2 * OpenH264 video encoder
3 * Copyright (C) 2014 Martin Storsjo
5 * This file is part of Libav.
7 * Libav 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 * Libav 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 Libav; 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 typedef struct SVCContext {
37 const AVClass *av_class;
48 #define OFFSET(x) offsetof(SVCContext, x)
49 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
50 static const AVOption options[] = {
51 { "slice_mode", "Slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_AUTO_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
52 { "fixed", "A fixed number of slices", 0, AV_OPT_TYPE_CONST, { .i64 = SM_FIXEDSLCNUM_SLICE }, 0, 0, VE, "slice_mode" },
53 { "rowmb", "One slice per row of macroblocks", 0, AV_OPT_TYPE_CONST, { .i64 = SM_ROWMB_SLICE }, 0, 0, VE, "slice_mode" },
54 { "auto", "Automatic number of slices according to number of threads", 0, AV_OPT_TYPE_CONST, { .i64 = SM_AUTO_SLICE }, 0, 0, VE, "slice_mode" },
55 { "dyn", "Dynamic slicing", 0, AV_OPT_TYPE_CONST, { .i64 = SM_DYN_SLICE }, 0, 0, VE, "slice_mode" },
56 { "loopfilter", "Enable loop filter", OFFSET(loopfilter), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
57 { "profile", "Set profile restrictions", OFFSET(profile), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, VE },
58 { "max_nal_size", "Set maximum NAL size in bytes", OFFSET(max_nal_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
59 { "allow_skip_frames", "Allow skipping frames to hit the target bitrate", OFFSET(skip_frames), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
60 { "cabac", "Enable cabac", OFFSET(cabac), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
64 static const AVClass class = {
65 "libopenh264enc", av_default_item_name, options, LIBAVUTIL_VERSION_INT
68 static av_cold int svc_encode_close(AVCodecContext *avctx)
70 SVCContext *s = avctx->priv_data;
73 WelsDestroySVCEncoder(s->encoder);
75 av_log(avctx, AV_LOG_WARNING, "%d frames skipped\n", s->skipped);
79 static av_cold int svc_encode_init(AVCodecContext *avctx)
81 SVCContext *s = avctx->priv_data;
82 SEncParamExt param = { 0 };
85 WelsTraceCallback callback_function;
86 AVCPBProperties *props;
88 if ((err = ff_libopenh264_check_version(avctx)) < 0)
90 // Use a default error for multiple error paths below
91 err = AVERROR_UNKNOWN;
93 if (WelsCreateSVCEncoder(&s->encoder)) {
94 av_log(avctx, AV_LOG_ERROR, "Unable to create encoder\n");
95 return AVERROR_UNKNOWN;
98 // Pass all libopenh264 messages to our callback, to allow ourselves to filter them.
99 log_level = WELS_LOG_DETAIL;
100 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_LEVEL, &log_level);
102 // Set the logging callback function to one that uses av_log() (see implementation above).
103 callback_function = (WelsTraceCallback) ff_libopenh264_trace_callback;
104 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK, (void *)&callback_function);
106 // Set the AVCodecContext as the libopenh264 callback context so that it can be passed to av_log().
107 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK_CONTEXT, (void *)&avctx);
109 (*s->encoder)->GetDefaultParams(s->encoder, ¶m);
111 #if FF_API_CODER_TYPE
112 FF_DISABLE_DEPRECATION_WARNINGS
114 s->cabac = avctx->coder_type == FF_CODER_TYPE_AC;
115 FF_ENABLE_DEPRECATION_WARNINGS
118 param.fMaxFrameRate = avctx->time_base.den / avctx->time_base.num;
119 param.iPicWidth = avctx->width;
120 param.iPicHeight = avctx->height;
121 param.iTargetBitrate = avctx->bit_rate;
122 param.iMaxBitrate = FFMAX(avctx->rc_max_rate, avctx->bit_rate);
123 param.iRCMode = RC_QUALITY_MODE;
124 param.iTemporalLayerNum = 1;
125 param.iSpatialLayerNum = 1;
126 param.bEnableDenoise = 0;
127 param.bEnableBackgroundDetection = 1;
128 param.bEnableAdaptiveQuant = 1;
129 param.bEnableFrameSkip = s->skip_frames;
130 param.bEnableLongTermReference = 0;
131 param.iLtrMarkPeriod = 30;
132 param.uiIntraPeriod = avctx->gop_size;
133 #if OPENH264_VER_AT_LEAST(1, 4)
134 param.eSpsPpsIdStrategy = CONSTANT_ID;
136 param.bEnableSpsPpsIdAddition = 0;
138 param.bPrefixNalAddingCtrl = 0;
139 param.iLoopFilterDisableIdc = !s->loopfilter;
140 param.iEntropyCodingModeFlag = 0;
141 param.iMultipleThreadIdc = avctx->thread_count;
142 if (s->profile && !strcmp(s->profile, "main"))
