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/avstring.h"
30 #include "libavcodec/avcodec.h"
31 #include "libavutil/pixdesc.h"
38 //These symbols may not be present
40 CFStringRef kCVImageBufferColorPrimaries_ITU_R_2020;
41 CFStringRef kCVImageBufferTransferFunction_ITU_R_2020;
42 CFStringRef kCVImageBufferYCbCrMatrix_ITU_R_2020;
44 CFStringRef kVTCompressionPropertyKey_H264EntropyMode;
45 CFStringRef kVTH264EntropyMode_CAVLC;
46 CFStringRef kVTH264EntropyMode_CABAC;
48 CFStringRef kVTProfileLevel_H264_Baseline_4_0;
49 CFStringRef kVTProfileLevel_H264_Baseline_4_2;
50 CFStringRef kVTProfileLevel_H264_Baseline_5_0;
51 CFStringRef kVTProfileLevel_H264_Baseline_5_1;
52 CFStringRef kVTProfileLevel_H264_Baseline_5_2;
53 CFStringRef kVTProfileLevel_H264_Baseline_AutoLevel;
54 CFStringRef kVTProfileLevel_H264_Main_4_2;
55 CFStringRef kVTProfileLevel_H264_Main_5_1;
56 CFStringRef kVTProfileLevel_H264_Main_5_2;
57 CFStringRef kVTProfileLevel_H264_Main_AutoLevel;
58 CFStringRef kVTProfileLevel_H264_High_3_0;
59 CFStringRef kVTProfileLevel_H264_High_3_1;
60 CFStringRef kVTProfileLevel_H264_High_3_2;
61 CFStringRef kVTProfileLevel_H264_High_4_0;
62 CFStringRef kVTProfileLevel_H264_High_4_1;
63 CFStringRef kVTProfileLevel_H264_High_4_2;
64 CFStringRef kVTProfileLevel_H264_High_5_1;
65 CFStringRef kVTProfileLevel_H264_High_5_2;
66 CFStringRef kVTProfileLevel_H264_High_AutoLevel;
68 CFStringRef kVTCompressionPropertyKey_RealTime;
70 CFStringRef kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder;
71 CFStringRef kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder;
74 #define GET_SYM(symbol, defaultVal) \
76 CFStringRef cfstr = *(CFStringRef*)dlsym(RTLD_DEFAULT, #symbol); \
78 compat_keys.symbol = CFSTR(defaultVal); \
80 compat_keys.symbol = cfstr; \
83 static pthread_once_t once_ctrl = PTHREAD_ONCE_INIT;
85 static void loadVTEncSymbols(){
86 GET_SYM(kCVImageBufferColorPrimaries_ITU_R_2020, "ITU_R_2020");
87 GET_SYM(kCVImageBufferTransferFunction_ITU_R_2020, "ITU_R_2020");
88 GET_SYM(kCVImageBufferYCbCrMatrix_ITU_R_2020, "ITU_R_2020");
90 GET_SYM(kVTCompressionPropertyKey_H264EntropyMode, "H264EntropyMode");
91 GET_SYM(kVTH264EntropyMode_CAVLC, "CAVLC");
92 GET_SYM(kVTH264EntropyMode_CABAC, "CABAC");
94 GET_SYM(kVTProfileLevel_H264_Baseline_4_0, "H264_Baseline_4_0");
95 GET_SYM(kVTProfileLevel_H264_Baseline_4_2, "H264_Baseline_4_2");
96 GET_SYM(kVTProfileLevel_H264_Baseline_5_0, "H264_Baseline_5_0");
97 GET_SYM(kVTProfileLevel_H264_Baseline_5_1, "H264_Baseline_5_1");
98 GET_SYM(kVTProfileLevel_H264_Baseline_5_2, "H264_Baseline_5_2");
99 GET_SYM(kVTProfileLevel_H264_Baseline_AutoLevel, "H264_Baseline_AutoLevel");
100 GET_SYM(kVTProfileLevel_H264_Main_4_2, "H264_Main_4_2");
101 GET_SYM(kVTProfileLevel_H264_Main_5_1, "H264_Main_5_1");
102 GET_SYM(kVTProfileLevel_H264_Main_5_2, "H264_Main_5_2");
103 GET_SYM(kVTProfileLevel_H264_Main_AutoLevel, "H264_Main_AutoLevel");
104 GET_SYM(kVTProfileLevel_H264_High_3_0, "H264_High_3_0");
105 GET_SYM(kVTProfileLevel_H264_High_3_1, "H264_High_3_1");
106 GET_SYM(kVTProfileLevel_H264_High_3_2, "H264_High_3_2");
107 GET_SYM(kVTProfileLevel_H264_High_4_0, "H264_High_4_0");
108 GET_SYM(kVTProfileLevel_H264_High_4_1, "H264_High_4_1");
109 GET_SYM(kVTProfileLevel_H264_High_4_2, "H264_High_4_2");
110 GET_SYM(kVTProfileLevel_H264_High_5_1, "H264_High_5_1");
111 GET_SYM(kVTProfileLevel_H264_High_5_2, "H264_High_5_2");
112 GET_SYM(kVTProfileLevel_H264_High_AutoLevel, "H264_High_AutoLevel");
114 GET_SYM(kVTCompressionPropertyKey_RealTime, "RealTime");
116 GET_SYM(kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder,
117 "EnableHardwareAcceleratedVideoEncoder");
118 GET_SYM(kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder,
119 "RequireHardwareAcceleratedVideoEncoder");
122 typedef enum VT_H264Profile {
130 typedef enum VTH264Entropy{
136 static const uint8_t start_code[] = { 0, 0, 0, 1 };
138 typedef struct ExtraSEI {
143 typedef struct BufNode {
144 CMSampleBufferRef cm_buffer;
146 struct BufNode* next;
150 typedef struct VTEncContext {
152 VTCompressionSessionRef session;
153 CFStringRef ycbcr_matrix;
154 CFStringRef color_primaries;
155 CFStringRef transfer_function;
157 pthread_mutex_t lock;
158 pthread_cond_t cv_sample_sent;
165 int64_t frame_ct_out;
175 int64_t frames_before;
176 int64_t frames_after;
182 bool warned_color_range;
186 static int vtenc_populate_extradata(AVCodecContext *avctx,
187 CMVideoCodecType codec_type,
188 CFStringRef profile_level,
189 CFNumberRef gamma_level,
190 CFDictionaryRef enc_info,
191 CFDictionaryRef pixel_buffer_info);
194 * NULL-safe release of *refPtr, and sets value to NULL.
196 static void vt_release_num(CFNumberRef* refPtr){
205 static void set_async_error(VTEncContext *vtctx, int err)
209 pthread_mutex_lock(&vtctx->lock);
211 vtctx->async_error = err;
213 info = vtctx->q_head;
214 vtctx->q_head = vtctx->q_tail = NULL;
217 BufNode *next = info->next;
218 CFRelease(info->cm_buffer);
223 pthread_mutex_unlock(&vtctx->lock);
226 static void clear_frame_queue(VTEncContext *vtctx)
228 set_async_error(vtctx, 0);
231 static int vtenc_q_pop(VTEncContext *vtctx, bool wait, CMSampleBufferRef *buf, ExtraSEI **sei)
235 pthread_mutex_lock(&vtctx->lock);
237 if (vtctx->async_error) {
238 pthread_mutex_unlock(&vtctx->lock);
239 return vtctx->async_error;
242 if (vtctx->flushing && vtctx->frame_ct_in == vtctx->frame_ct_out) {
245 pthread_mutex_unlock(&vtctx->lock);
249 while (!vtctx->q_head && !vtctx->async_error && wait) {
250 pthread_cond_wait(&vtctx->cv_sample_sent, &vtctx->lock);
253 if (!vtctx->q_head) {
254 pthread_mutex_unlock(&vtctx->lock);
259 info = vtctx->q_head;
260 vtctx->q_head = vtctx->q_head->next;
261 if (!vtctx->q_head) {
262 vtctx->q_tail = NULL;
265 pthread_mutex_unlock(&vtctx->lock);
267 *buf = info->cm_buffer;
270 } else if (info->sei) {
271 if (info->sei->data) av_free(info->sei->data);
276 vtctx->frame_ct_out++;
281 static void vtenc_q_push(VTEncContext *vtctx, CMSampleBufferRef buffer, ExtraSEI *sei)
283 BufNode *info = av_malloc(sizeof(BufNode));
285 set_async_error(vtctx, AVERROR(ENOMEM));
290 info->cm_buffer = buffer;
294 pthread_mutex_lock(&vtctx->lock);
295 pthread_cond_signal(&vtctx->cv_sample_sent);
297 if (!vtctx->q_head) {
298 vtctx->q_head = info;
300 vtctx->q_tail->next = info;
303 vtctx->q_tail = info;
305 pthread_mutex_unlock(&vtctx->lock);
308 static int count_nalus(size_t length_code_size,
309 CMSampleBufferRef sample_buffer,
316 size_t src_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
317 CMBlockBufferRef block = CMSampleBufferGetDataBuffer(sample_buffer);
319 if (length_code_size > 4)
320 return AVERROR_INVALIDDATA;
322 while (offset < src_size) {
327 status = CMBlockBufferCopyDataBytes(block,
332 for (i = 0; i < length_code_size; i++) {
334 box_len |= size_buf[i];
337 curr_src_len = box_len + length_code_size;
338 offset += curr_src_len;
347 static CMVideoCodecType get_cm_codec_type(enum AVCodecID id)
350 case AV_CODEC_ID_H264: return kCMVideoCodecType_H264;
356 * Get the parameter sets from a CMSampleBufferRef.
