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;
42 #define OFFSET(x) offsetof(SVCContext, x)
43 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
44 static const AVOption options[] = {
45 { "slice_mode", "Slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_AUTO_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
46 { "fixed", "A fixed number of slices", 0, AV_OPT_TYPE_CONST, { .i64 = SM_FIXEDSLCNUM_SLICE }, 0, 0, VE, "slice_mode" },
47 { "rowmb", "One slice per row of macroblocks", 0, AV_OPT_TYPE_CONST, { .i64 = SM_ROWMB_SLICE }, 0, 0, VE, "slice_mode" },
48 { "auto", "Automatic number of slices according to number of threads", 0, AV_OPT_TYPE_CONST, { .i64 = SM_AUTO_SLICE }, 0, 0, VE, "slice_mode" },
49 { "loopfilter", "Enable loop filter", OFFSET(loopfilter), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
50 { "profile", "Set profile restrictions", OFFSET(profile), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
54 static const AVClass class = {
55 "libopenh264enc", av_default_item_name, options, LIBAVUTIL_VERSION_INT
58 static av_cold int svc_encode_close(AVCodecContext *avctx)
60 SVCContext *s = avctx->priv_data;
63 WelsDestroySVCEncoder(s->encoder);
67 static av_cold int svc_encode_init(AVCodecContext *avctx)
69 SVCContext *s = avctx->priv_data;
70 SEncParamExt param = { 0 };
71 int err = AVERROR_UNKNOWN;
73 // Mingw GCC < 4.7 on x86_32 uses an incorrect/buggy ABI for the WelsGetCodecVersion
74 // function (for functions returning larger structs), thus skip the check in those
76 #if !defined(_WIN32) || !defined(__GNUC__) || !ARCH_X86_32 || AV_GCC_VERSION_AT_LEAST(4, 7)
77 OpenH264Version libver = WelsGetCodecVersion();
78 if (memcmp(&libver, &g_stCodecVersion, sizeof(libver))) {
79 av_log(avctx, AV_LOG_ERROR, "Incorrect library version loaded\n");
80 return AVERROR(EINVAL);
84 if (WelsCreateSVCEncoder(&s->encoder)) {
85 av_log(avctx, AV_LOG_ERROR, "Unable to create encoder\n");
86 return AVERROR_UNKNOWN;
89 (*s->encoder)->GetDefaultParams(s->encoder, ¶m);
91 param.fMaxFrameRate = avctx->time_base.den / avctx->time_base.num;
92 param.iPicWidth = avctx->width;
93 param.iPicHeight = avctx->height;
94 param.iTargetBitrate = avctx->bit_rate;
95 param.iMaxBitrate = FFMAX(avctx->rc_max_rate, avctx->bit_rate);
96 param.iRCMode = RC_QUALITY_MODE;
97 param.iTemporalLayerNum = 1;
98 param.iSpatialLayerNum = 1;
99 param.bEnableDenoise = 0;
100 param.bEnableBackgroundDetection = 1;
101 param.bEnableAdaptiveQuant = 1;
102 param.bEnableFrameSkip = 0;
103 param.bEnableLongTermReference = 0;
104 param.iLtrMarkPeriod = 30;
105 param.uiIntraPeriod = avctx->gop_size;
106 param.bEnableSpsPpsIdAddition = 0;
107 param.bPrefixNalAddingCtrl = 0;
108 param.iLoopFilterDisableIdc = !s->loopfilter;
109 param.iEntropyCodingModeFlag = 0;
110 param.iMultipleThreadIdc = avctx->thread_count;
111 if (s->profile && !strcmp(s->profile, "main"))
112 param.iEntropyCodingModeFlag = 1;
113 else if (!s->profile && avctx->coder_type == FF_CODER_TYPE_AC)
114 param.iEntropyCodingModeFlag = 1;
116 param.sSpatialLayers[0].iVideoWidth = param.iPicWidth;
117 param.sSpatialLayers[0].iVideoHeight = param.iPicHeight;
118 param.sSpatialLayers[0].fFrameRate = param.fMaxFrameRate;
119 param.sSpatialLayers[0].iSpatialBitrate = param.iTargetBitrate;
