Message ID | 1543729129-26319-2-git-send-email-mypopydev@gmail.com |
---|---|
State | Superseded |
Headers | show |
> On Dec 2, 2018, at 13:38, Jun Zhao <mypopydev@gmail.com> wrote: > > base on patch by Huang, Zhengxu from https://github.com/intel/SVT-HEVC > > Signed-off-by: Huang, Zhengxu <zhengxu.huang@intel.com> > Signed-off-by: hassene <hassene.tmar@intel.com> > Signed-off-by: Jun Zhao <jun.zhao@intel.com> > --- > configure | 4 + > libavcodec/Makefile | 1 + > libavcodec/allcodecs.c | 1 + > libavcodec/libsvt_hevc.c | 370 ++++++++++++++++++++++++++++++++++++++++++++++ > 4 files changed, 376 insertions(+), 0 deletions(-) > create mode 100644 libavcodec/libsvt_hevc.c > > diff --git a/configure b/configure > index 54b7e11..c3dca73 100755 > --- a/configure > +++ b/configure > @@ -263,6 +263,7 @@ External library support: > --enable-libspeex enable Speex de/encoding via libspeex [no] > --enable-libsrt enable Haivision SRT protocol via libsrt [no] > --enable-libssh enable SFTP protocol via libssh [no] > + --enable-libsvt enable HEVC encoding via svt [no] > --enable-libtensorflow enable TensorFlow as a DNN module backend > for DNN based filters like sr [no] > --enable-libtesseract enable Tesseract, needed for ocr filter [no] > @@ -1665,6 +1666,7 @@ EXTERNAL_LIBRARY_GPL_LIST=" > libcdio > libdavs2 > librubberband > + libsvt > libvidstab > libx264 > libx265 > @@ -3087,6 +3089,7 @@ pcm_mulaw_at_encoder_select="audio_frame_queue" > chromaprint_muxer_deps="chromaprint" > h264_videotoolbox_encoder_deps="pthreads" > h264_videotoolbox_encoder_select="videotoolbox_encoder" > +hevc_svt_encoder_deps="libsvt" > hevc_videotoolbox_encoder_deps="pthreads" > hevc_videotoolbox_encoder_select="videotoolbox_encoder" > libaom_av1_decoder_deps="libaom" > @@ -6147,6 +6150,7 @@ enabled libsoxr && require libsoxr soxr.h soxr_create -lsoxr > enabled libssh && require_pkg_config libssh libssh libssh/sftp.h sftp_init > enabled libspeex && require_pkg_config libspeex speex speex/speex.h speex_decoder_init > enabled libsrt && require_pkg_config libsrt "srt >= 1.3.0" srt/srt.h srt_socket > +enabled libsvt && require_pkg_config libsvt svt EbApi.h EbInitHandle > enabled libtensorflow && require libtensorflow tensorflow/c/c_api.h TF_Version -ltensorflow > enabled libtesseract && require_pkg_config libtesseract tesseract tesseract/capi.h TessBaseAPICreate > enabled libtheora && require libtheora theora/theoraenc.h th_info_init -ltheoraenc -ltheoradec -logg > diff --git a/libavcodec/Makefile b/libavcodec/Makefile > index 5feadac..1a8b547 100644 > --- a/libavcodec/Makefile > +++ b/libavcodec/Makefile > @@ -374,6 +374,7 @@ OBJS-$(CONFIG_HEVC_QSV_DECODER) += qsvdec_h2645.o > OBJS-$(CONFIG_HEVC_QSV_ENCODER) += qsvenc_hevc.o hevc_ps_enc.o \ > hevc_data.o > OBJS-$(CONFIG_HEVC_RKMPP_DECODER) += rkmppdec.o > +OBJS-$(CONFIG_HEVC_SVT_ENCODER) += libsvt_hevc.o > OBJS-$(CONFIG_HEVC_VAAPI_ENCODER) += vaapi_encode_h265.o h265_profile_level.o > OBJS-$(CONFIG_HEVC_V4L2M2M_DECODER) += v4l2_m2m_dec.o > OBJS-$(CONFIG_HEVC_V4L2M2M_ENCODER) += v4l2_m2m_enc.o > diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c > index d70646e..ff8ea94 100644 > --- a/libavcodec/allcodecs.c > +++ b/libavcodec/allcodecs.c > @@ -749,6 +749,7 @@ extern AVCodec ff_hevc_cuvid_decoder; > extern AVCodec