357 * @param dst If *dst isn't NULL, the parameters are copied into existing
358 * memory. *dst_size must be set accordingly when *dst != NULL.
359 * If *dst is NULL, it will be allocated.
360 * In all cases, *dst_size is set to the number of bytes used starting
363 static int get_params_size(
364 AVCodecContext *avctx,
365 CMVideoFormatDescriptionRef vid_fmt,
368 size_t total_size = 0;
370 int is_count_bad = 0;
373 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
385 for (i = 0; i < ps_count || is_count_bad; i++) {
388 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
396 * When ps_count is invalid, status != 0 ends the loop normally
397 * unless we didn't get any parameter sets.
399 if (i > 0 && is_count_bad) status = 0;
404 total_size += ps_size + sizeof(start_code);
408 av_log(avctx, AV_LOG_ERROR, "Error getting parameter set sizes: %d\n", status);
409 return AVERROR_EXTERNAL;
416 static int copy_param_sets(
417 AVCodecContext *avctx,
418 CMVideoFormatDescriptionRef vid_fmt,
423 int is_count_bad = 0;
428 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
441 for (i = 0; i < ps_count || is_count_bad; i++) {
446 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
453 if (i > 0 && is_count_bad) status = 0;
458 next_offset = offset + sizeof(start_code) + ps_size;
459 if (dst_size < next_offset) {
460 av_log(avctx, AV_LOG_ERROR, "Error: buffer too small for parameter sets.\n");
461 return AVERROR_BUFFER_TOO_SMALL;
464 memcpy(dst + offset, start_code, sizeof(start_code));
465 offset += sizeof(start_code);
467 memcpy(dst + offset, ps, ps_size);
468 offset = next_offset;
472 av_log(avctx, AV_LOG_ERROR, "Error getting parameter set data: %d\n", status);
473 return AVERROR_EXTERNAL;
479 static int set_extradata(AVCodecContext *avctx, CMSampleBufferRef sample_buffer)
481 CMVideoFormatDescriptionRef vid_fmt;
485 vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
487 av_log(avctx, AV_LOG_ERROR, "No video format.\n");
488 return AVERROR_EXTERNAL;
491 status = get_params_size(avctx, vid_fmt, &total_size);
493 av_log(avctx, AV_LOG_ERROR, "Could not get parameter sets.\n");
497 avctx->extradata = av_mallocz(total_size + AV_INPUT_BUFFER_PADDING_SIZE);
498 if (!avctx->extradata) {
499 return AVERROR(ENOMEM);
501 avctx->extradata_size = total_size;
503 status = copy_param_sets(avctx, vid_fmt, avctx->extradata, total_size);
506 av_log(avctx, AV_LOG_ERROR, "Could not copy param sets.\n");
513 static void vtenc_output_callback(
515 void *sourceFrameCtx,
517 VTEncodeInfoFlags flags,
518 CMSampleBufferRef sample_buffer)
520 AVCodecContext *avctx = ctx;
521 VTEncContext *vtctx = avctx->priv_data;
522 ExtraSEI *sei = sourceFrameCtx;
524 if (vtctx->async_error) {
525 if(sample_buffer) CFRelease(sample_buffer);
529 if (status || !sample_buffer) {
530 av_log(avctx, AV_LOG_ERROR, "Error encoding frame: %d\n", (int)status);
531 set_async_error(vtctx, AVERROR_EXTERNAL);
535 if (!avctx->extradata && (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)) {
536 int set_status = set_extradata(avctx, sample_buffer);
538 set_async_error(vtctx, set_status);
543 vtenc_q_push(vtctx, sample_buffer, sei);
546 static int get_length_code_size(
547 AVCodecContext *avctx,
548 CMSampleBufferRef sample_buffer,
551 CMVideoFormatDescriptionRef vid_fmt;
555 vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
557 av_log(avctx, AV_LOG_ERROR, "Error getting buffer format description.\n");
558 return AVERROR_EXTERNAL;
561 status = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(vid_fmt,
568 av_log(avctx, AV_LOG_ERROR, "Error getting length code size: %d\n", status);
569 return AVERROR_EXTERNAL;
577 * Returns true on success.
579 * If profile_level_val is NULL and this method returns true, don't specify the
580 * profile/level to the encoder.
582 static bool get_vt_profile_level(AVCodecContext *avctx,
583 CFStringRef *profile_level_val)
585 VTEncContext *vtctx = avctx->priv_data;
586 int64_t profile = vtctx->profile;
588 if (profile == H264_PROF_AUTO && vtctx->level) {
589 //Need to pick a profile if level is not auto-selected.
590 profile = vtctx->has_b_frames ? H264_PROF_MAIN : H264_PROF_BASELINE;
593 *profile_level_val = NULL;
599 case H264_PROF_BASELINE:
600 switch (vtctx->level) {
601 case 0: *profile_level_val =
602 compat_keys.kVTProfileLevel_H264_Baseline_AutoLevel; break;
603 case 13: *profile_level_val = kVTProfileLevel_H264_Baseline_1_3; break;
604 case 30: *profile_level_val = kVTProfileLevel_H264_Baseline_3_0; break;
605 case 31: *profile_level_val = kVTProfileLevel_H264_Baseline_3_1; break;
606 case 32: *profile_level_val = kVTProfileLevel_H264_Baseline_3_2; break;
607 case 40: *profile_level_val =
608 compat_keys.kVTProfileLevel_H264_Baseline_4_0; break;
609 case 41: *profile_level_val = kVTProfileLevel_H264_Baseline_4_1; break;
610 case 42: *profile_level_val =
611 compat_keys.kVTProfileLevel_H264_Baseline_4_2; break;
612 case 50: *profile_level_val =
613 compat_keys.kVTProfileLevel_H264_Baseline_5_0; break;
614 case 51: *profile_level_val =
615 compat_keys.kVTProfileLevel_H264_Baseline_5_1; break;
616 case 52: *profile_level_val =
617 compat_keys.kVTProfileLevel_H264_Baseline_5_2; break;
622 switch (vtctx->level) {
623 case 0: *profile_level_val =
624 compat_keys.kVTProfileLevel_H264_Main_AutoLevel; break;
625 case 30: *profile_level_val = kVTProfileLevel_H264_Main_3_0; break;
626 case 31: *profile_level_val = kVTProfileLevel_H264_Main_3_1; break;
627 case 32: *profile_level_val = kVTProfileLevel_H264_Main_3_2; break;
628 case 40: *profile_level_val = kVTProfileLevel_H264_Main_4_0; break;
629 case 41: *profile_level_val = kVTProfileLevel_H264_Main_4_1; break;
630 case 42: *profile_level_val =
631 compat_keys.kVTProfileLevel_H264_Main_4_2; break;
632 case 50: *profile_level_val = kVTProfileLevel_H264_Main_5_0; break;
633 case 51: *profile_level_val =
634 compat_keys.kVTProfileLevel_H264_Main_5_1; break;
635 case 52: *profile_level_val =
636 compat_keys.kVTProfileLevel_H264_Main_5_2; break;
641 switch (vtctx->level) {
642 case 0: *profile_level_val =
643 compat_keys.kVTProfileLevel_H264_High_AutoLevel; break;
644 case 30: *profile_level_val =
645 compat_keys.kVTProfileLevel_H264_High_3_0; break;
646 case 31: *profile_level_val =
647 compat_keys.kVTProfileLevel_H264_High_3_1; break;
648 case 32: *profile_level_val =
649 compat_keys.kVTProfileLevel_H264_High_3_2; break;
650 case 40: *profile_level_val =
651 compat_keys.kVTProfileLevel_H264_High_4_0; break;
652 case 41: *profile_level_val =
653 compat_keys.kVTProfileLevel_H264_High_4_1; break;
654 case 42: *profile_level_val =
655 compat_keys.kVTProfileLevel_H264_High_4_2; break;
656 case 50: *profile_level_val = kVTProfileLevel_H264_High_5_0; break;
657 case 51: *profile_level_val =
658 compat_keys.kVTProfileLevel_H264_High_5_1; break;
659 case 52: *profile_level_val =
660 compat_keys.kVTProfileLevel_H264_High_5_2; break;
665 if (!*profile_level_val) {
666 av_log(avctx, AV_LOG_ERROR, "Invalid Profile/Level.\n");
673 static int get_cv_pixel_format(AVCodecContext* avctx,
674 enum AVPixelFormat fmt,
675 enum AVColorRange range,
676 int* av_pixel_format,
679 if (range_guessed) *range_guessed = range != AVCOL_RANGE_MPEG &&
680 range != AVCOL_RANGE_JPEG;
682 //MPEG range is used when no range is set
683 if (fmt == AV_PIX_FMT_NV12) {
684 *av_pixel_format = range == AVCOL_RANGE_JPEG ?
