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/intreadwrite.h"
29 #include "libavutil/mathematics.h"
34 typedef struct SVCContext {
35 const AVClass *av_class;
45 #define OPENH264_VER_AT_LEAST(maj, min) \
46 ((OPENH264_MAJOR > (maj)) || \
47 (OPENH264_MAJOR == (maj) && OPENH264_MINOR >= (min)))
49 #define OFFSET(x) offsetof(SVCContext, x)
50 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
51 static const AVOption options[] = {
52 { "slice_mode", "set slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_AUTO_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
53 { "fixed", "a fixed number of slices", 0, AV_OPT_TYPE_CONST, { .i64 = SM_FIXEDSLCNUM_SLICE }, 0, 0, VE, "slice_mode" },
54 { "rowmb", "one slice per row of macroblocks", 0, AV_OPT_TYPE_CONST, { .i64 = SM_ROWMB_SLICE }, 0, 0, VE, "slice_mode" },
55 { "auto", "automatic number of slices according to number of threads", 0, AV_OPT_TYPE_CONST, { .i64 = SM_AUTO_SLICE }, 0, 0, VE, "slice_mode" },
56 { "dyn", "Dynamic slicing", 0, AV_OPT_TYPE_CONST, { .i64 = SM_DYN_SLICE }, 0, 0, VE, "slice_mode" },
57 { "loopfilter", "enable loop filter", OFFSET(loopfilter), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
58 { "profile", "set profile restrictions", OFFSET(profile), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, VE },
59 { "max_nal_size", "Set maximum NAL size in bytes", OFFSET(max_nal_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
60 { "allow_skip_frames", "Allow skipping frames to hit the target bitrate", OFFSET(skip_frames), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
64 static const AVClass class = {
65 "libopenh264enc", av_default_item_name, options, LIBAVUTIL_VERSION_INT
68 // Convert libopenh264 log level to equivalent ffmpeg log level.
69 static int libopenh264_to_ffmpeg_log_level(int libopenh264_log_level)
71 if (libopenh264_log_level >= WELS_LOG_DETAIL) return AV_LOG_TRACE;
72 else if (libopenh264_log_level >= WELS_LOG_DEBUG) return AV_LOG_DEBUG;
73 else if (libopenh264_log_level >= WELS_LOG_INFO) return AV_LOG_VERBOSE;
74 else if (libopenh264_log_level >= WELS_LOG_WARNING) return AV_LOG_WARNING;
75 else if (libopenh264_log_level >= WELS_LOG_ERROR) return AV_LOG_ERROR;
76 else return AV_LOG_QUIET;
79 // This function will be provided to the libopenh264 library. The function will be called
80 // when libopenh264 wants to log a message (error, warning, info, etc.). The signature for
81 // this function (defined in .../codec/api/svc/codec_api.h) is:
83 // typedef void (*WelsTraceCallback) (void* ctx, int level, const char* string);
85 static void libopenh264_trace_callback(void *ctx, int level, char const *msg)
87 // The message will be logged only if the requested EQUIVALENT ffmpeg log level is
88 // less than or equal to the current ffmpeg log level.
89 int equiv_ffmpeg_log_level = libopenh264_to_ffmpeg_log_level(level);
90 av_log(ctx, equiv_ffmpeg_log_level, "%s\n", msg);
93 static av_cold int svc_encode_close(AVCodecContext *avctx)
95 SVCContext *s = avctx->priv_data;
98 WelsDestroySVCEncoder(s->encoder);
100 av_log(avctx, AV_LOG_WARNING, "%d frames skipped\n", s->skipped);
104 static av_cold int svc_encode_init(AVCodecContext *avctx)
106 SVCContext *s = avctx->priv_data;
107 SEncParamExt param = { 0 };
108 int err = AVERROR_UNKNOWN;
110 WelsTraceCallback callback_function;
112 // Mingw GCC < 4.7 on x86_32 uses an incorrect/buggy ABI for the WelsGetCodecVersion
113 // function (for functions returning larger structs), thus skip the check in those
115 #if !defined(_WIN32) || !defined(__GNUC__) || !ARCH_X86_32 || AV_GCC_VERSION_AT_LEAST(4, 7)
116 OpenH264Version libver = WelsGetCodecVersion();
117 if (memcmp(&libver, &g_stCodecVersion, sizeof(libver))) {
118 av_log(avctx, AV_LOG_ERROR, "Incorrect library version loaded\n");
119 return AVERROR(EINVAL);
123 if (WelsCreateSVCEncoder(&s->encoder)) {
124 av_log(avctx, AV_LOG_ERROR, "Unable to create encoder\n");
125 return AVERROR_UNKNOWN;
128 // Pass all libopenh264 messages to our callback, to allow ourselves to filter them.