120 param.sSpatialLayers[0].iMaxSpatialBitrate = param.iMaxBitrate;
122 if (avctx->slices > 1)
123 s->slice_mode = SM_FIXEDSLCNUM_SLICE;
124 param.sSpatialLayers[0].sSliceCfg.uiSliceMode = s->slice_mode;
125 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceNum = avctx->slices;
127 if ((*s->encoder)->InitializeExt(s->encoder, ¶m) != cmResultSuccess) {
128 av_log(avctx, AV_LOG_ERROR, "Initialize failed\n");
132 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
133 SFrameBSInfo fbi = { 0 };
135 (*s->encoder)->EncodeParameterSets(s->encoder, &fbi);
136 for (i = 0; i < fbi.sLayerInfo[0].iNalCount; i++)
137 size += fbi.sLayerInfo[0].pNalLengthInByte[i];
138 avctx->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
139 if (!avctx->extradata) {
140 err = AVERROR(ENOMEM);
143 avctx->extradata_size = size;
144 memcpy(avctx->extradata, fbi.sLayerInfo[0].pBsBuf, size);
150 svc_encode_close(avctx);
154 static int svc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
155 const AVFrame *frame, int *got_packet)
157 SVCContext *s = avctx->priv_data;
158 SFrameBSInfo fbi = { 0 };
161 SSourcePicture sp = { 0 };
162 int size = 0, layer, first_layer = 0;
163 int layer_size[MAX_LAYER_NUM_OF_FRAME] = { 0 };
165 sp.iColorFormat = videoFormatI420;
166 for (i = 0; i < 3; i++) {
167 sp.iStride[i] = frame->linesize[i];
168 sp.pData[i] = frame->data[i];
170 sp.iPicWidth = avctx->width;
171 sp.iPicHeight = avctx->height;
173 encoded = (*s->encoder)->EncodeFrame(s->encoder, &sp, &fbi);
174 if (encoded != cmResultSuccess) {
175 av_log(avctx, AV_LOG_ERROR, "EncodeFrame failed\n");
176 return AVERROR_UNKNOWN;
178 if (fbi.eFrameType == videoFrameTypeSkip) {
179 av_log(avctx, AV_LOG_DEBUG, "frame skipped\n");
183 // Normal frames are returned with one single layer, while IDR
184 // frames have two layers, where the first layer contains the SPS/PPS.
185 // If using global headers, don't include the SPS/PPS in the returned
186 // packet - thus, only return one layer.
187 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
188 first_layer = fbi.iLayerNum - 1;
190 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
191 for (i = 0; i < fbi.sLayerInfo[layer].iNalCount; i++)
192 layer_size[layer] += fbi.sLayerInfo[layer].pNalLengthInByte[i];
193 size += layer_size[layer];
195 av_log(avctx, AV_LOG_DEBUG, "%d slices\n", fbi.sLayerInfo[fbi.iLayerNum - 1].iNalCount);
197 if ((ret = ff_alloc_packet(avpkt, size))) {
198 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
202 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
203 memcpy(avpkt->data + size, fbi.sLayerInfo[layer].pBsBuf, layer_size[layer]);
204 size += layer_size[layer];
206 avpkt->pts = frame->pts;
207 if (fbi.eFrameType == videoFrameTypeIDR)
208 avpkt->flags |= AV_PKT_FLAG_KEY;
213 AVCodec ff_libopenh264_encoder = {
214 .name = "libopenh264",
215 .long_name = NULL_IF_CONFIG_SMALL("OpenH264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
216 .type = AVMEDIA_TYPE_VIDEO,
217 .id = AV_CODEC_ID_H264,
218 .priv_data_size = sizeof(SVCContext),
219 .init = svc_encode_init,
220 .encode2 = svc_encode_frame,
221 .close = svc_encode_close,
222 .capabilities = CODEC_CAP_AUTO_THREADS,
223 .pix_fmts = (const enum PixelFormat[]){ AV_PIX_FMT_YUV420P,
225 .priv_class = &class,