685 kCVPixelFormatType_420YpCbCr8BiPlanarFullRange :
686 kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange;
687 } else if (fmt == AV_PIX_FMT_YUV420P) {
688 *av_pixel_format = range == AVCOL_RANGE_JPEG ?
689 kCVPixelFormatType_420YpCbCr8PlanarFullRange :
690 kCVPixelFormatType_420YpCbCr8Planar;
692 return AVERROR(EINVAL);
698 static void add_color_attr(AVCodecContext *avctx, CFMutableDictionaryRef dict) {
699 VTEncContext *vtctx = avctx->priv_data;
701 if (vtctx->color_primaries) {
702 CFDictionarySetValue(dict,
703 kCVImageBufferColorPrimariesKey,
704 vtctx->color_primaries);
707 if (vtctx->transfer_function) {
708 CFDictionarySetValue(dict,
709 kCVImageBufferTransferFunctionKey,
710 vtctx->transfer_function);
713 if (vtctx->ycbcr_matrix) {
714 CFDictionarySetValue(dict,
715 kCVImageBufferYCbCrMatrixKey,
716 vtctx->ycbcr_matrix);
720 static int create_cv_pixel_buffer_info(AVCodecContext* avctx,
721 CFMutableDictionaryRef* dict)
723 CFNumberRef cv_color_format_num = NULL;
724 CFNumberRef width_num = NULL;
725 CFNumberRef height_num = NULL;
726 CFMutableDictionaryRef pixel_buffer_info = NULL;
728 int status = get_cv_pixel_format(avctx,
733 if (status) return status;
735 pixel_buffer_info = CFDictionaryCreateMutable(
738 &kCFCopyStringDictionaryKeyCallBacks,
739 &kCFTypeDictionaryValueCallBacks);
741 if (!pixel_buffer_info) goto pbinfo_nomem;
743 cv_color_format_num = CFNumberCreate(kCFAllocatorDefault,
746 if (!cv_color_format_num) goto pbinfo_nomem;
748 CFDictionarySetValue(pixel_buffer_info,
749 kCVPixelBufferPixelFormatTypeKey,
750 cv_color_format_num);
751 vt_release_num(&cv_color_format_num);
753 width_num = CFNumberCreate(kCFAllocatorDefault,
756 if (!width_num) return AVERROR(ENOMEM);
758 CFDictionarySetValue(pixel_buffer_info,
759 kCVPixelBufferWidthKey,
761 vt_release_num(&width_num);
763 height_num = CFNumberCreate(kCFAllocatorDefault,
766 if (!height_num) goto pbinfo_nomem;
768 CFDictionarySetValue(pixel_buffer_info,
769 kCVPixelBufferHeightKey,
771 vt_release_num(&height_num);
773 add_color_attr(avctx, pixel_buffer_info);
775 *dict = pixel_buffer_info;
779 vt_release_num(&cv_color_format_num);
780 vt_release_num(&width_num);
781 vt_release_num(&height_num);
782 if (pixel_buffer_info) CFRelease(pixel_buffer_info);
784 return AVERROR(ENOMEM);
787 static int get_cv_color_primaries(AVCodecContext *avctx,
788 CFStringRef *primaries)
790 enum AVColorPrimaries pri = avctx->color_primaries;
792 case AVCOL_PRI_UNSPECIFIED:
796 case AVCOL_PRI_BT709:
797 *primaries = kCVImageBufferColorPrimaries_ITU_R_709_2;
800 case AVCOL_PRI_BT2020:
801 *primaries = compat_keys.kCVImageBufferColorPrimaries_ITU_R_2020;
805 av_log(avctx, AV_LOG_ERROR, "Color primaries %s is not supported.\n", av_color_primaries_name(pri));
813 static int get_cv_transfer_function(AVCodecContext *avctx,
814 CFStringRef *transfer_fnc,
815 CFNumberRef *gamma_level)
817 enum AVColorTransferCharacteristic trc = avctx->color_trc;
822 case AVCOL_TRC_UNSPECIFIED:
823 *transfer_fnc = NULL;
826 case AVCOL_TRC_BT709:
827 *transfer_fnc = kCVImageBufferTransferFunction_ITU_R_709_2;
830 case AVCOL_TRC_SMPTE240M:
831 *transfer_fnc = kCVImageBufferTransferFunction_SMPTE_240M_1995;
834 case AVCOL_TRC_GAMMA22:
836 *transfer_fnc = kCVImageBufferTransferFunction_UseGamma;
837 *gamma_level = CFNumberCreate(NULL, kCFNumberFloat32Type, &gamma);
840 case AVCOL_TRC_GAMMA28:
842 *transfer_fnc = kCVImageBufferTransferFunction_UseGamma;
843 *gamma_level = CFNumberCreate(NULL, kCFNumberFloat32Type, &gamma);
846 case AVCOL_TRC_BT2020_10:
847 case AVCOL_TRC_BT2020_12:
848 *transfer_fnc = compat_keys.kCVImageBufferTransferFunction_ITU_R_2020;
852 av_log(avctx, AV_LOG_ERROR, "Transfer function %s is not supported.\n", av_color_transfer_name(trc));
859 static int get_cv_ycbcr_matrix(AVCodecContext *avctx, CFStringRef *matrix) {
860 switch(avctx->colorspace) {
861 case AVCOL_SPC_BT709:
862 *matrix = kCVImageBufferYCbCrMatrix_ITU_R_709_2;
865 case AVCOL_SPC_UNSPECIFIED:
869 case AVCOL_SPC_BT470BG:
870 case AVCOL_SPC_SMPTE170M:
871 *matrix = kCVImageBufferYCbCrMatrix_ITU_R_601_4;
874 case AVCOL_SPC_SMPTE240M:
875 *matrix = kCVImageBufferYCbCrMatrix_SMPTE_240M_1995;
878 case AVCOL_SPC_BT2020_NCL:
879 *matrix = compat_keys.kCVImageBufferYCbCrMatrix_ITU_R_2020;
883 av_log(avctx, AV_LOG_ERROR, "Color space %s is not supported.\n", av_color_space_name(avctx->colorspace));
890 static int vtenc_create_encoder(AVCodecContext *avctx,
891 CMVideoCodecType codec_type,
892 CFStringRef profile_level,
893 CFNumberRef gamma_level,
894 CFDictionaryRef enc_info,
895 CFDictionaryRef pixel_buffer_info,
896 VTCompressionSessionRef *session)
898 VTEncContext *vtctx = avctx->priv_data;
899 SInt32 bit_rate = avctx->bit_rate;
900 SInt32 max_rate = avctx->rc_max_rate;
901 CFNumberRef bit_rate_num;
902 CFNumberRef bytes_per_second;
903 CFNumberRef one_second;
904 CFArrayRef data_rate_limits;
905 int64_t bytes_per_second_value = 0;
906 int64_t one_second_value = 0;
909 int status = VTCompressionSessionCreate(kCFAllocatorDefault,
916 vtenc_output_callback,
920 if (status || !vtctx->session) {
921 av_log(avctx, AV_LOG_ERROR, "Error: cannot create compression session: %d\n", status);
923 #if !TARGET_OS_IPHONE
924 if (!vtctx->allow_sw) {
925 av_log(avctx, AV_LOG_ERROR, "Try -allow_sw 1. The hardware encoder may be busy, or not supported.\n");
929 return AVERROR_EXTERNAL;
932 bit_rate_num = CFNumberCreate(kCFAllocatorDefault,
935 if (!bit_rate_num) return AVERROR(ENOMEM);
937 status = VTSessionSetProperty(vtctx->session,
938 kVTCompressionPropertyKey_AverageBitRate,
940 CFRelease(bit_rate_num);
943 av_log(avctx, AV_LOG_ERROR, "Error setting bitrate property: %d\n", status);
944 return AVERROR_EXTERNAL;
947 bytes_per_second_value = max_rate >> 3;
948 bytes_per_second = CFNumberCreate(kCFAllocatorDefault,
950 &bytes_per_second_value);
951 if (!bytes_per_second) {
952 return AVERROR(ENOMEM);
954 one_second_value = 1;
955 one_second = CFNumberCreate(kCFAllocatorDefault,
959 CFRelease(bytes_per_second);
960 return AVERROR(ENOMEM);
962 nums[0] = bytes_per_second;
963 nums[1] = one_second;
964 data_rate_limits = CFArrayCreate(kCFAllocatorDefault,
967 &kCFTypeArrayCallBacks);
969 if (!data_rate_limits) {
970 CFRelease(bytes_per_second);
971 CFRelease(one_second);
972 return AVERROR(ENOMEM);
974 status = VTSessionSetProperty(vtctx->session,
975 kVTCompressionPropertyKey_DataRateLimits,
978 CFRelease(bytes_per_second);
979 CFRelease(one_second);
980 CFRelease(data_rate_limits);
983 av_log(avctx, AV_LOG_ERROR, "Error setting max bitrate property: %d\n", status);
984 return AVERROR_EXTERNAL;
988 status = VTSessionSetProperty(vtctx->session,
989 kVTCompressionPropertyKey_ProfileLevel,
992 av_log(avctx, AV_LOG_ERROR, "Error setting profile/level property: %d\n", status);
