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;
43 #define OPENH264_VER_AT_LEAST(maj, min) \
44 ((OPENH264_MAJOR > (maj)) || \
45 (OPENH264_MAJOR == (maj) && OPENH264_MINOR >= (min)))
47 #define OFFSET(x) offsetof(SVCContext, x)
48 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
49 static const AVOption options[] = {
50 { "slice_mode", "set slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_AUTO_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
51 { "fixed", "a fixed number of slices", 0, AV_OPT_TYPE_CONST, { .i64 = SM_FIXEDSLCNUM_SLICE }, 0, 0, VE, "slice_mode" },
52 { "rowmb", "one slice per row of macroblocks", 0, AV_OPT_TYPE_CONST, { .i64 = SM_ROWMB_SLICE }, 0, 0, VE, "slice_mode" },
53 { "auto", "automatic number of slices according to number of threads", 0, AV_OPT_TYPE_CONST, { .i64 = SM_AUTO_SLICE }, 0, 0, VE, "slice_mode" },
54 { "dyn", "Dynamic slicing", 0, AV_OPT_TYPE_CONST, { .i64 = SM_DYN_SLICE }, 0, 0, VE, "slice_mode" },
55 { "loopfilter", "enable loop filter", OFFSET(loopfilter), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
56 { "profile", "set profile restrictions", OFFSET(profile), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
57 { "max_nal_size", "Set maximum NAL size in bytes", OFFSET(max_nal_size), AV_OPT_TYPE_INT, { 0 }, 0, INT_MAX, VE },
61 static const AVClass class = {
62 "libopenh264enc", av_default_item_name, options, LIBAVUTIL_VERSION_INT
65 // Convert libopenh264 log level to equivalent ffmpeg log level.
66 static int libopenh264_to_libav_log_level(int libopenh264_log_level)
68 if (libopenh264_log_level >= WELS_LOG_DETAIL) return AV_LOG_TRACE;
69 else if (libopenh264_log_level >= WELS_LOG_DEBUG) return AV_LOG_DEBUG;
70 else if (libopenh264_log_level >= WELS_LOG_INFO) return AV_LOG_VERBOSE;
71 else if (libopenh264_log_level >= WELS_LOG_WARNING) return AV_LOG_WARNING;
72 else if (libopenh264_log_level >= WELS_LOG_ERROR) return AV_LOG_ERROR;
73 else return AV_LOG_QUIET;
76 // This function will be provided to the libopenh264 library. The function will be called
77 // when libopenh264 wants to log a message (error, warning, info, etc.). The signature for
78 // this function (defined in .../codec/api/svc/codec_api.h) is:
80 // typedef void (*WelsTraceCallback) (void* ctx, int level, const char* string);
82 static void libopenh264_trace_callback(void *ctx, int level, char const *msg)
84 // The message will be logged only if the requested EQUIVALENT ffmpeg log level is
85 // less than or equal to the current ffmpeg log level.
86 int equiv_libav_log_level = libopenh264_to_libav_log_level(level);
87 av_log(ctx, equiv_libav_log_level, "%s\n", msg);
90 static av_cold int svc_encode_close(AVCodecContext *avctx)
92 SVCContext *s = avctx->priv_data;
95 WelsDestroySVCEncoder(s->encoder);
99 static av_cold int svc_encode_init(AVCodecContext *avctx)
101 SVCContext *s = avctx->priv_data;
102 SEncParamExt param = { 0 };
103 int err = AVERROR_UNKNOWN;
105 WelsTraceCallback callback_function;
107 // Mingw GCC < 4.7 on x86_32 uses an incorrect/buggy ABI for the WelsGetCodecVersion
108 // function (for functions returning larger structs), thus skip the check in those
110 #if !defined(_WIN32) || !defined(__GNUC__) || !ARCH_X86_32 || AV_GCC_VERSION_AT_LEAST(4, 7)
111 OpenH264Version libver = WelsGetCodecVersion();
112 if (memcmp(&libver, &g_stCodecVersion, sizeof(libver))) {
113 av_log(avctx, AV_LOG_ERROR, "Incorrect library version loaded\n");
114 return AVERROR(EINVAL);
118 if (WelsCreateSVCEncoder(&s->encoder)) {
119 av_log(avctx, AV_LOG_ERROR, "Unable to create encoder\n");
120 return AVERROR_UNKNOWN;
123 // Pass all libopenh264 messages to our callback, to allow ourselves to filter them.
