2 * copyright (c) 2015 Rick Kern <kernrj@gmail.com>
4 * This file is part of FFmpeg.
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.
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.
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
21 #include <VideoToolbox/VideoToolbox.h>
22 #include <CoreVideo/CoreVideo.h>
23 #include <CoreMedia/CoreMedia.h>
24 #include <TargetConditionals.h>
25 #include <Availability.h>
27 #include "libavutil/opt.h"
28 #include "libavutil/avassert.h"
29 #include "libavutil/atomic.h"
30 #include "libavutil/avstring.h"
31 #include "libavcodec/avcodec.h"
32 #include "libavutil/pixdesc.h"
39 //These symbols may not be present
41 CFStringRef kCVImageBufferColorPrimaries_ITU_R_2020;
42 CFStringRef kCVImageBufferTransferFunction_ITU_R_2020;
43 CFStringRef kCVImageBufferYCbCrMatrix_ITU_R_2020;
45 CFStringRef kVTCompressionPropertyKey_H264EntropyMode;
46 CFStringRef kVTH264EntropyMode_CAVLC;
47 CFStringRef kVTH264EntropyMode_CABAC;
49 CFStringRef kVTProfileLevel_H264_Baseline_4_0;
50 CFStringRef kVTProfileLevel_H264_Baseline_4_2;
51 CFStringRef kVTProfileLevel_H264_Baseline_5_0;
52 CFStringRef kVTProfileLevel_H264_Baseline_5_1;
53 CFStringRef kVTProfileLevel_H264_Baseline_5_2;
54 CFStringRef kVTProfileLevel_H264_Baseline_AutoLevel;
55 CFStringRef kVTProfileLevel_H264_Main_4_2;
56 CFStringRef kVTProfileLevel_H264_Main_5_1;
57 CFStringRef kVTProfileLevel_H264_Main_5_2;
58 CFStringRef kVTProfileLevel_H264_Main_AutoLevel;
59 CFStringRef kVTProfileLevel_H264_High_3_0;
60 CFStringRef kVTProfileLevel_H264_High_3_1;
61 CFStringRef kVTProfileLevel_H264_High_3_2;
62 CFStringRef kVTProfileLevel_H264_High_4_0;
63 CFStringRef kVTProfileLevel_H264_High_4_1;
64 CFStringRef kVTProfileLevel_H264_High_4_2;
65 CFStringRef kVTProfileLevel_H264_High_5_1;
66 CFStringRef kVTProfileLevel_H264_High_5_2;
67 CFStringRef kVTProfileLevel_H264_High_AutoLevel;
69 CFStringRef kVTCompressionPropertyKey_RealTime;
71 CFStringRef kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder;
72 CFStringRef kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder;
75 #define GET_SYM(symbol, defaultVal) \
77 CFStringRef cfstr = *(CFStringRef*)dlsym(RTLD_DEFAULT, #symbol); \
79 compat_keys.symbol = CFSTR(defaultVal); \
81 compat_keys.symbol = cfstr; \
84 static pthread_once_t once_ctrl = PTHREAD_ONCE_INIT;
86 static void loadVTEncSymbols(){
87 GET_SYM(kCVImageBufferColorPrimaries_ITU_R_2020, "ITU_R_2020");
88 GET_SYM(kCVImageBufferTransferFunction_ITU_R_2020, "ITU_R_2020");
89 GET_SYM(kCVImageBufferYCbCrMatrix_ITU_R_2020, "ITU_R_2020");
91 GET_SYM(kVTCompressionPropertyKey_H264EntropyMode, "H264EntropyMode");
92 GET_SYM(kVTH264EntropyMode_CAVLC, "CAVLC");
93 GET_SYM(kVTH264EntropyMode_CABAC, "CABAC");
95 GET_SYM(kVTProfileLevel_H264_Baseline_4_0, "H264_Baseline_4_0");
96 GET_SYM(kVTProfileLevel_H264_Baseline_4_2, "H264_Baseline_4_2");
97 GET_SYM(kVTProfileLevel_H264_Baseline_5_0, "H264_Baseline_5_0");
98 GET_SYM(kVTProfileLevel_H264_Baseline_5_1, "H264_Baseline_5_1");
99 GET_SYM(kVTProfileLevel_H264_Baseline_5_2, "H264_Baseline_5_2");
100 GET_SYM(kVTProfileLevel_H264_Baseline_AutoLevel, "H264_Baseline_AutoLevel");
101 GET_SYM(kVTProfileLevel_H264_Main_4_2, "H264_Main_4_2");
102 GET_SYM(kVTProfileLevel_H264_Main_5_1, "H264_Main_5_1");
103 GET_SYM(kVTProfileLevel_H264_Main_5_2, "H264_Main_5_2");
104 GET_SYM(kVTProfileLevel_H264_Main_AutoLevel, "H264_Main_AutoLevel");
105 GET_SYM(kVTProfileLevel_H264_High_3_0, "H264_High_3_0");
106 GET_SYM(kVTProfileLevel_H264_High_3_1, "H264_High_3_1");
107 GET_SYM(kVTProfileLevel_H264_High_3_2, "H264_High_3_2");
108 GET_SYM(kVTProfileLevel_H264_High_4_0, "H264_High_4_0");
109 GET_SYM(kVTProfileLevel_H264_High_4_1, "H264_High_4_1");
110 GET_SYM(kVTProfileLevel_H264_High_4_2, "H264_High_4_2");
111 GET_SYM(kVTProfileLevel_H264_High_5_1, "H264_High_5_1");
112 GET_SYM(kVTProfileLevel_H264_High_5_2, "H264_High_5_2");
113 GET_SYM(kVTProfileLevel_H264_High_AutoLevel, "H264_High_AutoLevel");
115 GET_SYM(kVTCompressionPropertyKey_RealTime, "RealTime");
117 GET_SYM(kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder,
118 "EnableHardwareAcceleratedVideoEncoder");
119 GET_SYM(kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder,
120 "RequireHardwareAcceleratedVideoEncoder");
123 typedef enum VT_H264Profile {
131 typedef enum VTH264Entropy{
137 static const uint8_t start_code[] = { 0, 0, 0, 1 };
139 typedef struct ExtraSEI {
144 typedef struct BufNode {
145 CMSampleBufferRef cm_buffer;
147 struct BufNode* next;
151 typedef struct VTEncContext {
153 VTCompressionSessionRef session;
154 CFStringRef ycbcr_matrix;
155 CFStringRef color_primaries;
156 CFStringRef transfer_function;
158 pthread_mutex_t lock;
159 pthread_cond_t cv_sample_sent;
166 int64_t frame_ct_out;
176 int64_t frames_before;
177 int64_t frames_after;
183 bool warned_color_range;
187 static int vtenc_populate_extradata(AVCodecContext *avctx,
188 CMVideoCodecType codec_type,
189 CFStringRef profile_level,
190 CFNumberRef gamma_level,
191 CFDictionaryRef enc_info,
192 CFDictionaryRef pixel_buffer_info);
195 * NULL-safe release of *refPtr, and sets value to NULL.
197 static void vt_release_num(CFNumberRef* refPtr){
206 static void set_async_error(VTEncContext *vtctx, int err)
210 pthread_mutex_lock(&vtctx->lock);
212 vtctx->async_error = err;
214 info = vtctx->q_head;
215 vtctx->q_head = vtctx->q_tail = NULL;
218 BufNode *next = info->next;
219 CFRelease(info->cm_buffer);
224 pthread_mutex_unlock(&vtctx->lock);
227 static void clear_frame_queue(VTEncContext *vtctx)
229 set_async_error(vtctx, 0);
232 static int vtenc_q_pop(VTEncContext *vtctx, bool wait, CMSampleBufferRef *buf, ExtraSEI **sei)
236 pthread_mutex_lock(&vtctx->lock);
238 if (vtctx->async_error) {
239 pthread_mutex_unlock(&vtctx->lock);
240 return vtctx->async_error;
243 if (vtctx->flushing && vtctx->frame_ct_in == vtctx->frame_ct_out) {
246 pthread_mutex_unlock(&vtctx->lock);
250 while (!vtctx->q_head && !vtctx->async_error && wait) {
251 pthread_cond_wait(&vtctx->cv_sample_sent, &vtctx->lock);
254 if (!vtctx->q_head) {
255 pthread_mutex_unlock(&vtctx->lock);
260 info = vtctx->q_head;
261 vtctx->q_head = vtctx->q_head->next;
262 if (!vtctx->q_head) {
263 vtctx->q_tail = NULL;
266 pthread_mutex_unlock(&vtctx->lock);
268 *buf = info->cm_buffer;
271 } else if (info->sei) {
272 if (info->sei->data) av_free(info->sei->data);
277 vtctx->frame_ct_out++;
282 static void vtenc_q_push(VTEncContext *vtctx, CMSampleBufferRef buffer, ExtraSEI *sei)
284 BufNode *info = av_malloc(sizeof(BufNode));
286 set_async_error(vtctx, AVERROR(ENOMEM));
291 info->cm_buffer = buffer;
295 pthread_mutex_lock(&vtctx->lock);
296 pthread_cond_signal(&vtctx->cv_sample_sent);
298 if (!vtctx->q_head) {
299 vtctx->q_head = info;
301 vtctx->q_tail->next = info;
304 vtctx->q_tail = info;
306 pthread_mutex_unlock(&vtctx->lock);
309 static int count_nalus(size_t length_code_size,
310 CMSampleBufferRef sample_buffer,
317 size_t src_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
318 CMBlockBufferRef block = CMSampleBufferGetDataBuffer(sample_buffer);
320 if (length_code_size > 4)
321 return AVERROR_INVALIDDATA;
323 while (offset < src_size) {
328 status = CMBlockBufferCopyDataBytes(block,
333 for (i = 0; i < length_code_size; i++) {
335 box_len |= size_buf[i];
338 curr_src_len = box_len + length_code_size;
339 offset += curr_src_len;
348 static CMVideoCodecType get_cm_codec_type(enum AVCodecID id)
351 case AV_CODEC_ID_H264: return kCMVideoCodecType_H264;
357 * Get the parameter sets from a CMSampleBufferRef.