996 if (avctx->gop_size > 0) {
997 CFNumberRef interval = CFNumberCreate(kCFAllocatorDefault,
1001 return AVERROR(ENOMEM);
1004 status = VTSessionSetProperty(vtctx->session,
1005 kVTCompressionPropertyKey_MaxKeyFrameInterval,
1007 CFRelease(interval);
1010 av_log(avctx, AV_LOG_ERROR, "Error setting 'max key-frame interval' property: %d\n", status);
1011 return AVERROR_EXTERNAL;
1015 if (vtctx->frames_before) {
1016 status = VTSessionSetProperty(vtctx->session,
1017 kVTCompressionPropertyKey_MoreFramesBeforeStart,
1020 if (status == kVTPropertyNotSupportedErr) {
1021 av_log(avctx, AV_LOG_WARNING, "frames_before property is not supported on this device. Ignoring.\n");
1022 } else if (status) {
1023 av_log(avctx, AV_LOG_ERROR, "Error setting frames_before property: %d\n", status);
1027 if (vtctx->frames_after) {
1028 status = VTSessionSetProperty(vtctx->session,
1029 kVTCompressionPropertyKey_MoreFramesAfterEnd,
1032 if (status == kVTPropertyNotSupportedErr) {
1033 av_log(avctx, AV_LOG_WARNING, "frames_after property is not supported on this device. Ignoring.\n");
1034 } else if (status) {
1035 av_log(avctx, AV_LOG_ERROR, "Error setting frames_after property: %d\n", status);
1039 if (avctx->sample_aspect_ratio.num != 0) {
1042 CFMutableDictionaryRef par;
1043 AVRational *avpar = &avctx->sample_aspect_ratio;
1045 av_reduce(&avpar->num, &avpar->den,
1046 avpar->num, avpar->den,
1049 num = CFNumberCreate(kCFAllocatorDefault,
1053 den = CFNumberCreate(kCFAllocatorDefault,
1059 par = CFDictionaryCreateMutable(kCFAllocatorDefault,
1061 &kCFCopyStringDictionaryKeyCallBacks,
1062 &kCFTypeDictionaryValueCallBacks);
1064 if (!par || !num || !den) {
1065 if (par) CFRelease(par);
1066 if (num) CFRelease(num);
1067 if (den) CFRelease(den);
1069 return AVERROR(ENOMEM);
1072 CFDictionarySetValue(
1074 kCMFormatDescriptionKey_PixelAspectRatioHorizontalSpacing,
1077 CFDictionarySetValue(
1079 kCMFormatDescriptionKey_PixelAspectRatioVerticalSpacing,
1082 status = VTSessionSetProperty(vtctx->session,
1083 kVTCompressionPropertyKey_PixelAspectRatio,
1093 "Error setting pixel aspect ratio to %d:%d: %d.\n",
1094 avctx->sample_aspect_ratio.num,
1095 avctx->sample_aspect_ratio.den,
1098 return AVERROR_EXTERNAL;
1103 if (vtctx->transfer_function) {
1104 status = VTSessionSetProperty(vtctx->session,
1105 kVTCompressionPropertyKey_TransferFunction,
1106 vtctx->transfer_function);
1109 av_log(avctx, AV_LOG_WARNING, "Could not set transfer function: %d\n", status);
1114 if (vtctx->ycbcr_matrix) {
1115 status = VTSessionSetProperty(vtctx->session,
1116 kVTCompressionPropertyKey_YCbCrMatrix,
1117 vtctx->ycbcr_matrix);
1120 av_log(avctx, AV_LOG_WARNING, "Could not set ycbcr matrix: %d\n", status);
1125 if (vtctx->color_primaries) {
1126 status = VTSessionSetProperty(vtctx->session,
1127 kVTCompressionPropertyKey_ColorPrimaries,
1128 vtctx->color_primaries);
1131 av_log(avctx, AV_LOG_WARNING, "Could not set color primaries: %d\n", status);
1136 status = VTSessionSetProperty(vtctx->session,
1137 kCVImageBufferGammaLevelKey,
1141 av_log(avctx, AV_LOG_WARNING, "Could not set gamma level: %d\n", status);
1145 if (!vtctx->has_b_frames) {
1146 status = VTSessionSetProperty(vtctx->session,
1147 kVTCompressionPropertyKey_AllowFrameReordering,
1151 av_log(avctx, AV_LOG_ERROR, "Error setting 'allow frame reordering' property: %d\n", status);
1152 return AVERROR_EXTERNAL;
1156 if (vtctx->entropy != VT_ENTROPY_NOT_SET) {
1157 CFStringRef entropy = vtctx->entropy == VT_CABAC ?
1158 compat_keys.kVTH264EntropyMode_CABAC:
1159 compat_keys.kVTH264EntropyMode_CAVLC;
1161 status = VTSessionSetProperty(vtctx->session,
1162 compat_keys.kVTCompressionPropertyKey_H264EntropyMode,
1166 av_log(avctx, AV_LOG_ERROR, "Error setting entropy property: %d\n", status);
1170 if (vtctx->realtime) {
1171 status = VTSessionSetProperty(vtctx->session,
1172 compat_keys.kVTCompressionPropertyKey_RealTime,
1176 av_log(avctx, AV_LOG_ERROR, "Error setting realtime property: %d\n", status);
1180 status = VTCompressionSessionPrepareToEncodeFrames(vtctx->session);
1182 av_log(avctx, AV_LOG_ERROR, "Error: cannot prepare encoder: %d\n", status);
1183 return AVERROR_EXTERNAL;
1189 static av_cold int vtenc_init(AVCodecContext *avctx)
1191 CFMutableDictionaryRef enc_info;
1192 CFMutableDictionaryRef pixel_buffer_info;
1193 CMVideoCodecType codec_type;
1194 VTEncContext *vtctx = avctx->priv_data;
1195 CFStringRef profile_level;
1196 CFBooleanRef has_b_frames_cfbool;
1197 CFNumberRef gamma_level = NULL;
1200 pthread_once(&once_ctrl, loadVTEncSymbols);
1202 codec_type = get_cm_codec_type(avctx->codec_id);
1204 av_log(avctx, AV_LOG_ERROR, "Error: no mapping for AVCodecID %d\n", avctx->codec_id);
1205 return AVERROR(EINVAL);
1208 vtctx->has_b_frames = avctx->max_b_frames > 0;
1209 if(vtctx->has_b_frames && vtctx->profile == H264_PROF_BASELINE){
1210 av_log(avctx, AV_LOG_WARNING, "Cannot use B-frames with baseline profile. Output will not contain B-frames.\n");
1211 vtctx->has_b_frames = false;
1214 if (vtctx->entropy == VT_CABAC && vtctx->profile == H264_PROF_BASELINE) {
1215 av_log(avctx, AV_LOG_WARNING, "CABAC entropy requires 'main' or 'high' profile, but baseline was requested. Encode will not use CABAC entropy.\n");
1216 vtctx->entropy = VT_ENTROPY_NOT_SET;
1219 if (!get_vt_profile_level(avctx, &profile_level)) return AVERROR(EINVAL);
1221 vtctx->session = NULL;
1223 enc_info = CFDictionaryCreateMutable(
1224 kCFAllocatorDefault,
1226 &kCFCopyStringDictionaryKeyCallBacks,
1227 &kCFTypeDictionaryValueCallBacks
1230 if (!enc_info) return AVERROR(ENOMEM);
1232 #if !TARGET_OS_IPHONE
1233 if (!vtctx->allow_sw) {
1234 CFDictionarySetValue(enc_info,
1235 compat_keys.kVTVideoEncoderSpecification_RequireHardwareAcceleratedVideoEncoder,
1238 CFDictionarySetValue(enc_info,
1239 compat_keys.kVTVideoEncoderSpecification_EnableHardwareAcceleratedVideoEncoder,
1244 if (avctx->pix_fmt != AV_PIX_FMT_VIDEOTOOLBOX) {
1245 status = create_cv_pixel_buffer_info(avctx, &pixel_buffer_info);
1249 pixel_buffer_info = NULL;
1252 pthread_mutex_init(&vtctx->lock, NULL);
1253 pthread_cond_init(&vtctx->cv_sample_sent, NULL);
1254 vtctx->dts_delta = vtctx->has_b_frames ? -1 : 0;
1256 get_cv_transfer_function(avctx, &vtctx->transfer_function, &gamma_level);
1257 get_cv_ycbcr_matrix(avctx, &vtctx->ycbcr_matrix);
1258 get_cv_color_primaries(avctx, &vtctx->color_primaries);
1261 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
1262 status = vtenc_populate_extradata(avctx,
1272 status = vtenc_create_encoder(avctx,
1283 status = VTSessionCopyProperty(vtctx->session,
1284 kVTCompressionPropertyKey_AllowFrameReordering,
1285 kCFAllocatorDefault,
1286 &has_b_frames_cfbool);
1289 //Some devices don't output B-frames for main profile, even if requested.