124 log_level = WELS_LOG_DETAIL;
125 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_LEVEL, &log_level);
127 // Set the logging callback function to one that uses av_log() (see implementation above).
128 callback_function = (WelsTraceCallback) libopenh264_trace_callback;
129 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK, (void *)&callback_function);
131 // Set the AVCodecContext as the libopenh264 callback context so that it can be passed to av_log().
132 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK_CONTEXT, (void *)&avctx);
134 (*s->encoder)->GetDefaultParams(s->encoder, ¶m);
136 param.fMaxFrameRate = 1/av_q2d(avctx->time_base);
137 param.iPicWidth = avctx->width;
138 param.iPicHeight = avctx->height;
139 param.iTargetBitrate = avctx->bit_rate;
140 param.iMaxBitrate = FFMAX(avctx->rc_max_rate, avctx->bit_rate);
141 param.iRCMode = RC_QUALITY_MODE;
142 param.iTemporalLayerNum = 1;
143 param.iSpatialLayerNum = 1;
144 param.bEnableDenoise = 0;
145 param.bEnableBackgroundDetection = 1;
146 param.bEnableAdaptiveQuant = 1;
147 param.bEnableFrameSkip = 0;
148 param.bEnableLongTermReference = 0;
149 param.iLtrMarkPeriod = 30;
150 param.uiIntraPeriod = avctx->gop_size;
151 #if OPENH264_VER_AT_LEAST(1, 4)
152 param.eSpsPpsIdStrategy = CONSTANT_ID;
154 param.bEnableSpsPpsIdAddition = 0;
156 param.bPrefixNalAddingCtrl = 0;
157 param.iLoopFilterDisableIdc = !s->loopfilter;
158 param.iEntropyCodingModeFlag = 0;
159 param.iMultipleThreadIdc = avctx->thread_count;
160 if (s->profile && !strcmp(s->profile, "main"))
161 param.iEntropyCodingModeFlag = 1;
162 else if (!s->profile && avctx->coder_type == FF_CODER_TYPE_AC)
163 param.iEntropyCodingModeFlag = 1;
165 param.sSpatialLayers[0].iVideoWidth = param.iPicWidth;
166 param.sSpatialLayers[0].iVideoHeight = param.iPicHeight;
167 param.sSpatialLayers[0].fFrameRate = param.fMaxFrameRate;
168 param.sSpatialLayers[0].iSpatialBitrate = param.iTargetBitrate;
169 param.sSpatialLayers[0].iMaxSpatialBitrate = param.iMaxBitrate;
171 if ((avctx->slices > 1) && (s->max_nal_size)){
172 av_log(avctx,AV_LOG_ERROR,"Invalid combination -slices %d and -max_nal_size %d.\n",avctx->slices,s->max_nal_size);
176 if (avctx->slices > 1)
177 s->slice_mode = SM_FIXEDSLCNUM_SLICE;
180 s->slice_mode = SM_DYN_SLICE;
182 param.sSpatialLayers[0].sSliceCfg.uiSliceMode = s->slice_mode;
183 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceNum = avctx->slices;
185 if (s->slice_mode == SM_DYN_SLICE) {
186 if (s->max_nal_size){
187 param.uiMaxNalSize = s->max_nal_size;
188 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
190 if (avctx->rtp_payload_size) {
191 av_log(avctx,AV_LOG_DEBUG,"Using RTP Payload size for uiMaxNalSize");
192 param.uiMaxNalSize = avctx->rtp_payload_size;
193 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = avctx->rtp_payload_size;
195 av_log(avctx,AV_LOG_ERROR,"Invalid -max_nal_size, specify a valid max_nal_size to use -slice_mode dyn\n");
