2 * OpenH264 video encoder
3 * Copyright (C) 2014 Martin Storsjo
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #include <wels/codec_api.h>
23 #include <wels/codec_ver.h>
25 #include "libavutil/attributes.h"
26 #include "libavutil/common.h"
27 #include "libavutil/opt.h"
28 #include "libavutil/internal.h"
29 #include "libavutil/intreadwrite.h"
30 #include "libavutil/mathematics.h"
34 #include "libopenh264.h"
36 #if !OPENH264_VER_AT_LEAST(1, 6)
37 #define SM_SIZELIMITED_SLICE SM_DYN_SLICE
40 typedef struct SVCContext {
41 const AVClass *av_class;
52 #define OFFSET(x) offsetof(SVCContext, x)
53 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
54 static const AVOption options[] = {
55 #if OPENH264_VER_AT_LEAST(1, 6)
56 { "slice_mode", "set slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_FIXEDSLCNUM_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
58 { "slice_mode", "set slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_AUTO_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
60 { "fixed", "a fixed number of slices", 0, AV_OPT_TYPE_CONST, { .i64 = SM_FIXEDSLCNUM_SLICE }, 0, 0, VE, "slice_mode" },
61 #if OPENH264_VER_AT_LEAST(1, 6)
62 { "dyn", "Size limited (compatibility name)", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
63 { "sizelimited", "Size limited", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
65 { "rowmb", "one slice per row of macroblocks", 0, AV_OPT_TYPE_CONST, { .i64 = SM_ROWMB_SLICE }, 0, 0, VE, "slice_mode" },
66 { "auto", "automatic number of slices according to number of threads", 0, AV_OPT_TYPE_CONST, { .i64 = SM_AUTO_SLICE }, 0, 0, VE, "slice_mode" },
67 { "dyn", "Dynamic slicing", 0, AV_OPT_TYPE_CONST, { .i64 = SM_DYN_SLICE }, 0, 0, VE, "slice_mode" },
69 { "loopfilter", "enable loop filter", OFFSET(loopfilter), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
70 { "profile", "set profile restrictions", OFFSET(profile), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, VE },
71 { "max_nal_size", "set maximum NAL size in bytes", OFFSET(max_nal_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
72 { "allow_skip_frames", "allow skipping frames to hit the target bitrate", OFFSET(skip_frames), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
73 { "cabac", "Enable cabac", OFFSET(cabac), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
77 static const AVClass class = {
78 .class_name = "libopenh264enc",
79 .item_name = av_default_item_name,
81 .version = LIBAVUTIL_VERSION_INT,
84 static av_cold int svc_encode_close(AVCodecContext *avctx)
86 SVCContext *s = avctx->priv_data;
89 WelsDestroySVCEncoder(s->encoder);
91 av_log(avctx, AV_LOG_WARNING, "%d frames skipped\n", s->skipped);
95 static av_cold int svc_encode_init(AVCodecContext *avctx)
97 SVCContext *s = avctx->priv_data;
98 SEncParamExt param = { 0 };
101 WelsTraceCallback callback_function;
102 AVCPBProperties *props;
104 if ((err = ff_libopenh264_check_version(avctx)) < 0)
107 if (WelsCreateSVCEncoder(&s->encoder)) {
108 av_log(avctx, AV_LOG_ERROR, "Unable to create encoder\n");
109 return AVERROR_UNKNOWN;
112 // Pass all libopenh264 messages to our callback, to allow ourselves to filter them.
113 log_level = WELS_LOG_DETAIL;
114 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_LEVEL, &log_level);
116 // Set the logging callback function to one that uses av_log() (see implementation above).
117 callback_function = (WelsTraceCallback) ff_libopenh264_trace_callback;
118 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK, &callback_function);
120 // Set the AVCodecContext as the libopenh264 callback context so that it can be passed to av_log().