1290 vtctx->has_b_frames = CFBooleanGetValue(has_b_frames_cfbool);
1291 CFRelease(has_b_frames_cfbool);
1293 avctx->has_b_frames = vtctx->has_b_frames;
1297 CFRelease(gamma_level);
1299 if (pixel_buffer_info)
1300 CFRelease(pixel_buffer_info);
1302 CFRelease(enc_info);
1307 static void vtenc_get_frame_info(CMSampleBufferRef buffer, bool *is_key_frame)
1309 CFArrayRef attachments;
1310 CFDictionaryRef attachment;
1311 CFBooleanRef not_sync;
1314 attachments = CMSampleBufferGetSampleAttachmentsArray(buffer, false);
1315 len = !attachments ? 0 : CFArrayGetCount(attachments);
1318 *is_key_frame = true;
1322 attachment = CFArrayGetValueAtIndex(attachments, 0);
1324 if (CFDictionaryGetValueIfPresent(attachment,
1325 kCMSampleAttachmentKey_NotSync,
1326 (const void **)¬_sync))
1328 *is_key_frame = !CFBooleanGetValue(not_sync);
1330 *is_key_frame = true;
1334 static int is_post_sei_nal_type(int nal_type){
1335 return nal_type != H264_NAL_SEI &&
1336 nal_type != H264_NAL_SPS &&
1337 nal_type != H264_NAL_PPS &&
1338 nal_type != H264_NAL_AUD;
1342 * Finds the sei message start/size of type find_sei_type.
1343 * If more than one of that type exists, the last one is returned.
1345 static int find_sei_end(AVCodecContext *avctx,
1351 size_t sei_payload_size = 0;
1352 int sei_payload_type = 0;
1354 uint8_t *nal_start = nal_data;
1359 nal_type = *nal_data & 0x1F;
1360 if (nal_type != H264_NAL_SEI)
1366 if (nal_data[nal_size - 1] == 0x80)
1369 while (nal_size > 0 && *nal_data > 0) {
1371 sei_payload_type += *nal_data;
1374 } while (nal_size > 0 && *nal_data == 0xFF);
1377 av_log(avctx, AV_LOG_ERROR, "Unexpected end of SEI NAL Unit parsing type.\n");
1378 return AVERROR_INVALIDDATA;
1382 sei_payload_size += *nal_data;
1385 } while (nal_size > 0 && *nal_data == 0xFF);
1387 if (nal_size < sei_payload_size) {
1388 av_log(avctx, AV_LOG_ERROR, "Unexpected end of SEI NAL Unit parsing size.\n");
1389 return AVERROR_INVALIDDATA;
1392 nal_data += sei_payload_size;
1393 nal_size -= sei_payload_size;
1396 *sei_end = nal_data;
1398 return nal_data - nal_start + 1;
1402 * Copies the data inserting emulation prevention bytes as needed.
1403 * Existing data in the destination can be taken into account by providing
1404 * dst with a dst_offset > 0.
1406 * @return The number of bytes copied on success. On failure, the negative of
1407 * the number of bytes needed to copy src is returned.
1409 static int copy_emulation_prev(const uint8_t *src,
1418 uint8_t* dst_end = dst + dst_size;
1419 const uint8_t* src_end = src + src_size;
1420 int start_at = dst_offset > 2 ? dst_offset - 2 : 0;
1422 for (i = start_at; i < dst_offset && i < dst_size; i++) {
1431 for (; src < src_end; src++, dst++) {
1433 int insert_ep3_byte = *src <= 3;
1434 if (insert_ep3_byte) {
1452 wrote_bytes = dst - dst_start;
1455 return -wrote_bytes;
1460 static int write_sei(const ExtraSEI *sei,
1465 uint8_t *sei_start = dst;
1466 size_t remaining_sei_size = sei->size;
1467 size_t remaining_dst_size = dst_size;
1472 if (!remaining_dst_size)
1473 return AVERROR_BUFFER_TOO_SMALL;
1475 while (sei_type && remaining_dst_size != 0) {
1476 int sei_byte = sei_type > 255 ? 255 : sei_type;
1479 sei_type -= sei_byte;
1481 remaining_dst_size--;
1485 return AVERROR_BUFFER_TOO_SMALL;
1487 while (remaining_sei_size && remaining_dst_size != 0) {
1488 int size_byte = remaining_sei_size > 255 ? 255 : remaining_sei_size;
1491 remaining_sei_size -= size_byte;
1493 remaining_dst_size--;
1496 if (remaining_dst_size < sei->size)
1497 return AVERROR_BUFFER_TOO_SMALL;
1499 header_bytes = dst - sei_start;
1501 offset = header_bytes;
1502 bytes_written = copy_emulation_prev(sei->data,
1507 if (bytes_written < 0)
1508 return AVERROR_BUFFER_TOO_SMALL;
1510 bytes_written += header_bytes;
1511 return bytes_written;
1515 * Copies NAL units and replaces length codes with
1516 * H.264 Annex B start codes. On failure, the contents of
1517 * dst_data may have been modified.
1519 * @param length_code_size Byte length of each length code
1520 * @param sample_buffer NAL units prefixed with length codes.
1521 * @param sei Optional A53 closed captions SEI data.
1522 * @param dst_data Must be zeroed before calling this function.
1523 * Contains the copied NAL units prefixed with
1524 * start codes when the function returns
1526 * @param dst_size Length of dst_data
1527 * @return 0 on success
1528 * AVERROR_INVALIDDATA if length_code_size is invalid
1529 * AVERROR_BUFFER_TOO_SMALL if dst_data is too small
1530 * or if a length_code in src_data specifies data beyond
1531 * the end of its buffer.
1533 static int copy_replace_length_codes(
1534 AVCodecContext *avctx,
1535 size_t length_code_size,
1536 CMSampleBufferRef sample_buffer,
1541 size_t src_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
1542 size_t remaining_src_size = src_size;
1543 size_t remaining_dst_size = dst_size;
1544 size_t src_offset = 0;
1547 uint8_t size_buf[4];
1549 CMBlockBufferRef block = CMSampleBufferGetDataBuffer(sample_buffer);
1551 if (length_code_size > 4) {
1552 return AVERROR_INVALIDDATA;
1555 while (remaining_src_size > 0) {
1556 size_t curr_src_len;
1557 size_t curr_dst_len;
1563 status = CMBlockBufferCopyDataBytes(block,
1568 av_log(avctx, AV_LOG_ERROR, "Cannot copy length: %d\n", status);
1569 return AVERROR_EXTERNAL;
1572 status = CMBlockBufferCopyDataBytes(block,
1573 src_offset + length_code_size,
1578 av_log(avctx, AV_LOG_ERROR, "Cannot copy type: %d\n", status);
1579 return AVERROR_EXTERNAL;
1584 for (i = 0; i < length_code_size; i++) {
1586 box_len |= size_buf[i];
1589 if (sei && !wrote_sei && is_post_sei_nal_type(nal_type)) {
1590 //No SEI NAL unit - insert.