358 * @param dst If *dst isn't NULL, the parameters are copied into existing
359 * memory. *dst_size must be set accordingly when *dst != NULL.
360 * If *dst is NULL, it will be allocated.
361 * In all cases, *dst_size is set to the number of bytes used starting
364 static int get_params_size(
365 AVCodecContext *avctx,
366 CMVideoFormatDescriptionRef vid_fmt,
369 size_t total_size = 0;
371 int is_count_bad = 0;
374 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
386 for (i = 0; i < ps_count || is_count_bad; i++) {
389 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
397 * When ps_count is invalid, status != 0 ends the loop normally
398 * unless we didn't get any parameter sets.
400 if (i > 0 && is_count_bad) status = 0;
405 total_size += ps_size + sizeof(start_code);
409 av_log(avctx, AV_LOG_ERROR, "Error getting parameter set sizes: %d\n", status);
410 return AVERROR_EXTERNAL;
417 static int copy_param_sets(
418 AVCodecContext *avctx,
419 CMVideoFormatDescriptionRef vid_fmt,
424 int is_count_bad = 0;
429 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
442 for (i = 0; i < ps_count || is_count_bad; i++) {
447 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
454 if (i > 0 && is_count_bad) status = 0;
459 next_offset = offset + sizeof(start_code) + ps_size;
460 if (dst_size < next_offset) {
461 av_log(avctx, AV_LOG_ERROR, "Error: buffer too small for parameter sets.\n");
462 return AVERROR_BUFFER_TOO_SMALL;
465 memcpy(dst + offset, start_code, sizeof(start_code));
466 offset += sizeof(start_code);
468 memcpy(dst + offset, ps, ps_size);
469 offset = next_offset;
473 av_log(avctx, AV_LOG_ERROR, "Error getting parameter set data: %d\n", status);
474 return AVERROR_EXTERNAL;
480 static int set_extradata(AVCodecContext *avctx, CMSampleBufferRef sample_buffer)
482 CMVideoFormatDescriptionRef vid_fmt;
486 vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
488 av_log(avctx, AV_LOG_ERROR, "No video format.\n");
489 return AVERROR_EXTERNAL;
492 status = get_params_size(avctx, vid_fmt, &total_size);
494 av_log(avctx, AV_LOG_ERROR, "Could not get parameter sets.\n");
498 avctx->extradata = av_mallocz(total_size + AV_INPUT_BUFFER_PADDING_SIZE);
499 if (!avctx->extradata) {
500 return AVERROR(ENOMEM);
502 avctx->extradata_size = total_size;
504 status = copy_param_sets(avctx, vid_fmt, avctx->extradata, total_size);
507 av_log(avctx, AV_LOG_ERROR, "Could not copy param sets.\n");
514 static void vtenc_output_callback(
516 void *sourceFrameCtx,
518 VTEncodeInfoFlags flags,
519 CMSampleBufferRef sample_buffer)
521 AVCodecContext *avctx = ctx;
522 VTEncContext *vtctx = avctx->priv_data;
523 ExtraSEI *sei = sourceFrameCtx;
525 if (vtctx->async_error) {
526 if(sample_buffer) CFRelease(sample_buffer);
530 if (status || !sample_buffer) {
531 av_log(avctx, AV_LOG_ERROR, "Error encoding frame: %d\n", (int)status);
532 set_async_error(vtctx, AVERROR_EXTERNAL);
536 if (!avctx->extradata && (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)) {
537 int set_status = set_extradata(avctx, sample_buffer);
539 set_async_error(vtctx, set_status);
544 vtenc_q_push(vtctx, sample_buffer, sei);
547 static int get_length_code_size(
548 AVCodecContext *avctx,
549 CMSampleBufferRef sample_buffer,
552 CMVideoFormatDescriptionRef vid_fmt;
556 vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
558 av_log(avctx, AV_LOG_ERROR, "Error getting buffer format description.\n");
559 return AVERROR_EXTERNAL;
562 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
569 av_log(avctx, AV_LOG_ERROR, "Error getting length code size: %d\n", status);
570 return AVERROR_EXTERNAL;
578 * Returns true on success.
580 * If profile_level_val is NULL and this method returns true, don't specify the
581 * profile/level to the encoder.
583 static bool get_vt_profile_level(AVCodecContext *avctx,
584 CFStringRef *profile_level_val)
586 VTEncContext *vtctx = avctx->priv_data;
587 int64_t profile = vtctx->profile;
589 if (profile == H264_PROF_AUTO && vtctx->level) {
590 //Need to pick a profile if level is not auto-selected.
591 profile = vtctx->has_b_frames ? H264_PROF_MAIN : H264_PROF_BASELINE;
594 *profile_level_val = NULL;
600 case H264_PROF_BASELINE:
601 switch (vtctx->level) {
602 case 0: *profile_level_val =
603 compat_keys.kVTProfileLevel_H264_Baseline_AutoLevel; break;
604 case 13: *profile_level_val = kVTProfileLevel_H264_Baseline_1_3; break;
605 case 30: *profile_level_val = kVTProfileLevel_H264_Baseline_3_0; break;
606 case 31: *profile_level_val = kVTProfileLevel_H264_Baseline_3_1; break;
607 case 32: *profile_level_val = kVTProfileLevel_H264_Baseline_3_2; break;
608 case 40: *profile_level_val =
609 compat_keys.kVTProfileLevel_H264_Baseline_4_0; break;
610 case 41: *profile_level_val = kVTProfileLevel_H264_Baseline_4_1; break;
611 case 42: *profile_level_val =
612 compat_keys.kVTProfileLevel_H264_Baseline_4_2; break;
613 case 50: *profile_level_val =
614 compat_keys.kVTProfileLevel_H264_Baseline_5_0; break;
615 case 51: *profile_level_val =
616 compat_keys.kVTProfileLevel_H264_Baseline_5_1; break;
617 case 52: *profile_level_val =
618 compat_keys.kVTProfileLevel_H264_Baseline_5_2; break;
623 switch (vtctx->level) {
624 case 0: *profile_level_val =
625 compat_keys.kVTProfileLevel_H264_Main_AutoLevel; break;
626 case 30: *profile_level_val = kVTProfileLevel_H264_Main_3_0; break;
627 case 31: *profile_level_val = kVTProfileLevel_H264_Main_3_1; break;
628 case 32: *profile_level_val = kVTProfileLevel_H264_Main_3_2; break;
629 case 40: *profile_level_val = kVTProfileLevel_H264_Main_4_0; break;
630 case 41: *profile_level_val = kVTProfileLevel_H264_Main_4_1; break;
631 case 42: *profile_level_val =
632 compat_keys.kVTProfileLevel_H264_Main_4_2; break;
633 case 50: *profile_level_val = kVTProfileLevel_H264_Main_5_0; break;
634 case 51: *profile_level_val =
635 compat_keys.kVTProfileLevel_H264_Main_5_1; break;
636 case 52: *profile_level_val =
637 compat_keys.kVTProfileLevel_H264_Main_5_2; break;
642 switch (vtctx->level) {
643 case 0: *profile_level_val =
644 compat_keys.kVTProfileLevel_H264_High_AutoLevel; break;
645 case 30: *profile_level_val =
646 compat_keys.kVTProfileLevel_H264_High_3_0; break;
647 case 31: *profile_level_val =
648 compat_keys.kVTProfileLevel_H264_High_3_1; break;
649 case 32: *profile_level_val =
650 compat_keys.kVTProfileLevel_H264_High_3_2; break;
651 case 40: *profile_level_val =
652 compat_keys.kVTProfileLevel_H264_High_4_0; break;
653 case 41: *profile_level_val =
654 compat_keys.kVTProfileLevel_H264_High_4_1; break;
655 case 42: *profile_level_val =
656 compat_keys.kVTProfileLevel_H264_High_4_2; break;
657 case 50: *profile_level_val = kVTProfileLevel_H264_High_5_0; break;
658 case 51: *profile_level_val =
659 compat_keys.kVTProfileLevel_H264_High_5_1; break;
660 case 52: *profile_level_val =
661 compat_keys.kVTProfileLevel_H264_High_5_2; break;
666 if (!*profile_level_val) {
667 av_log(avctx, AV_LOG_ERROR, "Invalid Profile/Level.\n");
674 static int get_cv_pixel_format(AVCodecContext* avctx,
675 enum AVPixelFormat fmt,
676 enum AVColorRange range,
677 int* av_pixel_format,
680 if (range_guessed) *range_guessed = range != AVCOL_RANGE_MPEG &&
681 range != AVCOL_RANGE_JPEG;
683 //MPEG range is used when no range is set
684 if (fmt == AV_PIX_FMT_NV12) {
685 *av_pixel_format = range == AVCOL_RANGE_JPEG ?