121 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK_CONTEXT, &avctx);
123 (*s->encoder)->GetDefaultParams(s->encoder, ¶m);
125 #if FF_API_CODER_TYPE
126 FF_DISABLE_DEPRECATION_WARNINGS
128 s->cabac = avctx->coder_type == FF_CODER_TYPE_AC;
129 FF_ENABLE_DEPRECATION_WARNINGS
132 param.fMaxFrameRate = 1/av_q2d(avctx->time_base);
133 param.iPicWidth = avctx->width;
134 param.iPicHeight = avctx->height;
135 param.iTargetBitrate = avctx->bit_rate;
136 param.iMaxBitrate = FFMAX(avctx->rc_max_rate, avctx->bit_rate);
137 param.iRCMode = RC_QUALITY_MODE;
138 param.iTemporalLayerNum = 1;
139 param.iSpatialLayerNum = 1;
140 param.bEnableDenoise = 0;
141 param.bEnableBackgroundDetection = 1;
142 param.bEnableAdaptiveQuant = 1;
143 param.bEnableFrameSkip = s->skip_frames;
144 param.bEnableLongTermReference = 0;
145 param.iLtrMarkPeriod = 30;
146 param.uiIntraPeriod = avctx->gop_size;
147 #if OPENH264_VER_AT_LEAST(1, 4)
148 param.eSpsPpsIdStrategy = CONSTANT_ID;
150 param.bEnableSpsPpsIdAddition = 0;
152 param.bPrefixNalAddingCtrl = 0;
153 param.iLoopFilterDisableIdc = !s->loopfilter;
154 param.iEntropyCodingModeFlag = 0;
155 param.iMultipleThreadIdc = avctx->thread_count;
156 if (s->profile && !strcmp(s->profile, "main"))
157 param.iEntropyCodingModeFlag = 1;
158 else if (!s->profile && s->cabac)
159 param.iEntropyCodingModeFlag = 1;
161 param.sSpatialLayers[0].iVideoWidth = param.iPicWidth;
162 param.sSpatialLayers[0].iVideoHeight = param.iPicHeight;
163 param.sSpatialLayers[0].fFrameRate = param.fMaxFrameRate;
164 param.sSpatialLayers[0].iSpatialBitrate = param.iTargetBitrate;
165 param.sSpatialLayers[0].iMaxSpatialBitrate = param.iMaxBitrate;
167 #if OPENH264_VER_AT_LEAST(1, 7)
168 if (avctx->sample_aspect_ratio.num && avctx->sample_aspect_ratio.den) {
170 static const AVRational sar_idc[] = {
171 { 0, 0 }, // Unspecified (never written here).
172 { 1, 1 }, { 12, 11 }, { 10, 11 }, { 16, 11 },
173 { 40, 33 }, { 24, 11 }, { 20, 11 }, { 32, 11 },
174 { 80, 33 }, { 18, 11 }, { 15, 11 }, { 64, 33 },
175 { 160, 99 }, // Last 3 are unknown to openh264: { 4, 3 }, { 3, 2 }, { 2, 1 },
177 static const ESampleAspectRatio asp_idc[] = {
179 ASP_1x1, ASP_12x11, ASP_10x11, ASP_16x11,
180 ASP_40x33, ASP_24x11, ASP_20x11, ASP_32x11,
181 ASP_80x33, ASP_18x11, ASP_15x11, ASP_64x33,
186 av_reduce(&num, &den, avctx->sample_aspect_ratio.num,
187 avctx->sample_aspect_ratio.den, 65535);
189 for (i = 1; i < FF_ARRAY_ELEMS(sar_idc); i++) {
190 if (num == sar_idc[i].num &&
191 den == sar_idc[i].den)
194 if (i == FF_ARRAY_ELEMS(sar_idc)) {
195 param.sSpatialLayers[0].eAspectRatio = ASP_EXT_SAR;
196 param.sSpatialLayers[0].sAspectRatioExtWidth = num;
197 param.sSpatialLayers[0].sAspectRatioExtHeight = den;
199 param.sSpatialLayers[0].eAspectRatio = asp_idc[i];
201 param.sSpatialLayers[0].bAspectRatioPresent = true;
204 param.sSpatialLayers[0].bAspectRatioPresent = false;
208 if ((avctx->slices > 1) && (s->max_nal_size)) {
209 av_log(avctx, AV_LOG_ERROR,
210 "Invalid combination -slices %d and -max_nal_size %d.\n",
211 avctx->slices, s->max_nal_size);
212 return AVERROR(EINVAL);
215 if (avctx->slices > 1)
216 s->slice_mode = SM_FIXEDSLCNUM_SLICE;
219 s->slice_mode = SM_SIZELIMITED_SLICE;
221 #if OPENH264_VER_AT_LEAST(1, 6)
222 param.sSpatialLayers[0].sSliceArgument.uiSliceMode = s->slice_mode;
223 param.sSpatialLayers[0].sSliceArgument.uiSliceNum = avctx->slices;
225 param.sSpatialLayers[0].sSliceCfg.uiSliceMode = s->slice_mode;
226 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceNum = avctx->slices;
229 if (s->slice_mode == SM_SIZELIMITED_SLICE) {
230 if (s->max_nal_size){
231 param.uiMaxNalSize = s->max_nal_size;