1593 memcpy(dst_data, start_code, sizeof(start_code));
1594 dst_data += sizeof(start_code);
1595 remaining_dst_size -= sizeof(start_code);
1597 *dst_data = H264_NAL_SEI;
1599 remaining_dst_size--;
1601 wrote_bytes = write_sei(sei,
1602 SEI_TYPE_USER_DATA_REGISTERED,
1604 remaining_dst_size);
1606 if (wrote_bytes < 0)
1609 remaining_dst_size -= wrote_bytes;
1610 dst_data += wrote_bytes;
1612 if (remaining_dst_size <= 0)
1613 return AVERROR_BUFFER_TOO_SMALL;
1618 remaining_dst_size--;
1623 curr_src_len = box_len + length_code_size;
1624 curr_dst_len = box_len + sizeof(start_code);
1626 if (remaining_src_size < curr_src_len) {
1627 return AVERROR_BUFFER_TOO_SMALL;
1630 if (remaining_dst_size < curr_dst_len) {
1631 return AVERROR_BUFFER_TOO_SMALL;
1634 dst_box = dst_data + sizeof(start_code);
1636 memcpy(dst_data, start_code, sizeof(start_code));
1637 status = CMBlockBufferCopyDataBytes(block,
1638 src_offset + length_code_size,
1643 av_log(avctx, AV_LOG_ERROR, "Cannot copy data: %d\n", status);
1644 return AVERROR_EXTERNAL;
1647 if (sei && !wrote_sei && nal_type == H264_NAL_SEI) {
1648 //Found SEI NAL unit - append.
1653 old_sei_length = find_sei_end(avctx, dst_box, box_len, &new_sei);
1654 if (old_sei_length < 0)
1657 wrote_bytes = write_sei(sei,
1658 SEI_TYPE_USER_DATA_REGISTERED,
1660 remaining_dst_size - old_sei_length);
1661 if (wrote_bytes < 0)
1664 if (new_sei + wrote_bytes >= dst_data + remaining_dst_size)
1665 return AVERROR_BUFFER_TOO_SMALL;
1667 new_sei[wrote_bytes++] = 0x80;
1668 extra_bytes = wrote_bytes - (dst_box + box_len - new_sei);
1670 dst_data += extra_bytes;
1671 remaining_dst_size -= extra_bytes;
1676 src_offset += curr_src_len;
1677 dst_data += curr_dst_len;
1679 remaining_src_size -= curr_src_len;
1680 remaining_dst_size -= curr_dst_len;
1687 * Returns a sufficient number of bytes to contain the sei data.
1688 * It may be greater than the minimum required.
1690 static int get_sei_msg_bytes(const ExtraSEI* sei, int type){
1695 copied_size = -copy_emulation_prev(sei->data,
1701 if ((sei->size % 255) == 0) //may result in an extra byte
1704 return copied_size + sei->size / 255 + 1 + type / 255 + 1;
1707 static int vtenc_cm_to_avpacket(
1708 AVCodecContext *avctx,
1709 CMSampleBufferRef sample_buffer,
1713 VTEncContext *vtctx = avctx->priv_data;
1718 size_t length_code_size;
1719 size_t header_size = 0;
1721 size_t out_buf_size;
1722 size_t sei_nalu_size = 0;
1724 int64_t time_base_num;
1728 CMVideoFormatDescriptionRef vid_fmt;
1731 vtenc_get_frame_info(sample_buffer, &is_key_frame);
1732 status = get_length_code_size(avctx, sample_buffer, &length_code_size);
1733 if (status) return status;
1735 add_header = is_key_frame && !(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER);
1738 vid_fmt = CMSampleBufferGetFormatDescription(sample_buffer);
1740 av_log(avctx, AV_LOG_ERROR, "Cannot get format description.\n");
1741 return AVERROR_EXTERNAL;
1744 int status = get_params_size(avctx, vid_fmt, &header_size);
1745 if (status) return status;
1748 status = count_nalus(length_code_size, sample_buffer, &nalu_count);
1753 size_t msg_size = get_sei_msg_bytes(sei,
1754 SEI_TYPE_USER_DATA_REGISTERED);
1756 sei_nalu_size = sizeof(start_code) + 1 + msg_size + 1;
1759 in_buf_size = CMSampleBufferGetTotalSampleSize(sample_buffer);
1760 out_buf_size = header_size +
1763 nalu_count * ((int)sizeof(start_code) - (int)length_code_size);
1765 status = ff_alloc_packet2(avctx, pkt, out_buf_size, out_buf_size);
1770 status = copy_param_sets(avctx, vid_fmt, pkt->data, out_buf_size);
1771 if(status) return status;
1774 status = copy_replace_length_codes(
1779 pkt->data + header_size,
1780 pkt->size - header_size
1784 av_log(avctx, AV_LOG_ERROR, "Error copying packet data: %d\n", status);
1789 pkt->flags |= AV_PKT_FLAG_KEY;
1792 pts = CMSampleBufferGetPresentationTimeStamp(sample_buffer);
1793 dts = CMSampleBufferGetDecodeTimeStamp (sample_buffer);
1795 if (CMTIME_IS_INVALID(dts)) {
1796 if (!vtctx->has_b_frames) {
1799 av_log(avctx, AV_LOG_ERROR, "DTS is invalid.\n");
1800 return AVERROR_EXTERNAL;
1804 dts_delta = vtctx->dts_delta >= 0 ? vtctx->dts_delta : 0;
1805 time_base_num = avctx->time_base.num;
1806 pkt->pts = pts.value / time_base_num;
1807 pkt->dts = dts.value / time_base_num - dts_delta;
1808 pkt->size = out_buf_size;
1814 * contiguous_buf_size is 0 if not contiguous, and the size of the buffer
1815 * containing all planes if so.
1817 static int get_cv_pixel_info(
1818 AVCodecContext *avctx,
1819 const AVFrame *frame,
1825 size_t *contiguous_buf_size)
1827 VTEncContext *vtctx = avctx->priv_data;
1828 int av_format = frame->format;
1829 int av_color_range = av_frame_get_color_range(frame);
1834 status = get_cv_pixel_format(avctx, av_format, av_color_range, color, &range_guessed);
1838 "Could not get pixel format for color format '%s' range '%s'.\n",
1839 av_get_pix_fmt_name(av_format),
1840 av_color_range > AVCOL_RANGE_UNSPECIFIED &&
1841 av_color_range < AVCOL_RANGE_NB ?
1842 av_color_range_name(av_color_range) :
1845 return AVERROR(EINVAL);
1848 if (range_guessed) {
1849 if (!vtctx->warned_color_range) {
1850 vtctx->warned_color_range = true;
1853 "Color range not set for %s. Using MPEG range.\n",
1854 av_get_pix_fmt_name(av_format));
1857 av_log(avctx, AV_LOG_WARNING, "");
1860 switch (av_format) {
1861 case AV_PIX_FMT_NV12:
1864 widths [0] = avctx->width;
1865 heights[0] = avctx->height;
1866 strides[0] = frame ? frame->linesize[0] : avctx->width;
1868 widths [1] = (avctx->width + 1) / 2;
1869 heights[1] = (avctx->height + 1) / 2;
1870 strides[1] = frame ? frame->linesize[1] : (avctx->width + 1) & -2;
1873 case AV_PIX_FMT_YUV420P:
1876 widths [0] = avctx->width;
1877 heights[0] = avctx->height;
1878 strides[0] = frame ? frame->linesize[0] : avctx->width;
1880 widths [1] = (avctx->width + 1) / 2;
1881 heights[1] = (avctx->height + 1) / 2;
1882 strides[1] = frame ? frame->linesize[1] : (avctx->width + 1) / 2;
1884 widths [2] = (avctx->width + 1) / 2;
1885 heights[2] = (avctx->height + 1) / 2;
1886 strides[2] = frame ? frame->linesize[2] : (avctx->width + 1) / 2;
1893 "Could not get frame format info for color %d range %d.\n",
1897 return AVERROR(EINVAL);
1900 *contiguous_buf_size = 0;
1901 for (i = 0; i < *plane_count; i++) {
1902 if (i < *plane_count - 1 &&
1903 frame->data[i] + strides[i] * heights[i] != frame->data[i + 1]) {
1904 *contiguous_buf_size = 0;
1908 *contiguous_buf_size += strides[i] * heights[i];
1914 #if !TARGET_OS_IPHONE
1915 //Not used on iOS - frame is always copied.
1916 static void free_avframe(
1921 const void *plane_addresses[])
1923 AVFrame *frame = release_ctx;
1924 av_frame_free(&frame);
1927 //Not used on OSX - frame is never copied.