686 kCVPixelFormatType_420YpCbCr8BiPlanarFullRange :
687 kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange;
688 } else if (fmt == AV_PIX_FMT_YUV420P) {
689 *av_pixel_format = range == AVCOL_RANGE_JPEG ?
690 kCVPixelFormatType_420YpCbCr8PlanarFullRange :
691 kCVPixelFormatType_420YpCbCr8Planar;
693 return AVERROR(EINVAL);
699 static void add_color_attr(AVCodecContext *avctx, CFMutableDictionaryRef dict) {
700 VTEncContext *vtctx = avctx->priv_data;
702 if (vtctx->color_primaries) {
703 CFDictionarySetValue(dict,
704 kCVImageBufferColorPrimariesKey,
705 vtctx->color_primaries);
708 if (vtctx->transfer_function) {
709 CFDictionarySetValue(dict,
710 kCVImageBufferTransferFunctionKey,
711 vtctx->transfer_function);
714 if (vtctx->ycbcr_matrix) {
715 CFDictionarySetValue(dict,
716 kCVImageBufferYCbCrMatrixKey,
717 vtctx->ycbcr_matrix);
721 static int create_cv_pixel_buffer_info(AVCodecContext* avctx,
722 CFMutableDictionaryRef* dict)
724 CFNumberRef cv_color_format_num = NULL;
725 CFNumberRef width_num = NULL;
726 CFNumberRef height_num = NULL;
727 CFMutableDictionaryRef pixel_buffer_info = NULL;
729 int status = get_cv_pixel_format(avctx,
734 if (status) return status;
736 pixel_buffer_info = CFDictionaryCreateMutable(
739 &kCFCopyStringDictionaryKeyCallBacks,
740 &kCFTypeDictionaryValueCallBacks);
742 if (!pixel_buffer_info) goto pbinfo_nomem;
744 cv_color_format_num = CFNumberCreate(kCFAllocatorDefault,
747 if (!cv_color_format_num) goto pbinfo_nomem;
749 CFDictionarySetValue(pixel_buffer_info,
750 kCVPixelBufferPixelFormatTypeKey,
751 cv_color_format_num);
752 vt_release_num(&cv_color_format_num);
754 width_num = CFNumberCreate(kCFAllocatorDefault,
757 if (!width_num) return AVERROR(ENOMEM);
759 CFDictionarySetValue(pixel_buffer_info,
760 kCVPixelBufferWidthKey,
762 vt_release_num(&width_num);
764 height_num = CFNumberCreate(kCFAllocatorDefault,
767 if (!height_num) goto pbinfo_nomem;
769 CFDictionarySetValue(pixel_buffer_info,
770 kCVPixelBufferHeightKey,
772 vt_release_num(&height_num);
774 add_color_attr(avctx, pixel_buffer_info);
776 *dict = pixel_buffer_info;
780 vt_release_num(&cv_color_format_num);
781 vt_release_num(&width_num);
782 vt_release_num(&height_num);
783 if (pixel_buffer_info) CFRelease(pixel_buffer_info);
785 return AVERROR(ENOMEM);
788 static int get_cv_color_primaries(AVCodecContext *avctx,
789 CFStringRef *primaries)
791 enum AVColorPrimaries pri = avctx->color_primaries;
793 case AVCOL_PRI_UNSPECIFIED:
797 case AVCOL_PRI_BT709:
798 *primaries = kCVImageBufferColorPrimaries_ITU_R_709_2;
801 case AVCOL_PRI_BT2020:
802 *primaries = compat_keys.kCVImageBufferColorPrimaries_ITU_R_2020;
806 av_log(avctx, AV_LOG_ERROR, "Color primaries %s is not supported.\n", av_color_primaries_name(pri));
814 static int get_cv_transfer_function(AVCodecContext *avctx,
815 CFStringRef *transfer_fnc,
816 CFNumberRef *gamma_level)
818 enum AVColorTransferCharacteristic trc = avctx->color_trc;
823 case AVCOL_TRC_UNSPECIFIED:
824 *transfer_fnc = NULL;
827 case AVCOL_TRC_BT709:
828 *transfer_fnc = kCVImageBufferTransferFunction_ITU_R_709_2;
831 case AVCOL_TRC_SMPTE240M:
832 *transfer_fnc = kCVImageBufferTransferFunction_SMPTE_240M_1995;
835 case AVCOL_TRC_GAMMA22:
837 *transfer_fnc = kCVImageBufferTransferFunction_UseGamma;
838 *gamma_level = CFNumberCreate(NULL, kCFNumberFloat32Type, &gamma);
841 case AVCOL_TRC_GAMMA28:
843 *transfer_fnc = kCVImageBufferTransferFunction_UseGamma;
844 *gamma_level = CFNumberCreate(NULL, kCFNumberFloat32Type, &gamma);
847 case AVCOL_TRC_BT2020_10:
848 case AVCOL_TRC_BT2020_12:
849 *transfer_fnc = compat_keys.kCVImageBufferTransferFunction_ITU_R_2020;
853 av_log(avctx, AV_LOG_ERROR, "Transfer function %s is not supported.\n", av_color_transfer_name(trc));
860 static int get_cv_ycbcr_matrix(AVCodecContext *avctx, CFStringRef *matrix) {
861 switch(avctx->colorspace) {
862 case AVCOL_SPC_BT709:
863 *matrix = kCVImageBufferYCbCrMatrix_ITU_R_709_2;
866 case AVCOL_SPC_UNSPECIFIED:
870 case AVCOL_SPC_BT470BG:
871 case AVCOL_SPC_SMPTE170M:
872 *matrix = kCVImageBufferYCbCrMatrix_ITU_R_601_4;
875 case AVCOL_SPC_SMPTE240M:
876 *matrix = kCVImageBufferYCbCrMatrix_SMPTE_240M_1995;
879 case AVCOL_SPC_BT2020_NCL:
880 *matrix = compat_keys.kCVImageBufferYCbCrMatrix_ITU_R_2020;
884 av_log(avctx, AV_LOG_ERROR, "Color space %s is not supported.\n", av_color_space_name(avctx->colorspace));
891 static int vtenc_create_encoder(AVCodecContext *avctx,
892 CMVideoCodecType codec_type,
893 CFStringRef profile_level,
894 CFNumberRef gamma_level,
895 CFDictionaryRef enc_info,
896 CFDictionaryRef pixel_buffer_info,
897 VTCompressionSessionRef *session)
899 VTEncContext *vtctx = avctx->priv_data;
900 SInt32 bit_rate = avctx->bit_rate;
901 SInt32 max_rate = avctx->rc_max_rate;
902 CFNumberRef bit_rate_num;
903 CFNumberRef bytes_per_second;
904 CFNumberRef one_second;
905 CFArrayRef data_rate_limits;
906 int64_t bytes_per_second_value = 0;
907 int64_t one_second_value = 0;
910 int status = VTCompressionSessionCreate(kCFAllocatorDefault,
917 vtenc_output_callback,
921 if (status || !vtctx->session) {
922 av_log(avctx, AV_LOG_ERROR, "Error: cannot create compression session: %d\n", status);
924 #if !TARGET_OS_IPHONE
925 if (!vtctx->allow_sw) {
926 av_log(avctx, AV_LOG_ERROR, "Try -allow_sw 1. The hardware encoder may be busy, or not supported.\n");
930 return AVERROR_EXTERNAL;
933 bit_rate_num = CFNumberCreate(kCFAllocatorDefault,
936 if (!bit_rate_num) return AVERROR(ENOMEM);
938 status = VTSessionSetProperty(vtctx->session,
939 kVTCompressionPropertyKey_AverageBitRate,
941 CFRelease(bit_rate_num);
944 av_log(avctx, AV_LOG_ERROR, "Error setting bitrate property: %d\n", status);
945 return AVERROR_EXTERNAL;
948 bytes_per_second_value = max_rate >> 3;
949 bytes_per_second = CFNumberCreate(kCFAllocatorDefault,
951 &bytes_per_second_value);
952 if (!bytes_per_second) {
953 return AVERROR(ENOMEM);
955 one_second_value = 1;
956 one_second = CFNumberCreate(kCFAllocatorDefault,
960 CFRelease(bytes_per_second);
961 return AVERROR(ENOMEM);
963 nums[0] = bytes_per_second;
964 nums[1] = one_second;
965 data_rate_limits = CFArrayCreate(kCFAllocatorDefault,
968 &kCFTypeArrayCallBacks);
970 if (!data_rate_limits) {
971 CFRelease(bytes_per_second);
972 CFRelease(one_second);
973 return AVERROR(ENOMEM);
975 status = VTSessionSetProperty(vtctx->session,
976 kVTCompressionPropertyKey_DataRateLimits,
979 CFRelease(bytes_per_second);
980 CFRelease(one_second);
981 CFRelease(data_rate_limits);
984 av_log(avctx, AV_LOG_ERROR, "Error setting max bitrate property: %d\n", status);
985 return AVERROR_EXTERNAL;
989 status = VTSessionSetProperty(vtctx->session,
990 kVTCompressionPropertyKey_ProfileLevel,
993 av_log(avctx, AV_LOG_ERROR, "Error setting profile/level property: %d\n", status);