129 log_level = WELS_LOG_DETAIL;
130 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_LEVEL, &log_level);
132 // Set the logging callback function to one that uses av_log() (see implementation above).
133 callback_function = (WelsTraceCallback) libopenh264_trace_callback;
134 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK, (void *)&callback_function);
136 // Set the AVCodecContext as the libopenh264 callback context so that it can be passed to av_log().
137 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK_CONTEXT, (void *)&avctx);
139 (*s->encoder)->GetDefaultParams(s->encoder, ¶m);
141 param.fMaxFrameRate = 1/av_q2d(avctx->time_base);
142 param.iPicWidth = avctx->width;
143 param.iPicHeight = avctx->height;
144 param.iTargetBitrate = avctx->bit_rate;
145 param.iMaxBitrate = FFMAX(avctx->rc_max_rate, avctx->bit_rate);
146 param.iRCMode = RC_QUALITY_MODE;
147 param.iTemporalLayerNum = 1;
148 param.iSpatialLayerNum = 1;
149 param.bEnableDenoise = 0;
150 param.bEnableBackgroundDetection = 1;
151 param.bEnableAdaptiveQuant = 1;
152 param.bEnableFrameSkip = s->skip_frames;
153 param.bEnableLongTermReference = 0;
154 param.iLtrMarkPeriod = 30;
155 param.uiIntraPeriod = avctx->gop_size;
156 #if OPENH264_VER_AT_LEAST(1, 4)
157 param.eSpsPpsIdStrategy = CONSTANT_ID;
159 param.bEnableSpsPpsIdAddition = 0;
161 param.bPrefixNalAddingCtrl = 0;
162 param.iLoopFilterDisableIdc = !s->loopfilter;
163 param.iEntropyCodingModeFlag = 0;
164 param.iMultipleThreadIdc = avctx->thread_count;
165 if (s->profile && !strcmp(s->profile, "main"))
166 param.iEntropyCodingModeFlag = 1;
167 else if (!s->profile && avctx->coder_type == FF_CODER_TYPE_AC)
168 param.iEntropyCodingModeFlag = 1;
170 param.sSpatialLayers[0].iVideoWidth = param.iPicWidth;
171 param.sSpatialLayers[0].iVideoHeight = param.iPicHeight;
172 param.sSpatialLayers[0].fFrameRate = param.fMaxFrameRate;
173 param.sSpatialLayers[0].iSpatialBitrate = param.iTargetBitrate;
174 param.sSpatialLayers[0].iMaxSpatialBitrate = param.iMaxBitrate;
176 if ((avctx->slices > 1) && (s->max_nal_size)){
177 av_log(avctx,AV_LOG_ERROR,"Invalid combination -slices %d and -max_nal_size %d.\n",avctx->slices,s->max_nal_size);
181 if (avctx->slices > 1)
182 s->slice_mode = SM_FIXEDSLCNUM_SLICE;
185 s->slice_mode = SM_DYN_SLICE;
187 param.sSpatialLayers[0].sSliceCfg.uiSliceMode = s->slice_mode;
188 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceNum = avctx->slices;
190 if (s->slice_mode == SM_DYN_SLICE) {
191 if (s->max_nal_size){
192 param.uiMaxNalSize = s->max_nal_size;
193 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
195 if (avctx->rtp_payload_size) {
196 av_log(avctx,AV_LOG_DEBUG,"Using RTP Payload size for uiMaxNalSize");
197 param.uiMaxNalSize = avctx->rtp_payload_size;
198 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = avctx->rtp_payload_size;
200 av_log(avctx,AV_LOG_ERROR,"Invalid -max_nal_size, specify a valid max_nal_size to use -slice_mode dyn\n");
