *avctx)
1009 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
1013 ff_videotoolbox_uninit(avctx);
1016 videotoolbox_stop(avctx);
1018 av_buffer_unref(&vtctx->cached_hw_frames_ctx);
1019 av_freep(&vtctx->vt_ctx);
1024 static enum AVPixelFormat videotoolbox_best_pixel_format(AVCodecContext *avctx) {
1025 const AVPixFmtDescriptor *descriptor = av_pix_fmt_desc_get(avctx->pix_fmt);
1027 return AV_PIX_FMT_NV12; // same as av_videotoolbox_alloc_context()
1029 int depth = descriptor->comp[0].depth;
1031 return AV_PIX_FMT_P010;
1034 return AV_PIX_FMT_NV12;
1037 static int videotoolbox_common_init(AVCodecContext *avctx)
1039 VTContext *vtctx = avctx->internal->hwaccel_priv_data;
1040 AVHWFramesContext *hw_frames;
1043 // Old API - do nothing.
1044 if (avctx->hwaccel_context)
1047 if (!avctx->hw_frames_ctx && !avctx->hw_device_ctx) {
1048 av_log(avctx, AV_LOG_ERROR,
1049 "Either hw_frames_ctx or hw_device_ctx must be set.\n");
1050 return AVERROR(EINVAL);
1053 vtctx->vt_ctx = av_videotoolbox_alloc_context();
1054 if (!vtctx->vt_ctx) {
1055 err = AVERROR(ENOMEM);
1059 if (avctx->hw_frames_ctx) {
1060 hw_frames = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
1062 avctx->hw_frames_ctx = av_hwframe_ctx_alloc(avctx->hw_device_ctx);
1063 if (!avctx->hw_frames_ctx) {
1064 err = AVERROR(ENOMEM);
1068 hw_frames = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
1069 hw_frames->format = AV_PIX_FMT_VIDEOTOOLBOX;
1070 hw_frames->sw_format = videotoolbox_best_pixel_format(avctx);
1071 hw_frames->width = avctx->width;
1072 hw_frames->height = avctx->height;
1074 err = av_hwframe_ctx_init(avctx->hw_frames_ctx);
1076 av_buffer_unref(&avctx->hw_frames_ctx);
1081 vtctx->cached_hw_frames_ctx = av_buffer_ref(avctx->hw_frames_ctx);
1082 if (!vtctx->cached_hw_frames_ctx) {
1083 err = AVERROR(ENOMEM);
1087 vtctx->vt_ctx->cv_pix_fmt_type =
1088 av_map_videotoolbox_format_from_pixfmt(hw_frames->sw_format);
1089 if (!vtctx->vt_ctx->cv_pix_fmt_type) {
1090 av_log(avctx, AV_LOG_ERROR, "Unknown sw_format.\n");
1091 err = AVERROR(EINVAL);
1095 err = videotoolbox_start(avctx);
1102 videotoolbox_uninit(avctx);
1106 static int videotoolbox_frame_params(AVCodecContext *avctx,
1107 AVBufferRef *hw_frames_ctx)
1109 AVHWFramesContext *frames_ctx = (AVHWFramesContext*)hw_frames_ctx->data;
1111 frames_ctx->format = AV_PIX_FMT_VIDEOTOOLBOX;
1112 frames_ctx->width = avctx->coded_width;
1113 frames_ctx->height = avctx->coded_height;
1114 frames_ctx->sw_format = videotoolbox_best_pixel_format(avctx);
1119 const AVHWAccel ff_h263_videotoolbox_hwaccel = {
1120 .name = "h263_videotoolbox",
1121 .type = AVMEDIA_TYPE_VIDEO,
1122 .id = AV_CODEC_ID_H263,
1123 .pix_fmt = AV_PIX_FMT_VIDEOTOOLBOX,
1124 .alloc_frame = ff_videotoolbox_alloc_frame,
1125 .start_frame = videotoolbox_mpeg_start_frame,
1126 .decode_slice = videotoolbox_mpeg_decode_slice,
1127 .end_frame = videotoolbox_mpeg_end_frame,
1128 .frame_params = videotoolbox_frame_params,
1129 .init = videotoolbox_common_init,
1130 .uninit = videotoolbox_uninit,
1131 .priv_data_size = sizeof(VTContext),
1134 const AVHWAccel ff_hevc_videotoolbox_hwaccel = {
1135 .name = "hevc_videotoolbox",
1136 .type = AVMEDIA_TYPE_VIDEO,
1137 .id = AV_CODEC_ID_HEVC,
1138 .pix_fmt = AV_PIX_FMT_VIDEOTOOLBOX,
1139 .alloc_frame = ff_videotoolbox_alloc_frame,
1140 .start_frame = videotoolbox_hevc_start_frame,
1141 .decode_slice = videotoolbox_hevc_decode_slice,
1142 .decode_params = videotoolbox_hevc_decode_params,
1143 .end_frame = videotoolbox_hevc_end_frame,
1144 .frame_params = videotoolbox_frame_params,
1145 .init = videotoolbox_common_init,