201 if ((*s->encoder)->InitializeExt(s->encoder, ¶m) != cmResultSuccess) {
202 av_log(avctx, AV_LOG_ERROR, "Initialize failed\n");
206 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
207 SFrameBSInfo fbi = { 0 };
209 (*s->encoder)->EncodeParameterSets(s->encoder, &fbi);
210 for (i = 0; i < fbi.sLayerInfo[0].iNalCount; i++)
211 size += fbi.sLayerInfo[0].pNalLengthInByte[i];
212 avctx->extradata = av_mallocz(size + AV_INPUT_BUFFER_PADDING_SIZE);
213 if (!avctx->extradata) {
214 err = AVERROR(ENOMEM);
217 avctx->extradata_size = size;
218 memcpy(avctx->extradata, fbi.sLayerInfo[0].pBsBuf, size);
224 svc_encode_close(avctx);
228 static int svc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
229 const AVFrame *frame, int *got_packet)
231 SVCContext *s = avctx->priv_data;
232 SFrameBSInfo fbi = { 0 };
235 SSourcePicture sp = { 0 };
236 int size = 0, layer, first_layer = 0;
237 int layer_size[MAX_LAYER_NUM_OF_FRAME] = { 0 };
239 sp.iColorFormat = videoFormatI420;
240 for (i = 0; i < 3; i++) {
241 sp.iStride[i] = frame->linesize[i];
242 sp.pData[i] = frame->data[i];
244 sp.iPicWidth = avctx->width;
245 sp.iPicHeight = avctx->height;
247 encoded = (*s->encoder)->EncodeFrame(s->encoder, &sp, &fbi);
248 if (encoded != cmResultSuccess) {
249 av_log(avctx, AV_LOG_ERROR, "EncodeFrame failed\n");
250 return AVERROR_UNKNOWN;
252 if (fbi.eFrameType == videoFrameTypeSkip) {
253 av_log(avctx, AV_LOG_DEBUG, "frame skipped\n");
257 // Normal frames are returned with one single layer, while IDR
258 // frames have two layers, where the first layer contains the SPS/PPS.
259 // If using global headers, don't include the SPS/PPS in the returned
260 // packet - thus, only return one layer.
261 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
262 first_layer = fbi.iLayerNum - 1;
264 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
265 for (i = 0; i < fbi.sLayerInfo[layer].iNalCount; i++)
266 layer_size[layer] += fbi.sLayerInfo[layer].pNalLengthInByte[i];
267 size += layer_size[layer];
269 av_log(avctx, AV_LOG_DEBUG, "%d slices\n", fbi.sLayerInfo[fbi.iLayerNum - 1].iNalCount);
271 if ((ret = ff_alloc_packet2(avctx, avpkt, size, size))) {
272 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
276 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
277 memcpy(avpkt->data + size, fbi.sLayerInfo[layer].pBsBuf, layer_size[layer]);
278 size += layer_size[layer];
280 avpkt->pts = frame->pts;
281 if (fbi.eFrameType == videoFrameTypeIDR)
282 avpkt->flags |= AV_PKT_FLAG_KEY;
287 AVCodec ff_libopenh264_encoder = {
288 .name = "libopenh264",
289 .long_name = NULL_IF_CONFIG_SMALL("OpenH264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
290 .type = AVMEDIA_TYPE_VIDEO,
291 .id = AV_CODEC_ID_H264,
292 .priv_data_size = sizeof(SVCContext),
293 .init = svc_encode_init,
294 .encode2 = svc_encode_frame,
295 .close = svc_encode_close,
296 .capabilities = AV_CODEC_CAP_AUTO_THREADS,
297 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P,
299 .priv_class = &class,