997 if (avctx->gop_size > 0) {
998 CFNumberRef interval = CFNumberCreate(kCFAllocatorDefault,
1002 return AVERROR(ENOMEM);
1005 status = VTSessionSetProperty(vtctx->session,
1006 kVTCompressionPropertyKey_MaxKeyFrameInterval,
1008 CFRelease(interval);
1011 av_log(avctx, AV_LOG_ERROR, "Error setting 'max key-frame interval' property: %d\n", status);
1012 return AVERROR_EXTERNAL;
1016 if (vtctx->frames_before) {
1017 status = VTSessionSetProperty(vtctx->session,
1018 kVTCompressionPropertyKey_MoreFramesBeforeStart,
1021 if (status == kVTPropertyNotSupportedErr) {
1022 av_log(avctx, AV_LOG_WARNING, "frames_before property is not supported on this device. Ignoring.\n");
1023 } else if (status) {
1024 av_log(avctx, AV_LOG_ERROR, "Error setting frames_before property: %d\n", status);
1028 if (vtctx->frames_after) {
1029 status = VTSessionSetProperty(vtctx->session,
1030 kVTCompressionPropertyKey_MoreFramesAfterEnd,
1033 if (status == kVTPropertyNotSupportedErr) {
1034 av_log(avctx, AV_LOG_WARNING, "frames_after property is not supported on this device. Ignoring.\n");
1035 } else if (status) {
1036 av_log(avctx, AV_LOG_ERROR, "Error setting frames_after property: %d\n", status);
1040 if (avctx->sample_aspect_ratio.num != 0) {
1043 CFMutableDictionaryRef par;
1044 AVRational *avpar = &avctx->sample_aspect_ratio;
1046 av_reduce(&avpar->num, &avpar->den,
1047 avpar->num, avpar->den,
1050 num = CFNumberCreate(kCFAllocatorDefault,
1054 den = CFNumberCreate(kCFAllocatorDefault,
1060 par = CFDictionaryCreateMutable(kCFAllocatorDefault,
1062 &kCFCopyStringDictionaryKeyCallBacks,
1063 &kCFTypeDictionaryValueCallBacks);
1065 if (!par || !num || !den) {
1066 if (par) CFRelease(par);
1067 if (num) CFRelease(num);
1068 if (den) CFRelease(den);
1070 return AVERROR(ENOMEM);
1073 CFDictionarySetValue(
1075 kCMFormatDescriptionKey_PixelAspectRatioHorizontalSpacing,
1078 CFDictionarySetValue(
1080 kCMFormatDescriptionKey_PixelAspectRatioVerticalSpacing,
1083 status = VTSessionSetProperty(vtctx->session,
1084 kVTCompressionPropertyKey_PixelAspectRatio,
1094 "Error setting pixel aspect ratio to %d:%d: %d.\n",
1095 avctx->sample_aspect_ratio.num,
1096 avctx->sample_aspect_ratio.den,
1099 return AVERROR_EXTERNAL;
1104 if (vtctx->transfer_function) {
1105 status = VTSessionSetProperty(vtctx->session,
1106 kVTCompressionPropertyKey_TransferFunction,
1107 vtctx->transfer_function);
1110 av_log(avctx, AV_LOG_WARNING, "Could not set transfer function: %d\n", status);
1115 if (vtctx->ycbcr_matrix) {
1116 status = VTSessionSetProperty(vtctx->session,
1117 kVTCompressionPropertyKey_YCbCrMatrix,
1118 vtctx->ycbcr_matrix);
1121 av_log(avctx, AV_LOG_WARNING, "Could not set ycbcr matrix: %d\n", status);
1126 if (vtctx->color_primaries) {
1127 status = VTSessionSetProperty(vtctx->session,
1128 kVTCompressionPropertyKey_ColorPrimaries,
1129 vtctx->color_primaries);
1132 av_log(avctx, AV_LOG_WARNING, "Could not set color primaries: %d\n", status);
1137 status = VTSessionSetProperty(vtctx->session,
1138 kCVImageBufferGammaLevelKey,
1142 av_log(avctx, AV_LOG_WARNING, "Could not set gamma level: %d\n", status);
1146 if (!vtctx->has_b_frames) {
1147 status = VTSessionSetProperty(vtctx->session,
1148 kVTCompressionPropertyKey_AllowFrameReordering,
1152 av_log(avctx, AV_LOG_ERROR, "Error setting 'allow frame reordering' property: %d\n", status);
1153 return AVERROR_EXTERNAL;
1157 if (vtctx->entropy != VT_ENTROPY_NOT_SET) {
1158 CFStringRef entropy = vtctx->entropy == VT_CABAC ?
1159 compat_keys.kVTH264EntropyMode_CABAC:
1160 compat_keys.kVTH264EntropyMode_CAVLC;
1162 status = VTSessionSetProperty(vtctx->session,
1163 compat_keys.kVTCompressionPropertyKey_H264EntropyMode,
1167 av_log(avctx, AV_LOG_ERROR, "Error setting entropy property: %d\n", status);
1171 if (vtctx->realtime) {
1172 status = VTSessionSetProperty(vtctx->session,
1173 compat_keys.kVTCompressionPropertyKey_RealTime,
1177 av_log(avctx, AV_LOG_ERROR, "Error setting realtime property: %d\n", status);
1181 status = VTCompressionSessionPrepareToEncodeFrames(vtctx->session);
1183 av_log(avctx, AV_LOG_ERROR, "Error: cannot prepare encoder: %d\n", status);
1184 return AVERROR_EXTERNAL;
1190 static av_cold int vtenc_init(AVCodecContext *avctx)
1192 CFMutableDictionaryRef enc_info;
1193 CFMutableDictionaryRef pixel_buffer_info;
1194 CMVideoCodecType codec_type;
1195 VTEncContext *vtctx = avctx->priv_data;
1196 CFStringRef profile_level;
1197 CFBooleanRef has_b_frames_cfbool;
1198 CFNumberRef gamma_level = NULL;
1201 pthread_once(&once_ctrl, loadVTEncSymbols);
1203 codec_type = get_cm_codec_type(avctx->codec_id);
1205 av_log(avctx, AV_LOG_ERROR, "Error: no mapping for AVCodecID %d\n", avctx->codec_id);
1206 return AVERROR(EINVAL);
1209 vtctx->has_b_frames = avctx->max_b_frames > 0;
1210 if(vtctx->has_b_frames && vtctx->profile == H264_PROF_BASELINE){
1211 av_log(avctx, AV_LOG_WARNING, "Cannot use B-frames with baseline profile. Output will not contain B-frames.\n");
1212 vtctx->has_b_frames = false;
1215 if (vtctx->entropy == VT_CABAC && vtctx->profile == H264_PROF_BASELINE) {
1216 av_log(avctx, AV_LOG_WARNING, "CABAC entropy requires 'main' or 'high' profile, but baseline was requested. Encode will not use CABAC entropy.\n");
1217 vtctx->entropy = VT_ENTROPY_NOT_SET;
1220 if (!get_vt_profile_level(avctx, &profile_level)) return AVERROR(EINVAL);
1222 vtctx->session = NULL;
1224 enc_info = CFDictionaryCreateMutable(
1225 kCFAllocatorDefault,
1227 &kCFCopyStringDictionaryKeyCallBacks,
1228 &kCFTypeDictionaryValueCallBacks
1231 if (!enc_info) return AVERROR(ENOMEM);
1233 #if !TARGET_OS_IPHONE
1234 if (!vtctx->allow_sw) {
1235 CFDictionarySetValue(enc_info,
1236 compat_keys.kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder,
1239 CFDictionarySetValue(enc_info,
1240 compat_keys.kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder,
1245 if (avctx->pix_fmt != AV_PIX_FMT_VIDEOTOOLBOX) {
1246 status = create_cv_pixel_buffer_info(avctx, &pixel_buffer_info);
1250 pixel_buffer_info = NULL;
1253 pthread_mutex_init(&vtctx->lock, NULL);
1254 pthread_cond_init(&vtctx->cv_sample_sent, NULL);
1255 vtctx->dts_delta = vtctx->has_b_frames ? -1 : 0;
1257 get_cv_transfer_function(avctx, &vtctx->transfer_function, &gamma_level);
1258 get_cv_ycbcr_matrix(avctx, &vtctx->ycbcr_matrix);
1259 get_cv_color_primaries(avctx, &vtctx->color_primaries);
1262 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
1263 status = vtenc_populate_extradata(avctx,
1273 status = vtenc_create_encoder(avctx,
1284 status = VTSessionCopyProperty(vtctx->session,
1285 kVTCompressionPropertyKey_AllowFrameReordering,
1286 kCFAllocatorDefault,
1287 &has_b_frames_cfbool);
1290 //Some devices don't output B-frames for main profile, even if requested.