1928 static int copy_avframe_to_pixel_buffer(AVCodecContext *avctx,
1929 const AVFrame *frame,
1930 CVPixelBufferRef cv_img,
1931 const size_t *plane_strides,
1932 const size_t *plane_rows)
1944 status = CVPixelBufferLockBaseAddress(cv_img, 0);
1949 "Error: Could not lock base address of CVPixelBuffer: %d.\n",
1954 if (CVPixelBufferIsPlanar(cv_img)) {
1955 plane_count = CVPixelBufferGetPlaneCount(cv_img);
1956 for (i = 0; frame->data[i]; i++) {
1957 if (i == plane_count) {
1958 CVPixelBufferUnlockBaseAddress(cv_img, 0);
1961 "Error: different number of planes in AVFrame and CVPixelBuffer.\n"
1964 return AVERROR_EXTERNAL;
1967 dst_addr = (uint8_t*)CVPixelBufferGetBaseAddressOfPlane(cv_img, i);
1968 src_addr = (uint8_t*)frame->data[i];
1969 dst_stride = CVPixelBufferGetBytesPerRowOfPlane(cv_img, i);
1970 src_stride = plane_strides[i];
1971 rows = plane_rows[i];
1973 if (dst_stride == src_stride) {
1974 memcpy(dst_addr, src_addr, src_stride * rows);
1976 copy_bytes = dst_stride < src_stride ? dst_stride : src_stride;
1978 for (j = 0; j < rows; j++) {
1979 memcpy(dst_addr + j * dst_stride, src_addr + j * src_stride, copy_bytes);
1984 if (frame->data[1]) {
1985 CVPixelBufferUnlockBaseAddress(cv_img, 0);
1988 "Error: different number of planes in AVFrame and non-planar CVPixelBuffer.\n"
1991 return AVERROR_EXTERNAL;
1994 dst_addr = (uint8_t*)CVPixelBufferGetBaseAddress(cv_img);
1995 src_addr = (uint8_t*)frame->data[0];
1996 dst_stride = CVPixelBufferGetBytesPerRow(cv_img);
1997 src_stride = plane_strides[0];
1998 rows = plane_rows[0];
2000 if (dst_stride == src_stride) {
2001 memcpy(dst_addr, src_addr, src_stride * rows);
2003 copy_bytes = dst_stride < src_stride ? dst_stride : src_stride;
2005 for (j = 0; j < rows; j++) {
2006 memcpy(dst_addr + j * dst_stride, src_addr + j * src_stride, copy_bytes);
2011 status = CVPixelBufferUnlockBaseAddress(cv_img, 0);
2013 av_log(avctx, AV_LOG_ERROR, "Error: Could not unlock CVPixelBuffer base address: %d.\n", status);
2014 return AVERROR_EXTERNAL;
2019 #endif //!TARGET_OS_IPHONE
2021 static int create_cv_pixel_buffer(AVCodecContext *avctx,
2022 const AVFrame *frame,
2023 CVPixelBufferRef *cv_img)
2027 size_t widths [AV_NUM_DATA_POINTERS];
2028 size_t heights[AV_NUM_DATA_POINTERS];
2029 size_t strides[AV_NUM_DATA_POINTERS];
2031 size_t contiguous_buf_size;
2032 #if TARGET_OS_IPHONE
2033 CVPixelBufferPoolRef pix_buf_pool;
2034 VTEncContext* vtctx = avctx->priv_data;
2036 CFMutableDictionaryRef pix_buf_attachments = CFDictionaryCreateMutable(
2037 kCFAllocatorDefault,
2039 &kCFCopyStringDictionaryKeyCallBacks,
2040 &kCFTypeDictionaryValueCallBacks);
2042 if (!pix_buf_attachments) return AVERROR(ENOMEM);
2045 if (avctx->pix_fmt == AV_PIX_FMT_VIDEOTOOLBOX) {
2046 av_assert0(frame->format == AV_PIX_FMT_VIDEOTOOLBOX);
2048 *cv_img = (CVPixelBufferRef)frame->data[3];
2049 av_assert0(*cv_img);
2055 memset(widths, 0, sizeof(widths));
2056 memset(heights, 0, sizeof(heights));
2057 memset(strides, 0, sizeof(strides));
2059 status = get_cv_pixel_info(
2067 &contiguous_buf_size
2074 "Error: Cannot convert format %d color_range %d: %d\n",
2076 av_frame_get_color_range(frame),
2080 return AVERROR_EXTERNAL;
2083 #if TARGET_OS_IPHONE
2084 pix_buf_pool = VTCompressionSessionGetPixelBufferPool(vtctx->session);
2085 if (!pix_buf_pool) {
2086 av_log(avctx, AV_LOG_ERROR, "Could not get pixel buffer pool.\n");
2087 return AVERROR_EXTERNAL;
2090 status = CVPixelBufferPoolCreatePixelBuffer(NULL,
2096 av_log(avctx, AV_LOG_ERROR, "Could not create pixel buffer from pool: %d.\n", status);
2097 return AVERROR_EXTERNAL;
2100 status = copy_avframe_to_pixel_buffer(avctx, frame, *cv_img, strides, heights);
2107 AVFrame *enc_frame = av_frame_alloc();
2108 if (!enc_frame) return AVERROR(ENOMEM);
2110 status = av_frame_ref(enc_frame, frame);
2112 av_frame_free(&enc_frame);
2116 status = CVPixelBufferCreateWithPlanarBytes(
2117 kCFAllocatorDefault,
2122 contiguous_buf_size,
2124 (void **)enc_frame->data,
2134 add_color_attr(avctx, pix_buf_attachments);
2135 CVBufferSetAttachments(*cv_img, pix_buf_attachments, kCVAttachmentMode_ShouldPropagate);
2136 CFRelease(pix_buf_attachments);
2139 av_log(avctx, AV_LOG_ERROR, "Error: Could not create CVPixelBuffer: %d\n", status);
2140 return AVERROR_EXTERNAL;
2147 static int create_encoder_dict_h264(const AVFrame *frame,
2148 CFDictionaryRef* dict_out)
2150 CFDictionaryRef dict = NULL;
2151 if (frame->pict_type == AV_PICTURE_TYPE_I) {
2152 const void *keys[] = { kVTEncodeFrameOptionKey_ForceKeyFrame };
2153 const void *vals[] = { kCFBooleanTrue };
2155 dict = CFDictionaryCreate(NULL, keys, vals, 1, NULL, NULL);
2156 if(!dict) return AVERROR(ENOMEM);
2163 static int vtenc_send_frame(AVCodecContext *avctx,
2164 VTEncContext *vtctx,
2165 const AVFrame *frame)
2168 CFDictionaryRef frame_dict;
2169 CVPixelBufferRef cv_img = NULL;
2170 AVFrameSideData *side_data = NULL;
2171 ExtraSEI *sei = NULL;
2172 int status = create_cv_pixel_buffer(avctx, frame, &cv_img);
2174 if (status) return status;
2176 status = create_encoder_dict_h264(frame, &frame_dict);
2182 side_data = av_frame_get_side_data(frame, AV_FRAME_DATA_A53_CC);
2183 if (vtctx->a53_cc && side_data && side_data->size) {
2184 sei = av_mallocz(sizeof(*sei));
2186 av_log(avctx, AV_LOG_ERROR, "Not enough memory for closed captions, skipping\n");
2188 int ret = ff_alloc_a53_sei(frame, 0, &sei->data, &sei->size);
2190 av_log(avctx, AV_LOG_ERROR, "Not enough memory for closed captions, skipping\n");
2197 time = CMTimeMake(frame->pts * avctx->time_base.num, avctx->time_base.den);
2198 status = VTCompressionSessionEncodeFrame(
2208 if (frame_dict) CFRelease(frame_dict);
2212 av_log(avctx, AV_LOG_ERROR, "Error: cannot encode frame: %d\n", status);
2213 return AVERROR_EXTERNAL;
2219 static av_cold int vtenc_frame(
2220 AVCodecContext *avctx,
2222 const AVFrame *frame,
2225 VTEncContext *vtctx = avctx->priv_data;
2228 CMSampleBufferRef buf = NULL;
2229 ExtraSEI *sei = NULL;
2232 status = vtenc_send_frame(avctx, vtctx, frame);
2235 status = AVERROR_EXTERNAL;
2239 if (vtctx->frame_ct_in == 0) {
2240 vtctx->first_pts = frame->pts;
2241 } else if(vtctx->frame_ct_in == 1 && vtctx->has_b_frames) {
2242 vtctx->dts_delta = frame->pts - vtctx->first_pts;
2245 vtctx->frame_ct_in++;
2246 } else if(!vtctx->flushing) {
2247 vtctx->flushing = true;
2249 status = VTCompressionSessionCompleteFrames(vtctx->session,
2253 av_log(avctx, AV_LOG_ERROR, "Error flushing frames: %d\n", status);
2254 status = AVERROR_EXTERNAL;
2260 get_frame = vtctx->dts_delta >= 0 || !frame;
2266 status = vtenc_q_pop(vtctx, !frame, &buf, &sei);
2267 if (status) goto end_nopkt;
2268 if (!buf) goto end_nopkt;
2270 status = vtenc_cm_to_avpacket(avctx, buf, pkt, sei);
2272 if (sei->data) av_free(sei->data);
2276 if (status) goto end_nopkt;
2282 av_packet_unref(pkt);
2286 static int vtenc_populate_extradata(AVCodecContext *avctx,
2287 CMVideoCodecType codec_type,
2288 CFStringRef profile_level,
2289 CFNumberRef gamma_level,
2290 CFDictionaryRef enc_info,
2291 CFDictionaryRef pixel_buffer_info)
2293 VTEncContext *vtctx = avctx->priv_data;
2294 AVFrame *frame = av_frame_alloc();
2295 int y_size = avctx->width * avctx->height;
2296 int chroma_size = (avctx->width / 2) * (avctx->height / 2);
2297 CMSampleBufferRef buf = NULL;
2301 return AVERROR(ENOMEM);
2303 frame->buf[0] = av_buffer_alloc(y_size + 2 * chroma_size);
2306 status = AVERROR(ENOMEM);
2310 status = vtenc_create_encoder(avctx,
2320 frame->data[0] = frame->buf[0]->data;
2321 memset(frame->data[0], 0, y_size);
2323 frame->data[1] = frame->buf[0]->data + y_size;
2324 memset(frame->data[1], 128, chroma_size);
2327 if (avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
2328 frame->data[2] = frame->buf[0]->data + y_size + chroma_size;
2329 memset(frame->data[2], 128, chroma_size);
2332 frame->linesize[0] = avctx->width;
2334 if (avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
2335 frame->linesize[1] =
2336 frame->linesize[2] = (avctx->width + 1) / 2;
2338 frame->linesize[1] = (avctx->width + 1) / 2;
2341 frame->format = avctx->pix_fmt;
2342 frame->width = avctx->width;
2343 frame->height = avctx->height;
2344 av_frame_set_colorspace(frame, avctx->colorspace);
2345 av_frame_set_color_range(frame, avctx->color_range);
2346 frame->color_trc = avctx->color_trc;
2347 frame->color_primaries = avctx->color_primaries;
2350 status = vtenc_send_frame(avctx, vtctx, frame);
2352 av_log(avctx, AV_LOG_ERROR, "Error sending frame: %d\n", status);
2356 //Populates extradata - output frames are flushed and param sets are available.