206 if ((*s->encoder)->InitializeExt(s->encoder, ¶m) != cmResultSuccess) {
207 av_log(avctx, AV_LOG_ERROR, "Initialize failed\n");
211 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
212 SFrameBSInfo fbi = { 0 };
214 (*s->encoder)->EncodeParameterSets(s->encoder, &fbi);
215 for (i = 0; i < fbi.sLayerInfo[0].iNalCount; i++)
216 size += fbi.sLayerInfo[0].pNalLengthInByte[i];
217 avctx->extradata = av_mallocz(size + AV_INPUT_BUFFER_PADDING_SIZE);
218 if (!avctx->extradata) {
219 err = AVERROR(ENOMEM);
222 avctx->extradata_size = size;
223 memcpy(avctx->extradata, fbi.sLayerInfo[0].pBsBuf, size);
229 svc_encode_close(avctx);
233 static int svc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
234 const AVFrame *frame, int *got_packet)
236 SVCContext *s = avctx->priv_data;
237 SFrameBSInfo fbi = { 0 };
240 SSourcePicture sp = { 0 };
241 int size = 0, layer, first_layer = 0;
242 int layer_size[MAX_LAYER_NUM_OF_FRAME] = { 0 };
244 sp.iColorFormat = videoFormatI420;
245 for (i = 0; i < 3; i++) {
246 sp.iStride[i] = frame->linesize[i];
247 sp.pData[i] = frame->data[i];
249 sp.iPicWidth = avctx->width;
250 sp.iPicHeight = avctx->height;
252 encoded = (*s->encoder)->EncodeFrame(s->encoder, &sp, &fbi);
253 if (encoded != cmResultSuccess) {
254 av_log(avctx, AV_LOG_ERROR, "EncodeFrame failed\n");
255 return AVERROR_UNKNOWN;
257 if (fbi.eFrameType == videoFrameTypeSkip) {
259 av_log(avctx, AV_LOG_DEBUG, "frame skipped\n");
263 // Normal frames are returned with one single layer, while IDR
264 // frames have two layers, where the first layer contains the SPS/PPS.
265 // If using global headers, don't include the SPS/PPS in the returned
266 // packet - thus, only return one layer.
267 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
268 first_layer = fbi.iLayerNum - 1;
270 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
271 for (i = 0; i < fbi.sLayerInfo[layer].iNalCount; i++)
272 layer_size[layer] += fbi.sLayerInfo[layer].pNalLengthInByte[i];
273 size += layer_size[layer];
275 av_log(avctx, AV_LOG_DEBUG, "%d slices\n", fbi.sLayerInfo[fbi.iLayerNum - 1].iNalCount);
277 if ((ret = ff_alloc_packet2(avctx, avpkt, size, size))) {
278 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
282 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
283 memcpy(avpkt->data + size, fbi.sLayerInfo[layer].pBsBuf, layer_size[layer]);
284 size += layer_size[layer];
286 avpkt->pts = frame->pts;
287 if (fbi.eFrameType == videoFrameTypeIDR)
288 avpkt->flags |= AV_PKT_FLAG_KEY;
293 AVCodec ff_libopenh264_encoder = {
294 .name = "libopenh264",
295 .long_name = NULL_IF_CONFIG_SMALL("OpenH264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
296 .type = AVMEDIA_TYPE_VIDEO,
297 .id = AV_CODEC_ID_H264,
298 .priv_data_size = sizeof(SVCContext),
299 .init = svc_encode_init,
300 .encode2 = svc_encode_frame,
301 .close = svc_encode_close,
302 .capabilities = AV_CODEC_CAP_AUTO_THREADS,
303 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P,
305 .priv_class = &class,