1291 vtctx->has_b_frames = CFBooleanGetValue(has_b_frames_cfbool);
1292 CFRelease(has_b_frames_cfbool);
1294 avctx->has_b_frames = vtctx->has_b_frames;
1298 CFRelease(gamma_level);
1300 if (pixel_buffer_info)
1301 CFRelease(pixel_buffer_info);
1303 CFRelease(enc_info);
1308 static void vtenc_get_frame_info(CMSampleBufferRef buffer, bool *is_key_frame)
1310 CFArrayRef attachments;
1311 CFDictionaryRef attachment;
1312 CFBooleanRef not_sync;
1315 attachments = CMSampleBufferGetSampleAttachmentsArray(buffer, false);
1316 len = !attachments ? 0 : CFArrayGetCount(attachments);
1319 *is_key_frame = true;
1323 attachment = CFArrayGetValueAtIndex(attachments, 0);
1325 if (CFDictionaryGetValueIfPresent(attachment,
1326 kCMSampleAttachmentKey_NotSync,
1327 (const void **)¬_sync))
1329 *is_key_frame = !CFBooleanGetValue(not_sync);
1331 *is_key_frame = true;
1335 static int is_post_sei_nal_type(int nal_type){
1336 return nal_type != H264_NAL_SEI &&
1337 nal_type != H264_NAL_SPS &&
1338 nal_type != H264_NAL_PPS &&
1339 nal_type != H264_NAL_AUD;
1343 * Finds the sei message start/size of type find_sei_type.
1344 * If more than one of that type exists, the last one is returned.
1346 static int find_sei_end(AVCodecContext *avctx,
1352 size_t sei_payload_size = 0;
1353 int sei_payload_type = 0;
1355 uint8_t *nal_start = nal_data;
1360 nal_type = *nal_data & 0x1F;
1361 if (nal_type != H264_NAL_SEI)
1367 if (nal_data[nal_size - 1] == 0x80)
1370 while (nal_size > 0 && *nal_data > 0) {
1372 sei_payload_type += *nal_data;
1375 } while (nal_size > 0 && *nal_data == 0xFF);
1378 av_log(avctx, AV_LOG_ERROR, "Unexpected end of SEI NAL Unit parsing type.\n");
1379 return AVERROR_INVALIDDATA;
1383 sei_payload_size += *nal_data;
1386 } while (nal_size > 0 && *nal_data == 0xFF);
1388 if (nal_size < sei_payload_size) {
1389 av_log(avctx, AV_LOG_ERROR, "Unexpected end of SEI NAL Unit parsing size.\n");
1390 return AVERROR_INVALIDDATA;
1393 nal_data += sei_payload_size;
1394 nal_size -= sei_payload_size;
1397 *sei_end = nal_data;
1399 return nal_data - nal_start + 1;
1403 * Copies the data inserting emulation prevention bytes as needed.
1404 * Existing data in the destination can be taken into account by providing
1405 * dst with a dst_offset > 0.
1407 * @return The number of bytes copied on success. On failure, the negative of
1408 * the number of bytes needed to copy src is returned.
1410 static int copy_emulation_prev(const uint8_t *src,
1419 uint8_t* dst_end = dst + dst_size;
1420 const uint8_t* src_end = src + src_size;
1421 int start_at = dst_offset > 2 ? dst_offset - 2 : 0;
1423 for (i = start_at; i < dst_offset && i < dst_size; i++) {
1432 for (; src < src_end; src++, dst++) {
1434 int insert_ep3_byte = *src <= 3;
1435 if (insert_ep3_byte) {
1453 wrote_bytes = dst - dst_start;
1456 return -wrote_bytes;
1461 static int write_sei(const ExtraSEI *sei,
1466 uint8_t *sei_start = dst;
1467 size_t remaining_sei_size = sei->size;
1468 size_t remaining_dst_size = dst_size;
1473 if (!remaining_dst_size)
1474 return AVERROR_BUFFER_TOO_SMALL;
1476 while (sei_type && remaining_dst_size != 0) {
1477 int sei_byte = sei_type > 255 ? 255 : sei_type;
1480 sei_type -= sei_byte;
1482 remaining_dst_size--;
1486 return AVERROR_BUFFER_TOO_SMALL;
1488 while (remaining_sei_size && remaining_dst_size != 0) {
1489 int size_byte = remaining_sei_size > 255 ? 255 : remaining_sei_size;
1492 remaining_sei_size -= size_byte;
1494 remaining_dst_size--;
1497 if (remaining_dst_size < sei->size)
1498 return AVERROR_BUFFER_TOO_SMALL;
1500 header_bytes = dst - sei_start;
1502 offset = header_bytes;
1503 bytes_written = copy_emulation_prev(sei->data,
1508 if (bytes_written < 0)
1509 return AVERROR_BUFFER_TOO_SMALL;
1511 bytes_written += header_bytes;
1512 return bytes_written;
1516 * Copies NAL units and replaces length codes with
1517 * H.264 Annex B start codes. On failure, the contents of
1518 * dst_data may have been modified.
1520 * @param length_code_size Byte length of each length code
1521 * @param sample_buffer NAL units prefixed with length codes.
1522 * @param sei Optional A53 closed captions SEI data.
1523 * @param dst_data Must be zeroed before calling this function.
1524 * Contains the copied NAL units prefixed with
1525 * start codes when the function returns
1527 * @param dst_size Length of dst_data
1528 * @return 0 on success
1529 * AVERROR_INVALIDDATA if length_code_size is invalid
1530 * AVERROR_BUFFER_TOO_SMALL if dst_data is too small
1531 * or if a length_code in src_data specifies data beyond
1532 * the end of its buffer.
1534 static int copy_replace_length_codes(
1535 AVCodecContext *avctx,
1536 size_t length_code_size,
1537 CMSampleBufferRef sample_buffer,
1542 size_t src_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
1543 size_t remaining_src_size = src_size;
1544 size_t remaining_dst_size = dst_size;
1545 size_t src_offset = 0;
1548 uint8_t size_buf[4];
1550 CMBlockBufferRef block = CMSampleBufferGetDataBuffer(sample_buffer);
1552 if (length_code_size > 4) {
1553 return AVERROR_INVALIDDATA;
1556 while (remaining_src_size > 0) {
1557 size_t curr_src_len;
1558 size_t curr_dst_len;
1564 status = CMBlockBufferCopyDataBytes(block,
1569 av_log(avctx, AV_LOG_ERROR, "Cannot copy length: %d\n", status);
1570 return AVERROR_EXTERNAL;
1573 status = CMBlockBufferCopyDataBytes(block,
1574 src_offset + length_code_size,
1579 av_log(avctx, AV_LOG_ERROR, "Cannot copy type: %d\n", status);
1580 return AVERROR_EXTERNAL;
1585 for (i = 0; i < length_code_size; i++) {
1587 box_len |= size_buf[i];
1590 if (sei && !wrote_sei && is_post_sei_nal_type(nal_type)) {
1591 //No SEI NAL unit - insert.