2357 status = VTCompressionSessionCompleteFrames(vtctx->session,
2363 status = vtenc_q_pop(vtctx, 0, &buf, NULL);
2365 av_log(avctx, AV_LOG_ERROR, "popping: %d\n", status);
2375 CFRelease(vtctx->session);
2377 vtctx->session = NULL;
2378 vtctx->frame_ct_out = 0;
2380 av_frame_unref(frame);
2381 av_frame_free(&frame);
2383 av_assert0(status != 0 || (avctx->extradata && avctx->extradata_size > 0));
2388 static av_cold int vtenc_close(AVCodecContext *avctx)
2390 VTEncContext *vtctx = avctx->priv_data;
2392 if(!vtctx->session) return 0;
2394 VTCompressionSessionCompleteFrames(vtctx->session,
2396 clear_frame_queue(vtctx);
2397 pthread_cond_destroy(&vtctx->cv_sample_sent);
2398 pthread_mutex_destroy(&vtctx->lock);
2399 CFRelease(vtctx->session);
2400 vtctx->session = NULL;
2402 if (vtctx->color_primaries) {
2403 CFRelease(vtctx->color_primaries);
2404 vtctx->color_primaries = NULL;
2407 if (vtctx->transfer_function) {
2408 CFRelease(vtctx->transfer_function);
2409 vtctx->transfer_function = NULL;
2412 if (vtctx->ycbcr_matrix) {
2413 CFRelease(vtctx->ycbcr_matrix);
2414 vtctx->ycbcr_matrix = NULL;
2420 static const enum AVPixelFormat pix_fmts[] = {
2421 AV_PIX_FMT_VIDEOTOOLBOX,
2427 #define OFFSET(x) offsetof(VTEncContext, x)
2428 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
2429 static const AVOption options[] = {
2430 { "profile", "Profile", OFFSET(profile), AV_OPT_TYPE_INT, { .i64 = H264_PROF_AUTO }, H264_PROF_AUTO, H264_PROF_COUNT, VE, "profile" },
2431 { "baseline", "Baseline Profile", 0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_BASELINE }, INT_MIN, INT_MAX, VE, "profile" },
2432 { "main", "Main Profile", 0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_MAIN }, INT_MIN, INT_MAX, VE, "profile" },
2433 { "high", "High Profile", 0, AV_OPT_TYPE_CONST, { .i64 = H264_PROF_HIGH }, INT_MIN, INT_MAX, VE, "profile" },
2435 { "level", "Level", OFFSET(level), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 52, VE, "level" },
2436 { "1.3", "Level 1.3, only available with Baseline Profile", 0, AV_OPT_TYPE_CONST, { .i64 = 13 }, INT_MIN, INT_MAX, VE, "level" },
2437 { "3.0", "Level 3.0", 0, AV_OPT_TYPE_CONST, { .i64 = 30 }, INT_MIN, INT_MAX, VE, "level" },
2438 { "3.1", "Level 3.1", 0, AV_OPT_TYPE_CONST, { .i64 = 31 }, INT_MIN, INT_MAX, VE, "level" },
2439 { "3.2", "Level 3.2", 0, AV_OPT_TYPE_CONST, { .i64 = 32 }, INT_MIN, INT_MAX, VE, "level" },
2440 { "4.0", "Level 4.0", 0, AV_OPT_TYPE_CONST, { .i64 = 40 }, INT_MIN, INT_MAX, VE, "level" },
2441 { "4.1", "Level 4.1", 0, AV_OPT_TYPE_CONST, { .i64 = 41 }, INT_MIN, INT_MAX, VE, "level" },
2442 { "4.2", "Level 4.2", 0, AV_OPT_TYPE_CONST, { .i64 = 42 }, INT_MIN, INT_MAX, VE, "level" },
2443 { "5.0", "Level 5.0", 0, AV_OPT_TYPE_CONST, { .i64 = 50 }, INT_MIN, INT_MAX, VE, "level" },
2444 { "5.1", "Level 5.1", 0, AV_OPT_TYPE_CONST, { .i64 = 51 }, INT_MIN, INT_MAX, VE, "level" },
2445 { "5.2", "Level 5.2", 0, AV_OPT_TYPE_CONST, { .i64 = 52 }, INT_MIN, INT_MAX, VE, "level" },
2447 { "allow_sw", "Allow software encoding", OFFSET(allow_sw), AV_OPT_TYPE_BOOL,
2448 { .i64 = 0 }, 0, 1, VE },
2450 { "coder", "Entropy coding", OFFSET(entropy), AV_OPT_TYPE_INT, { .i64 = VT_ENTROPY_NOT_SET }, VT_ENTROPY_NOT_SET, VT_CABAC, VE, "coder" },
2451 { "cavlc", "CAVLC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CAVLC }, INT_MIN, INT_MAX, VE, "coder" },
2452 { "vlc", "CAVLC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CAVLC }, INT_MIN, INT_MAX, VE, "coder" },
2453 { "cabac", "CABAC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CABAC }, INT_MIN, INT_MAX, VE, "coder" },
2454 { "ac", "CABAC entropy coding", 0, AV_OPT_TYPE_CONST, { .i64 = VT_CABAC }, INT_MIN, INT_MAX, VE, "coder" },
2456 { "realtime", "Hint that encoding should happen in real-time if not faster (e.g. capturing from camera).",
2457 OFFSET(realtime), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
2459 { "frames_before", "Other frames will come before the frames in this session. This helps smooth concatenation issues.",
2460 OFFSET(frames_before), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
2461 { "frames_after", "Other frames will come after the frames in this session. This helps smooth concatenation issues.",
2462 OFFSET(frames_after), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
2464 { "a53cc", "Use A53 Closed Captions (if available)", OFFSET(a53_cc), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, VE },
2469 static const AVClass h264_videotoolbox_class = {
2470 .class_name = "h264_videotoolbox",
2471 .item_name = av_default_item_name,
2473 .version = LIBAVUTIL_VERSION_INT,
2476 AVCodec ff_h264_videotoolbox_encoder = {
2477 .name = "h264_videotoolbox",
2478 .long_name = NULL_IF_CONFIG_SMALL("VideoToolbox H.264 Encoder"),
2479 .type = AVMEDIA_TYPE_VIDEO,
2480 .id = AV_CODEC_ID_H264,
2481 .priv_data_size = sizeof(VTEncContext),
2482 .pix_fmts = pix_fmts,
2484 .encode2 = vtenc_frame,
2485 .close = vtenc_close,
2486 .capabilities = AV_CODEC_CAP_DELAY,
2487 .priv_class = &h264_videotoolbox_class,
2488 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE |
2489 FF_CODEC_CAP_INIT_CLEANUP,