1146 .uninit = ff_videotoolbox_uninit,
1147 .priv_data_size = sizeof(VTContext),
1150 const AVHWAccel ff_h264_videotoolbox_hwaccel = {
1151 .name = "h264_videotoolbox",
1152 .type = AVMEDIA_TYPE_VIDEO,
1153 .id = AV_CODEC_ID_H264,
1154 .pix_fmt = AV_PIX_FMT_VIDEOTOOLBOX,
1155 .alloc_frame = ff_videotoolbox_alloc_frame,
1156 .start_frame = ff_videotoolbox_h264_start_frame,
1157 .decode_slice = ff_videotoolbox_h264_decode_slice,
1158 .decode_params = videotoolbox_h264_decode_params,
1159 .end_frame = videotoolbox_h264_end_frame,
1160 .frame_params = videotoolbox_frame_params,
1161 .init = videotoolbox_common_init,
1162 .uninit = videotoolbox_uninit,
1163 .priv_data_size = sizeof(VTContext),
1166 const AVHWAccel ff_mpeg1_videotoolbox_hwaccel = {
1167 .name = "mpeg1_videotoolbox",
1168 .type = AVMEDIA_TYPE_VIDEO,
1169 .id = AV_CODEC_ID_MPEG1VIDEO,
1170 .pix_fmt = AV_PIX_FMT_VIDEOTOOLBOX,
1171 .alloc_frame = ff_videotoolbox_alloc_frame,
1172 .start_frame = videotoolbox_mpeg_start_frame,
1173 .decode_slice = videotoolbox_mpeg_decode_slice,
1174 .end_frame = videotoolbox_mpeg_end_frame,
1175 .frame_params = videotoolbox_frame_params,
1176 .init = videotoolbox_common_init,
1177 .uninit = videotoolbox_uninit,
1178 .priv_data_size = sizeof(VTContext),
1181 const AVHWAccel ff_mpeg2_videotoolbox_hwaccel = {
1182 .name = "mpeg2_videotoolbox",
1183 .type = AVMEDIA_TYPE_VIDEO,
1184 .id = AV_CODEC_ID_MPEG2VIDEO,
1185 .pix_fmt = AV_PIX_FMT_VIDEOTOOLBOX,
1186 .alloc_frame = ff_videotoolbox_alloc_frame,
1187 .start_frame = videotoolbox_mpeg_start_frame,
1188 .decode_slice = videotoolbox_mpeg_decode_slice,
1189 .end_frame = videotoolbox_mpeg_end_frame,
1190 .frame_params = videotoolbox_frame_params,
1191 .init = videotoolbox_common_init,
1192 .uninit = videotoolbox_uninit,
1193 .priv_data_size = sizeof(VTContext),
1196 const AVHWAccel ff_mpeg4_videotoolbox_hwaccel = {
1197 .name = "mpeg4_videotoolbox",
1198 .type = AVMEDIA_TYPE_VIDEO,
1199 .id = AV_CODEC_ID_MPEG4,
1200 .pix_fmt = AV_PIX_FMT_VIDEOTOOLBOX,
1201 .alloc_frame = ff_videotoolbox_alloc_frame,
1202 .start_frame = videotoolbox_mpeg_start_frame,
1203 .decode_slice = videotoolbox_mpeg_decode_slice,
1204 .end_frame = videotoolbox_mpeg_end_frame,
1205 .frame_params = videotoolbox_frame_params,
1206 .init = videotoolbox_common_init,
1207 .uninit = videotoolbox_uninit,
1208 .priv_data_size = sizeof(VTContext),
1211 static AVVideotoolboxContext *av_videotoolbox_alloc_context_with_pix_fmt(enum AVPixelFormat pix_fmt)
1213 AVVideotoolboxContext *ret = av_mallocz(sizeof(*ret));
1216 ret->output_callback = videotoolbox_decoder_callback;
1218 OSType cv_pix_fmt_type = av_map_videotoolbox_format_from_pixfmt(pix_fmt);
1219 if (cv_pix_fmt_type == 0) {
1220 cv_pix_fmt_type = kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange;
1222 ret->cv_pix_fmt_type = cv_pix_fmt_type;
1228 AVVideotoolboxContext *av_videotoolbox_alloc_context(void)
1230 return av_videotoolbox_alloc_context_with_pix_fmt(AV_PIX_FMT_NONE);
1233 int av_videotoolbox_default_init(AVCodecContext *avctx)
1235 return av_videotoolbox_default_init2(avctx, NULL);
1238 int av_videotoolbox_default_init2(AVCodecContext *avctx, AVVideotoolboxContext *vtctx)
1240 avctx->hwaccel_context = vtctx ?: av_videotoolbox_alloc_context_with_pix_fmt(videotoolbox_best_pixel_format(avctx));
1241 if (!avctx->hwaccel_context)
1242 return AVERROR(ENOMEM);
1243 return videotoolbox_start(avctx);
1246 void av_videotoolbox_default_free(AVCodecContext *avctx)
1249 videotoolbox_stop(avctx);
1250 av_freep(&avctx->hwaccel_context);
1252 #endif /* CONFIG_VIDEOTOOLBOX */