1594 memcpy(dst_data, start_code, sizeof(start_code));
1595 dst_data += sizeof(start_code);
1596 remaining_dst_size -= sizeof(start_code);
1598 *dst_data = H264_NAL_SEI;
1600 remaining_dst_size--;
1602 wrote_bytes = write_sei(sei,
1603 SEI_TYPE_USER_DATA_REGISTERED,
1605 remaining_dst_size);
1607 if (wrote_bytes < 0)
1610 remaining_dst_size -= wrote_bytes;
1611 dst_data += wrote_bytes;
1613 if (remaining_dst_size <= 0)
1614 return AVERROR_BUFFER_TOO_SMALL;
1619 remaining_dst_size--;
1624 curr_src_len = box_len + length_code_size;
1625 curr_dst_len = box_len + sizeof(start_code);
1627 if (remaining_src_size < curr_src_len) {
1628 return AVERROR_BUFFER_TOO_SMALL;
1631 if (remaining_dst_size < curr_dst_len) {
1632 return AVERROR_BUFFER_TOO_SMALL;
1635 dst_box = dst_data + sizeof(start_code);
1637 memcpy(dst_data, start_code, sizeof(start_code));
1638 status = CMBlockBufferCopyDataBytes(block,
1639 src_offset + length_code_size,
1644 av_log(avctx, AV_LOG_ERROR, "Cannot copy data: %d\n", status);
1645 return AVERROR_EXTERNAL;
1648 if (sei && !wrote_sei && nal_type == H264_NAL_SEI) {
1649 //Found SEI NAL unit - append.
1654 old_sei_length = find_sei_end(avctx, dst_box, box_len, &new_sei);
1655 if (old_sei_length < 0)
1658 wrote_bytes = write_sei(sei,
1659 SEI_TYPE_USER_DATA_REGISTERED,
1661 remaining_dst_size - old_sei_length);
1662 if (wrote_bytes < 0)
1665 if (new_sei + wrote_bytes >= dst_data + remaining_dst_size)
1666 return AVERROR_BUFFER_TOO_SMALL;
1668 new_sei[wrote_bytes++] = 0x80;
1669 extra_bytes = wrote_bytes - (dst_box + box_len - new_sei);
1671 dst_data += extra_bytes;
1672 remaining_dst_size -= extra_bytes;
1677 src_offset += curr_src_len;
1678 dst_data += curr_dst_len;
1680 remaining_src_size -= curr_src_len;
1681 remaining_dst_size -= curr_dst_len;
1688 * Returns a sufficient number of bytes to contain the sei data.
1689 * It may be greater than the minimum required.
1691 static int get_sei_msg_bytes(const ExtraSEI* sei, int type){
1696 copied_size = -copy_emulation_prev(sei->data,
1702 if ((sei->size % 255) == 0) //may result in an extra byte
1705 return copied_size + sei->size / 255 + 1 + type / 255 + 1;
1708 static int vtenc_cm_to_avpacket(
1709 AVCodecContext *avctx,
1710 CMSampleBufferRef sample_buffer,
1714 VTEncContext *vtctx = avctx->priv_data;
1719 size_t length_code_size;
1720 size_t header_size = 0;
1722 size_t out_buf_size;
1723 size_t sei_nalu_size = 0;
1725 int64_t time_base_num;
1729 CMVideoFormatDescriptionRef vid_fmt;
1732 vtenc_get_frame_info(sample_buffer, &is_key_frame);
1733 status = get_length_code_size(avctx, sample_buffer, &length_code_size);
1734 if (status) return status;
1736 add_header = is_key_frame && !(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER);
1739 vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
1741 av_log(avctx, AV_LOG_ERROR, "Cannot get format description.\n");
1742 return AVERROR_EXTERNAL;
1745 int status = get_params_size(avctx, vid_fmt, &header_size);
1746 if (status) return status;
1749 status = count_nalus(length_code_size, sample_buffer, &nalu_count);
1754 size_t msg_size = get_sei_msg_bytes(sei,
1755 SEI_TYPE_USER_DATA_REGISTERED);
1757 sei_nalu_size = sizeof(start_code) + 1 + msg_size + 1;
1760 in_buf_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
1761 out_buf_size = header_size +
1764 nalu_count * ((int)sizeof(start_code) - (int)length_code_size);
1766 status = ff_alloc_packet2(avctx, pkt, out_buf_size, out_buf_size);
1771 status = copy_param_sets(avctx, vid_fmt, pkt->data, out_buf_size);
1772 if(status) return status;
1775 status = copy_replace_length_codes(
1780 pkt->data + header_size,
1781 pkt->size - header_size
1785 av_log(avctx, AV_LOG_ERROR, "Error copying packet data: %d\n", status);
1790 pkt->flags |= AV_PKT_FLAG_KEY;
1793 pts = CMSampleBufferGetPresentationTimeStamp(sample_buffer);
1794 dts = CMSampleBufferGetDecodeTimeStamp (sample_buffer);
1796 if (CMTIME_IS_INVALID(dts)) {
1797 if (!vtctx->has_b_frames) {
1800 av_log(avctx, AV_LOG_ERROR, "DTS is invalid.\n");
1801 return AVERROR_EXTERNAL;
1805 dts_delta = vtctx->dts_delta >= 0 ? vtctx->dts_delta : 0;
1806 time_base_num = avctx->time_base.num;
1807 pkt->pts = pts.value / time_base_num;
1808 pkt->dts = dts.value / time_base_num - dts_delta;
1809 pkt->size = out_buf_size;
1815 * contiguous_buf_size is 0 if not contiguous, and the size of the buffer
1816 * containing all planes if so.
1818 static int get_cv_pixel_info(
1819 AVCodecContext *avctx,
1820 const AVFrame *frame,
1826 size_t *contiguous_buf_size)
1828 VTEncContext *vtctx = avctx->priv_data;
1829 int av_format = frame->format;
1830 int av_color_range = av_frame_get_color_range(frame);
1835 status = get_cv_pixel_format(avctx, av_format, av_color_range, color, &range_guessed);
1839 "Could not get pixel format for color format '%s' range '%s'.\n",
1840 av_get_pix_fmt_name(av_format),
1841 av_color_range > AVCOL_RANGE_UNSPECIFIED &&
1842 av_color_range < AVCOL_RANGE_NB ?
1843 av_color_range_name(av_color_range) :
1846 return AVERROR(EINVAL);
1849 if (range_guessed) {
1850 if (!vtctx->warned_color_range) {
1851 vtctx->warned_color_range = true;
1854 "Color range not set for %s. Using MPEG range.\n",
1855 av_get_pix_fmt_name(av_format));
1858 av_log(avctx, AV_LOG_WARNING, "");
1861 switch (av_format) {
1862 case AV_PIX_FMT_NV12:
1865 widths [0] = avctx->width;
1866 heights[0] = avctx->height;
1867 strides[0] = frame ? frame->linesize[0] : avctx->width;
1869 widths [1] = (avctx->width + 1) / 2;
1870 heights[1] = (avctx->height + 1) / 2;
1871 strides[1] = frame ? frame->linesize[1] : (avctx->width + 1) & -2;
1874 case AV_PIX_FMT_YUV420P:
1877 widths [0] = avctx->width;
1878 heights[0] = avctx->height;
1879 strides[0] = frame ? frame->linesize[0] : avctx->width;
1881 widths [1] = (avctx->width + 1) / 2;
1882 heights[1] = (avctx->height + 1) / 2;
1883 strides[1] = frame ? frame->linesize[1] : (avctx->width + 1) / 2;
1885 widths [2] = (avctx->width + 1) / 2;
1886 heights[2] = (avctx->height + 1) / 2;
1887 strides[2] = frame ? frame->linesize[2] : (avctx->width + 1) / 2;
1894 "Could not get frame format info for color %d range %d.\n",
1898 return AVERROR(EINVAL);
1901 *contiguous_buf_size = 0;
1902 for (i = 0; i < *plane_count; i++) {
1903 if (i < *plane_count - 1 &&
1904 frame->data[i] + strides[i] * heights[i] != frame->data[i + 1]) {
1905 *contiguous_buf_size = 0;
1909 *contiguous_buf_size += strides[i] * heights[i];
1915 #if !TARGET_OS_IPHONE
1916 //Not used on iOS - frame is always copied.
1917 static void free_avframe(
1922 const void *plane_addresses[])
1924 AVFrame *frame = release_ctx;
1925 av_frame_free(&frame);
1928 //Not used on OSX - frame is never copied.