143 param.iEntropyCodingModeFlag = 1;
144 else if (!s->profile && s->cabac)
145 param.iEntropyCodingModeFlag = 1;
147 param.sSpatialLayers[0].iVideoWidth = param.iPicWidth;
148 param.sSpatialLayers[0].iVideoHeight = param.iPicHeight;
149 param.sSpatialLayers[0].fFrameRate = param.fMaxFrameRate;
150 param.sSpatialLayers[0].iSpatialBitrate = param.iTargetBitrate;
151 param.sSpatialLayers[0].iMaxSpatialBitrate = param.iMaxBitrate;
153 if ((avctx->slices > 1) && (s->max_nal_size)) {
154 av_log(avctx, AV_LOG_ERROR,
155 "Invalid combination -slices %d and -max_nal_size %d.\n",
156 avctx->slices, s->max_nal_size);
160 if (avctx->slices > 1)
161 s->slice_mode = SM_FIXEDSLCNUM_SLICE;
164 s->slice_mode = SM_DYN_SLICE;
166 param.sSpatialLayers[0].sSliceCfg.uiSliceMode = s->slice_mode;
167 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceNum = avctx->slices;
169 if (s->slice_mode == SM_DYN_SLICE) {
170 if (s->max_nal_size){
171 param.uiMaxNalSize = s->max_nal_size;
172 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
174 av_log(avctx, AV_LOG_ERROR, "Invalid -max_nal_size, "
175 "specify a valid max_nal_size to use -slice_mode dyn\n");
180 if ((*s->encoder)->InitializeExt(s->encoder, ¶m) != cmResultSuccess) {
181 av_log(avctx, AV_LOG_ERROR, "Initialize failed\n");
185 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
186 SFrameBSInfo fbi = { 0 };
188 (*s->encoder)->EncodeParameterSets(s->encoder, &fbi);
189 for (i = 0; i < fbi.sLayerInfo[0].iNalCount; i++)
190 size += fbi.sLayerInfo[0].pNalLengthInByte[i];
191 avctx->extradata = av_mallocz(size + AV_INPUT_BUFFER_PADDING_SIZE);
192 if (!avctx->extradata) {
193 err = AVERROR(ENOMEM);
196 avctx->extradata_size = size;
197 memcpy(avctx->extradata, fbi.sLayerInfo[0].pBsBuf, size);
200 props = ff_add_cpb_side_data(avctx);
202 err = AVERROR(ENOMEM);
205 props->max_bitrate = param.iMaxBitrate;
206 props->avg_bitrate = param.iTargetBitrate;
211 svc_encode_close(avctx);
215 static int svc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
216 const AVFrame *frame, int *got_packet)
218 SVCContext *s = avctx->priv_data;
219 SFrameBSInfo fbi = { 0 };
222 SSourcePicture sp = { 0 };
223 int size = 0, layer, first_layer = 0;
224 int layer_size[MAX_LAYER_NUM_OF_FRAME] = { 0 };
226 sp.iColorFormat = videoFormatI420;
227 for (i = 0; i < 3; i++) {
228 sp.iStride[i] = frame->linesize[i];
229 sp.pData[i] = frame->data[i];
231 sp.iPicWidth = avctx->width;
232 sp.iPicHeight = avctx->height;
234 encoded = (*s->encoder)->EncodeFrame(s->encoder, &sp, &fbi);
235 if (encoded != cmResultSuccess) {
236 av_log(avctx, AV_LOG_ERROR, "EncodeFrame failed\n");
237 return AVERROR_UNKNOWN;
239 if (fbi.eFrameType == videoFrameTypeSkip) {
241 av_log(avctx, AV_LOG_DEBUG, "frame skipped\n");
245 // Normal frames are returned with one single layer, while IDR
246 // frames have two layers, where the first layer contains the SPS/PPS.
247 // If using global headers, don't include the SPS/PPS in the returned
248 // packet - thus, only return one layer.
249 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
250 first_layer = fbi.iLayerNum - 1;
252 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
253 for (i = 0; i < fbi.sLayerInfo[layer].iNalCount; i++)
254 layer_size[layer] += fbi.sLayerInfo[layer].pNalLengthInByte[i];
255 size += layer_size[layer];
257 av_log(avctx, AV_LOG_DEBUG, "%d slices\n", fbi.sLayerInfo[fbi.iLayerNum - 1].iNalCount);
259 if ((ret = ff_alloc_packet(avpkt, size))) {
260 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
264 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
265 memcpy(avpkt->data + size, fbi.sLayerInfo[layer].pBsBuf, layer_size[layer]);
266 size += layer_size[layer];
268 avpkt->pts = frame->pts;
269 if (fbi.eFrameType == videoFrameTypeIDR)
270 avpkt->flags |= AV_PKT_FLAG_KEY;
275 AVCodec ff_libopenh264_encoder = {
276 .name = "libopenh264",
277 .long_name = NULL_IF_CONFIG_SMALL("OpenH264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
278 .type = AVMEDIA_TYPE_VIDEO,
279 .id = AV_CODEC_ID_H264,
280 .priv_data_size = sizeof(SVCContext),
281 .init = svc_encode_init,
282 .encode2 = svc_encode_frame,
283 .close = svc_encode_close,
284 .capabilities = AV_CODEC_CAP_AUTO_THREADS,
285 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P,
287 .priv_class = &class,