232 #if OPENH264_VER_AT_LEAST(1, 6)
233 param.sSpatialLayers[0].sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
235 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
238 av_log(avctx, AV_LOG_ERROR, "Invalid -max_nal_size, "
239 "specify a valid max_nal_size to use -slice_mode dyn\n");
240 return AVERROR(EINVAL);
244 if ((*s->encoder)->InitializeExt(s->encoder, ¶m) != cmResultSuccess) {
245 av_log(avctx, AV_LOG_ERROR, "Initialize failed\n");
246 return AVERROR_UNKNOWN;
249 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
250 SFrameBSInfo fbi = { 0 };
252 (*s->encoder)->EncodeParameterSets(s->encoder, &fbi);
253 for (i = 0; i < fbi.sLayerInfo[0].iNalCount; i++)
254 size += fbi.sLayerInfo[0].pNalLengthInByte[i];
255 avctx->extradata = av_mallocz(size + AV_INPUT_BUFFER_PADDING_SIZE);
256 if (!avctx->extradata)
257 return AVERROR(ENOMEM);
258 avctx->extradata_size = size;
259 memcpy(avctx->extradata, fbi.sLayerInfo[0].pBsBuf, size);
262 props = ff_add_cpb_side_data(avctx);
264 return AVERROR(ENOMEM);
265 props->max_bitrate = param.iMaxBitrate;
266 props->avg_bitrate = param.iTargetBitrate;
271 static int svc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
272 const AVFrame *frame, int *got_packet)
274 SVCContext *s = avctx->priv_data;
275 SFrameBSInfo fbi = { 0 };
278 SSourcePicture sp = { 0 };
279 int size = 0, layer, first_layer = 0;
280 int layer_size[MAX_LAYER_NUM_OF_FRAME] = { 0 };
282 sp.iColorFormat = videoFormatI420;
283 for (i = 0; i < 3; i++) {
284 sp.iStride[i] = frame->linesize[i];
285 sp.pData[i] = frame->data[i];
287 sp.iPicWidth = avctx->width;
288 sp.iPicHeight = avctx->height;
290 if (frame->pict_type == AV_PICTURE_TYPE_I) {
291 (*s->encoder)->ForceIntraFrame(s->encoder, true);
294 encoded = (*s->encoder)->EncodeFrame(s->encoder, &sp, &fbi);
295 if (encoded != cmResultSuccess) {
296 av_log(avctx, AV_LOG_ERROR, "EncodeFrame failed\n");
297 return AVERROR_UNKNOWN;
299 if (fbi.eFrameType == videoFrameTypeSkip) {
301 av_log(avctx, AV_LOG_DEBUG, "frame skipped\n");
305 // Normal frames are returned with one single layer, while IDR
306 // frames have two layers, where the first layer contains the SPS/PPS.
307 // If using global headers, don't include the SPS/PPS in the returned
308 // packet - thus, only return one layer.
309 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
310 first_layer = fbi.iLayerNum - 1;
312 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
313 for (i = 0; i < fbi.sLayerInfo[layer].iNalCount; i++)
314 layer_size[layer] += fbi.sLayerInfo[layer].pNalLengthInByte[i];
315 size += layer_size[layer];
317 av_log(avctx, AV_LOG_DEBUG, "%d slices\n", fbi.sLayerInfo[fbi.iLayerNum - 1].iNalCount);
319 if ((ret = ff_alloc_packet2(avctx, avpkt, size, size))) {
320 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
324 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
325 memcpy(avpkt->data + size, fbi.sLayerInfo[layer].pBsBuf, layer_size[layer]);
326 size += layer_size[layer];
328 avpkt->pts = frame->pts;
329 if (fbi.eFrameType == videoFrameTypeIDR)
330 avpkt->flags |= AV_PKT_FLAG_KEY;
335 AVCodec ff_libopenh264_encoder = {
336 .name = "libopenh264",
337 .long_name = NULL_IF_CONFIG_SMALL("OpenH264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
338 .type = AVMEDIA_TYPE_VIDEO,
339 .id = AV_CODEC_ID_H264,
340 .priv_data_size = sizeof(SVCContext),
341 .init = svc_encode_init,
342 .encode2 = svc_encode_frame,
343 .close = svc_encode_close,
344 .capabilities = AV_CODEC_CAP_AUTO_THREADS,
345 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
346 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P,
348 .priv_class = &class,
349 .wrapper_name = "libopenh264",