1929 static int copy_avframe_to_pixel_buffer(AVCodecContext *avctx,
1930 const AVFrame *frame,
1931 CVPixelBufferRef cv_img,
1932 const size_t *plane_strides,
1933 const size_t *plane_rows)
1945 status = CVPixelBufferLockBaseAddress(cv_img, 0);
1950 "Error: Could not lock base address of CVPixelBuffer: %d.\n",
1955 if (CVPixelBufferIsPlanar(cv_img)) {
1956 plane_count = CVPixelBufferGetPlaneCount(cv_img);
1957 for (i = 0; frame->data[i]; i++) {
1958 if (i == plane_count) {
1959 CVPixelBufferUnlockBaseAddress(cv_img, 0);
1962 "Error: different number of planes in AVFrame and CVPixelBuffer.\n"
1965 return AVERROR_EXTERNAL;
1968 dst_addr = (uint8_t*)CVPixelBufferGetBaseAddressOfPlane(cv_img, i);
1969 src_addr = (uint8_t*)frame->data[i];
1970 dst_stride = CVPixelBufferGetBytesPerRowOfPlane(cv_img, i);
1971 src_stride = plane_strides[i];
1972 rows = plane_rows[i];
1974 if (dst_stride == src_stride) {
1975 memcpy(dst_addr, src_addr, src_stride * rows);
1977 copy_bytes = dst_stride < src_stride ? dst_stride : src_stride;
1979 for (j = 0; j < rows; j++) {
1980 memcpy(dst_addr + j * dst_stride, src_addr + j * src_stride, copy_bytes);
1985 if (frame->data[1]) {
1986 CVPixelBufferUnlockBaseAddress(cv_img, 0);
1989 "Error: different number of planes in AVFrame and non-planar CVPixelBuffer.\n"
1992 return AVERROR_EXTERNAL;
1995 dst_addr = (uint8_t*)CVPixelBufferGetBaseAddress(cv_img);
1996 src_addr = (uint8_t*)frame->data[0];
1997 dst_stride = CVPixelBufferGetBytesPerRow(cv_img);
1998 src_stride = plane_strides[0];
1999 rows = plane_rows[0];
2001 if (dst_stride == src_stride) {
2002 memcpy(dst_addr, src_addr, src_stride * rows);
2004 copy_bytes = dst_stride < src_stride ? dst_stride : src_stride;
2006 for (j = 0; j < rows; j++) {
2007 memcpy(dst_addr + j * dst_stride, src_addr + j * src_stride, copy_bytes);
2012 status = CVPixelBufferUnlockBaseAddress(cv_img, 0);
2014 av_log(avctx, AV_LOG_ERROR, "Error: Could not unlock CVPixelBuffer base address: %d.\n", status);
2015 return AVERROR_EXTERNAL;
2020 #endif //!TARGET_OS_IPHONE
2022 static int create_cv_pixel_buffer(AVCodecContext *avctx,
2023 const AVFrame *frame,
2024 CVPixelBufferRef *cv_img)
2028 size_t widths [AV_NUM_DATA_POINTERS];
2029 size_t heights[AV_NUM_DATA_POINTERS];
2030 size_t strides[AV_NUM_DATA_POINTERS];
2032 size_t contiguous_buf_size;
2033 #if TARGET_OS_IPHONE
2034 CVPixelBufferPoolRef pix_buf_pool;
2035 VTEncContext* vtctx = avctx->priv_data;
2037 CFMutableDictionaryRef pix_buf_attachments = CFDictionaryCreateMutable(
2038 kCFAllocatorDefault,
2040 &kCFCopyStringDictionaryKeyCallBacks,
2041 &kCFTypeDictionaryValueCallBacks);
2043 if (!pix_buf_attachments) return AVERROR(ENOMEM);
2046 if (avctx->pix_fmt == AV_PIX_FMT_VIDEOTOOLBOX) {
2047 av_assert0(frame->format == AV_PIX_FMT_VIDEOTOOLBOX);
2049 *cv_img = (CVPixelBufferRef)frame->data[3];
2050 av_assert0(*cv_img);
2056 memset(widths, 0, sizeof(widths));
2057 memset(heights, 0, sizeof(heights));
2058 memset(strides, 0, sizeof(strides));
2060 status = get_cv_pixel_info(
2068 &contiguous_buf_size
2075 "Error: Cannot convert format %d color_range %d: %d\n",
2077 av_frame_get_color_range(frame),
2081 return AVERROR_EXTERNAL;
2084 #if TARGET_OS_IPHONE
2085 pix_buf_pool = VTCompressionSessionGetPixelBufferPool(vtctx->session);
2086 if (!pix_buf_pool) {
2087 av_log(avctx, AV_LOG_ERROR, "Could not get pixel buffer pool.\n");
2088 return AVERROR_EXTERNAL;
2091 status = CVPixelBufferPoolCreatePixelBuffer(NULL,
2097 av_log(avctx, AV_LOG_ERROR, "Could not create pixel buffer from pool: %d.\n", status);
2098 return AVERROR_EXTERNAL;
2101 status = copy_avframe_to_pixel_buffer(avctx, frame, *cv_img, strides, heights);
2108 AVFrame *enc_frame = av_frame_alloc();
2109 if (!enc_frame) return AVERROR(ENOMEM);
2111 status = av_frame_ref(enc_frame, frame);
2113 av_frame_free(&enc_frame);
2117 status = CVPixelBufferCreateWithPlanarBytes(
2118 kCFAllocatorDefault,
2123 contiguous_buf_size,
2125 (void **)enc_frame->data,
2135 add_color_attr(avctx, pix_buf_attachments);
2136 CVBufferSetAttachments(*cv_img, pix_buf_attachments, kCVAttachmentMode_ShouldPropagate);
2137 CFRelease(pix_buf_attachments);
2140 av_log(avctx, AV_LOG_ERROR, "Error: Could not create CVPixelBuffer: %d\n", status);
2141 return AVERROR_EXTERNAL;
2148 static int create_encoder_dict_h264(const AVFrame *frame,
2149 CFDictionaryRef* dict_out)
2151 CFDictionaryRef dict = NULL;
2152 if (frame->pict_type == AV_PICTURE_TYPE_I) {
2153 const void *keys[] = { kVTEncodeFrameOptionKey_ForceKeyFrame };
2154 const void *vals[] = { kCFBooleanTrue };
2156 dict = CFDictionaryCreate(NULL, keys, vals, 1, NULL, NULL);
2157 if(!dict) return AVERROR(ENOMEM);
2164 static int vtenc_send_frame(AVCodecContext *avctx,
2165 VTEncContext *vtctx,
2166 const AVFrame *frame)
2169 CFDictionaryRef frame_dict;
2170 CVPixelBufferRef cv_img = NULL;
2171 AVFrameSideData *side_data = NULL;
2172 ExtraSEI *sei = NULL;
2173 int status = create_cv_pixel_buffer(avctx, frame, &cv_img);
2175 if (status) return status;
2177 status = create_encoder_dict_h264(frame, &frame_dict);
2183 side_data = av_frame_get_side_data(frame, AV_FRAME_DATA_A53_CC);
2184 if (vtctx->a53_cc && side_data && side_data->size) {
2185 sei = av_mallocz(sizeof(*sei));
2187 av_log(avctx, AV_LOG_ERROR, "Not enough memory for closed captions, skipping\n");
2189 int ret = ff_alloc_a53_sei(frame, 0, &sei->data, &sei->size);
2191 av_log(avctx, AV_LOG_ERROR, "Not enough memory for closed captions, skipping\n");
2198 time = CMTimeMake(frame->pts * avctx->time_base.num, avctx->time_base.den);
2199 status = VTCompressionSessionEncodeFrame(
2209 if (frame_dict) CFRelease(frame_dict);
2213 av_log(avctx, AV_LOG_ERROR, "Error: cannot encode frame: %d\n", status);
2214 return AVERROR_EXTERNAL;
2220 static av_cold int vtenc_frame(
2221 AVCodecContext *avctx,
2223 const AVFrame *frame,
2226 VTEncContext *vtctx = avctx->priv_data;
2229 CMSampleBufferRef buf = NULL;
2230 ExtraSEI *sei = NULL;
2233 status = vtenc_send_frame(avctx, vtctx, frame);
2236 status = AVERROR_EXTERNAL;
2240 if (vtctx->frame_ct_in == 0) {
2241 vtctx->first_pts = frame->pts;
2242 } else if(vtctx->frame_ct_in == 1 && vtctx->has_b_frames) {
2243 vtctx->dts_delta = frame->pts - vtctx->first_pts;
2246 vtctx->frame_ct_in++;
2247 } else if(!vtctx->flushing) {
2248 vtctx->flushing = true;
2250 status = VTCompressionSessionCompleteFrames(vtctx->session,
2254 av_log(avctx, AV_LOG_ERROR, "Error flushing frames: %d\n", status);
2255 status = AVERROR_EXTERNAL;
2261 get_frame = vtctx->dts_delta >= 0 || !frame;
2267 status = vtenc_q_pop(vtctx, !frame, &buf, &sei);
2268 if (status) goto end_nopkt;
2269 if (!buf) goto end_nopkt;
2271 status = vtenc_cm_to_avpacket(avctx, buf, pkt, sei);
2273 if (sei->data) av_free(sei->data);
2277 if (status) goto end_nopkt;
2283 av_packet_unref(pkt);
2287 static int vtenc_populate_extradata(AVCodecContext *avctx,
2288 CMVideoCodecType codec_type,
2289 CFStringRef profile_level,
2290 CFNumberRef gamma_level,
2291 CFDictionaryRef enc_info,
2292 CFDictionaryRef pixel_buffer_info)
2294 VTEncContext *vtctx = avctx->priv_data;
2295 AVFrame *frame = av_frame_alloc();
2296 int y_size = avctx->width * avctx->height;
2297 int chroma_size = (avctx->width / 2) * (avctx->height / 2);
2298 CMSampleBufferRef buf = NULL;
2302 return AVERROR(ENOMEM);
2304 frame->buf[0] = av_buffer_alloc(y_size + 2 * chroma_size);
2307 status = AVERROR(ENOMEM);
2311 status = vtenc_create_encoder(avctx,
2321 frame->data[0] = frame->buf[0]->data;
2322 memset(frame->data[0], 0, y_size);
2324 frame->data[1] = frame->buf[0]->data + y_size;
2325 memset(frame->data[1], 128, chroma_size);
2328 if (avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
2329 frame->data[2] = frame->buf[0]->data + y_size + chroma_size;
2330 memset(frame->data[2], 128, chroma_size);
2333 frame->linesize[0] = avctx->width;
2335 if (avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
2336 frame->linesize[1] =
2337 frame->linesize[2] = (avctx->width + 1) / 2;
2339 frame->linesize[1] = (avctx->width + 1) / 2;
2342 frame->format = avctx->pix_fmt;
2343 frame->width = avctx->width;
2344 frame->height = avctx->height;
2345 av_frame_set_colorspace(frame, avctx->colorspace);
2346 av_frame_set_color_range(frame, avctx->color_range);
2347 frame->color_trc = avctx->color_trc;
2348 frame->color_primaries = avctx->color_primaries;
2351 status = vtenc_send_frame(avctx, vtctx, frame);
2353 av_log(avctx, AV_LOG_ERROR, "Error sending frame: %d\n", status);
2357 //Populates extradata - output frames are flushed and param sets are available.
2358 status = VTCompressionSessionCompleteFrames(vtctx->session,
2364 status = vtenc_q_pop(vtctx, 0, &buf, NULL);
2366 av_log(avctx, AV_LOG_ERROR, "popping: %d\n", status);
2376 CFRelease(vtctx->session);
2378 vtctx->session = NULL;
2379 vtctx->frame_ct_out = 0;
2381 av_frame_unref(frame);
2382 av_frame_free(&frame);
2384 av_assert0(status != 0 || (avctx->extradata && avctx->extradata_size > 0));
2389 static av_cold int vtenc_close(AVCodecContext *avctx)
2391 VTEncContext *vtctx = avctx->priv_data;
2393 if(!vtctx->session) return 0;
2395 VTCompressionSessionCompleteFrames(vtctx->session,
2397 clear_frame_queue(vtctx);
2398 pthread_cond_destroy(&vtctx->cv_sample_sent);
2399 pthread_mutex_destroy(&vtctx->lock);
2400 CFRelease(vtctx->session);
2401 vtctx->session = NULL;
2403 if (vtctx->color_primaries) {
2404 CFRelease(vtctx->color_primaries);
2405 vtctx->color_primaries = NULL;
2408 if (vtctx->transfer_function) {
2409 CFRelease(vtctx->transfer_function);
2410 vtctx->transfer_function = NULL;
2413 if (vtctx->ycbcr_matrix) {
2414 CFRelease(vtctx->ycbcr_matrix);
2415 vtctx->ycbcr_matrix = NULL;
2421 static const enum AVPixelFormat pix_fmts[] = {
2422 AV_PIX_FMT_VIDEOTOOLBOX,
2428 #define OFFSET(x) offsetof(VTEncContext, x)
2429 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
2430 static const AVOption options[] = {
2431 { "profile", "Profile", OFFSET(profile), AV_OPT_TYPE_INT, { .i64 = H264_PROF_AUTO }, H264_PROF_AUTO, H264_PROF_COUNT, VE, "profile" },
2432 { "baseline", "Baseline Profile", 0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_BASELINE }, INT_MIN, INT_MAX, VE, "profile" },
2433 { "main", "Main Profile", 0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_MAIN }, INT_MIN, INT_MAX, VE, "profile" },
2434 { "high", "High Profile", 0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_HIGH }, INT_MIN, INT_MAX, VE, "profile" },
2436 { "level", "Level", OFFSET(level), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 52, VE, "level" },
2437 { "1.3", "Level 1.3, only available with Baseline Profile", 0, AV_OPT_TYPE_CONST, { .i64 = 13 }, INT_MIN, INT_MAX, VE, "level" },
2438 { "3.0", "Level 3.0", 0, AV_OPT_TYPE_CONST, { .i64 = 30 }, INT_MIN, INT_MAX, VE, "level" },
2439 { "3.1", "Level 3.1", 0, AV_OPT_TYPE_CONST, { .i64 = 31 }, INT_MIN, INT_MAX, VE, "level" },
2440 { "3.2", "Level 3.2", 0, AV_OPT_TYPE_CONST, { .i64 = 32 }, INT_MIN, INT_MAX, VE, "level" },
2441 { "4.0", "Level 4.0", 0, AV_OPT_TYPE_CONST, { .i64 = 40 }, INT_MIN, INT_MAX, VE, "level" },
2442 { "4.1", "Level 4.1", 0, AV_OPT_TYPE_CONST, { .i64 = 41 }, INT_MIN, INT_MAX, VE, "level" },
2443 { "4.2", "Level 4.2", 0, AV_OPT_TYPE_CONST, { .i64 = 42 }, INT_MIN, INT_MAX, VE, "level" },
2444 { "5.0", "Level 5.0", 0, AV_OPT_TYPE_CONST, { .i64 = 50 }, INT_MIN, INT_MAX, VE, "level" },
2445 { "5.1", "Level 5.1", 0, AV_OPT_TYPE_CONST, { .i64 = 51 }, INT_MIN, INT_MAX, VE, "level" },
2446 { "5.2", "Level 5.2", 0, AV_OPT_TYPE_CONST, { .i64 = 52 }, INT_MIN, INT_MAX, VE, "level" },
2448 { "allow_sw", "Allow software encoding", OFFSET(allow_sw), AV_OPT_TYPE_BOOL,
2449 { .i64 = 0 }, 0, 1, VE },
2451 { "coder", "Entropy coding", OFFSET(entropy), AV_OPT_TYPE_INT, { .i64 = VT_ENTROPY_NOT_SET }, VT_ENTROPY_NOT_SET, VT_CABAC, VE, "coder" },
2452 { "cavlc", "CAVLC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CAVLC }, INT_MIN, INT_MAX, VE, "coder" },
2453 { "vlc", "CAVLC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CAVLC }, INT_MIN, INT_MAX, VE, "coder" },
2454 { "cabac", "CABAC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CABAC }, INT_MIN, INT_MAX, VE, "coder" },
2455 { "ac", "CABAC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CABAC }, INT_MIN, INT_MAX, VE, "coder" },
2457 { "realtime", "Hint that encoding should happen in real-time if not faster (e.g. capturing from camera).",
2458 OFFSET(realtime), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
2460 { "frames_before", "Other frames will come before the frames in this session. This helps smooth concatenation issues.",
2461 OFFSET(frames_before), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
2462 { "frames_after", "Other frames will come after the frames in this session. This helps smooth concatenation issues.",
2463 OFFSET(frames_after), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
2465 { "a53cc", "Use A53 Closed Captions (if available)", OFFSET(a53_cc), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, VE },
2470 static const AVClass h264_videotoolbox_class = {
2471 .class_name = "h264_videotoolbox",
2472 .item_name = av_default_item_name,
2474 .version = LIBAVUTIL_VERSION_INT,
2477 AVCodec ff_h264_videotoolbox_encoder = {
2478 .name = "h264_videotoolbox",
2479 .long_name = NULL_IF_CONFIG_SMALL("VideoToolbox H.264 Encoder"),
2480 .type = AVMEDIA_TYPE_VIDEO,
2481 .id = AV_CODEC_ID_H264,
2482 .priv_data_size = sizeof(VTEncContext),
2483 .pix_fmts = pix_fmts,
2485 .encode2 = vtenc_frame,
2486 .close = vtenc_close,
2487 .capabilities = AV_CODEC_CAP_DELAY,
2488 .priv_class = &h264_videotoolbox_class,
2489 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE |
2490 FF_CODEC_CAP_INIT_CLEANUP,