ff_hevc_mediacodec_decoder; > extern AVCodec ff_hevc_nvenc_encoder; > extern AVCodec ff_hevc_qsv_encoder; > +extern AVCodec ff_hevc_svt_encoder; > extern AVCodec ff_hevc_v4l2m2m_encoder; > extern AVCodec ff_hevc_vaapi_encoder; > extern AVCodec ff_hevc_videotoolbox_encoder; > diff --git a/libavcodec/libsvt_hevc.c b/libavcodec/libsvt_hevc.c > new file mode 100644 > index 0000000..1e10111 > --- /dev/null > +++ b/libavcodec/libsvt_hevc.c > @@ -0,0 +1,370 @@ > +/* > +* Scalable Video Technology for HEVC encoder library plugin > +* > +* Copyright (c) 2018 Intel Corporation > +* > +* This program is free software; you can redistribute it and/or > +* modify it under the terms of the GNU Lesser General Public > +* License as published by the Free Software Foundation; either > +* version 2.1 of the License, or (at your option) any later version. > +* > +* This program is distributed in the hope that it will be useful, > +* but WITHOUT ANY WARRANTY; without even the implied warranty of > +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > +* Lesser General Public License for more details. > +* > +* You should have received a copy of the GNU Lesser General Public > +* License along with this program; if not, write to the Free Software > +* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA > +*/ > + > +#include "EbErrorCodes.h" > +#include "EbTime.h" > +#include "EbApi.h" > + > +#include "libavutil/common.h" > +#include "libavutil/frame.h" > +#include "libavutil/opt.h" > + > +#include "internal.h" > +#include "avcodec.h" > + > +typedef struct SvtEncoder { > + EB_H265_ENC_CONFIGURATION enc_params; > + EB_COMPONENTTYPE *svt_handle; > + EB_BUFFERHEADERTYPE *in_buf; > + EB_BUFFERHEADERTYPE *out_buf; > + int raw_size; > +} SvtEncoder; > + > +typedef struct SvtParams { > + int vui_info; > + int hierarchical_level; > + int la_depth; > + int intra_ref_type; > + int enc_mode; > + int rc_mode; > + int scd; > + int tune; > + int qp; > + int profile; > + int base_layer_switch_mode; > +}SvtParams; > + > +typedef struct SvtContext { > + AVClass *class; > + SvtEncoder *svt_enc; > + SvtParams svt_param; > + int eos_flag; > +} SvtContext; > + > +static void free_buffer(SvtEncoder *svt_enc) > +{ > + if (svt_enc->in_buf) { > + EB_H265_ENC_INPUT *in_data = (EB_H265_ENC_INPUT *)svt_enc->in_buf->pBuffer; > + av_freep(&in_data); > + av_freep(&svt_enc->in_buf); > + } > + av_freep(&svt_enc->out_buf); > +} > + > +static EB_ERRORTYPE alloc_buffer(EB_H265_ENC_CONFIGURATION *config, SvtEncoder *svt_enc) > +{ > + EB_ERRORTYPE ret = EB_ErrorNone; > + > + const int pack_mode_10bit = (config->encoderBitDepth > 8) && (config->compressedTenBitFormat == 0) ? 1 : 0; > + const size_t luma_size_8bit = config->sourceWidth * config->sourceHeight * (1 << pack_mode_10bit); > + const size_t luma_size_10bit = (config->encoderBitDepth > 8 && pack_mode_10bit == 0) ? luma_size_8bit : 0; > + > + svt_enc->raw_size = (luma_size_8bit + luma_size_10bit) * 3 / 2; > + > + // allocate buffer for in and out > + svt_enc->in_buf = av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); > + svt_enc->out_buf = av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); > + if (!svt_enc->in_buf || !svt_enc->out_buf) > + goto failed; > + > + svt_enc->in_buf->pBuffer = av_mallocz(sizeof(EB_H265_ENC_INPUT)); > + if (!svt_enc->in_buf->pBuffer) > + goto failed; > + > + svt_enc->in_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); > + svt_enc->in_buf->pAppPrivate = NULL; > + svt_enc->out_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); > + svt_enc->out_buf->nAllocLen = svt_enc->raw_size; > + svt_enc->out_buf->pAppPrivate = NULL; > + > + return ret; > + > +failed: > + free_buffer(svt_enc); > + return AVERROR(ENOMEM); > +} > + > +static int error_mapping(int val) > +{ > + if (val == EB_ErrorInsufficientResources) > + return AVERROR(ENOMEM); > + if ((val == EB_ErrorUndefined) || (val == EB_ErrorInvalidComponent) || > + (val == EB_ErrorBadParameter)) > + return AVERROR(EINVAL); > + return AVERROR_EXTERNAL; > +} > + > +static EB_ERRORTYPE config_enc_params(EB_H265_ENC_CONFIGURATION *param, AVCodecContext *avctx) > +{ > + SvtContext *q = avctx->priv_data; > + SvtEncoder *svt_enc = q->svt_enc; > + EB_ERRORTYPE ret = EB_ErrorNone; > + int tenBits = 0; Camelcase ? > + > + param->sourceWidth = avctx->width; > + param->sourceHeight = avctx->height; > + > + if (avctx->pix_fmt == AV_PIX_FMT_YUV420P10LE) { > + av_log(avctx, AV_LOG_DEBUG , "Encoder 10 bits depth input\n"); > + param->compressedTenBitFormat = 0; > + tenBits = 1; > + } > + maybe need check q->svt_param first. > > + // Update param from options > + param->hierarchicalLevels = q->svt_param.hierarchical_level; > + param->encMode = q->svt_param.enc_mode; > + param->intraRefreshType = q->svt_param.intra_ref_type; > + param->profile = q->svt_param.profile; > + param->rateControlMode = q->svt_param.rc_mode; > + param->sceneChangeDetection = q->svt_param.scd; > + param->tune = q->svt_param.tune; > + param->baseLayerSwitchMode = q->svt_param.base_layer_switch_mode; > + param->qp = q->svt_param.qp; > + > + param->targetBitRate = avctx->bit_rate; > + param->intraPeriodLength = avctx->gop_size-1; > + param->frameRateNumerator = avctx->time_base.den; > + param->frameRateDenominator = avctx->time_base.num * avctx->ticks_per_frame; > + > + param->codeVpsSpsPps = 0; > + > + if (q->svt_param.vui_info) > + param->videoUsabilityInfo = q->svt_param.vui_info; > + if (q->svt_param.la_depth != -1) > + param->lookAheadDistance = q->svt_param.la_depth; > + > + if (tenBits == 1) { > + param->encoderBitDepth = 10; > + param->profile = 2; > + } > + > + ret = alloc_buffer(param, svt_enc); > + > + return ret; > +} > + > +static void read_in_data(EB_H265_ENC_CONFIGURATION *config, const AVFrame* frame, EB_BUFFERHEADERTYPE *headerPtr) > +{ > + unsigned int is16bit = config->encoderBitDepth > 8; > + unsigned long long lumaReadSize = (unsigned long long)config->sourceWidth * config->sourceHeight<< is16bit; what about use uint64_t? > + EB_H265_ENC_INPUT *in_data = (EB_H265_ENC_INPUT*)headerPtr->pBuffer; > + > + // support yuv420p and yuv420p010 > + in_data->luma = frame->data[0]; > + in_data->cb = frame->data[1]; > + in_data->cr = frame->data[2]; > + > + // stride info > + in_data->yStride = frame->linesize[0] >> is16bit; > + in_data->cbStride = frame->linesize[1] >> is16bit; > + in_data->crStride = frame->linesize[2] >> is16bit; > + > + headerPtr->nFilledLen += lumaReadSize * 3/2u; > +} > + > +static av_cold int eb_enc_init(AVCodecContext *avctx) > +{ > + SvtContext *q = avctx->priv_data; > + SvtEncoder *svt_enc = NULL; > + EB_ERRORTYPE ret = EB_ErrorNone; > + > + q->svt_enc = av_mallocz(sizeof(*q->svt_enc)); > + if (!q->svt_enc) > + return AVERROR(ENOMEM); > + svt_enc = q->svt_enc; > + > + q->eos_flag = 0; > + > + ret = EbInitHandle(&svt_enc->svt_handle, q, &svt_enc->enc_params); > + if (ret != EB_ErrorNone) > + goto failed_init; > + > + ret = config_enc_params(&svt_enc->enc_params, avctx); > + if (ret != EB_ErrorNone) > + goto failed_init; > + > + ret = EbH265EncSetParameter(svt_enc->svt_handle, &svt_enc->enc_params); > + if (ret != EB_ErrorNone) > + goto failed_init; > + > + ret = EbInitEncoder(svt_enc->svt_handle); > + if (ret != EB_ErrorNone) > + goto failed_init; > + > + if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) { > + > + EB_BUFFERHEADERTYPE headerPtr; > + headerPtr.nSize = sizeof(EB_BUFFERHEADERTYPE); > + headerPtr.nFilledLen = 0; > + headerPtr.pBuffer = av_malloc(10 * 1024 * 1024); > + headerPtr.nAllocLen = (10 * 1024 * 1024); > + > + if (!headerPtr.pBuffer) > + return AVERROR(ENOMEM); > + > + ret = EbH265EncStreamHeader(svt_enc->svt_handle, &headerPtr); > + if (ret != EB_ErrorNone) { > + av_freep(&headerPtr.pBuffer); > + goto failed_init; > + } > + avctx->extradata_size = headerPtr.nFilledLen; > + avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE); > + if (!avctx->extradata) { > + av_log(avctx, AV_LOG_ERROR, > + "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size); > + return AVERROR(ENOMEM); > + } > + memcpy(avctx->extradata, headerPtr.pBuffer, avctx->extradata_size); > + av_freep(&headerPtr.pBuffer); > + } > + return 0; > + > +failed_init: add av_free(q->svt_enc) ? > + return error_mapping(ret); > +} > + > +static int eb_send_frame(AVCodecContext *avctx, const AVFrame *frame) > +{ > + SvtContext *q = avctx->priv_data; > + SvtEncoder *svt_enc = q->svt_enc; > + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->in_buf; > + int ret = 0; > + > + if (!frame) { > + EB_BUFFERHEADERTYPE headerPtrLast; > + headerPtrLast.nAllocLen = 0; > + headerPtrLast.nFilledLen = 0; > + headerPtrLast.nTickCount = 0; > + headerPtrLast.pAppPrivate = NULL; > + headerPtrLast.nOffset = 0; > + headerPtrLast.pBuffer = NULL; > + headerPtrLast.nFlags = EB_BUFFERFLAG_EOS; > + EbH265EncSendPicture(svt_enc->svt_handle, &headerPtrLast); > + q->eos_flag = 1; > + av_log(avctx, AV_LOG_DEBUG, "Finish sending frames!!!\n"); > + return ret; > + } > + > + read_in_data(&svt_enc->enc_params, frame, headerPtr); > + > + headerPtr->nOffset = 0; > + headerPtr->nFlags = 0; > + headerPtr->pAppPrivate = NULL; > + headerPtr->pts = frame->pts; > + headerPtr->sliceType = INVALID_SLICE; > + EbH265EncSendPicture(svt_enc->svt_handle, headerPtr); > + > + return ret; > +} > + > +static int eb_receive_packet(AVCodecContext *avctx, AVPacket *pkt) > +{ > + SvtContext *q = avctx->priv_data; > + SvtEncoder *svt_enc = q->svt_enc; > + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->out_buf; > + EB_ERRORTYPE stream_status = EB_ErrorNone; > + int ret = 0; > + > + if ((ret = ff_alloc_packet2(avctx, pkt, svt_enc->raw_size, 0)) < 0) { > + av_log(avctx, AV_LOG_ERROR, "Failed to allocate output packet.\n"); > + return ret; > + } > + headerPtr->pBuffer = pkt->data; > + stream_status = EbH265GetPacket(svt_enc->svt_handle, headerPtr, q->eos_flag); > + if (stream_status == EB_NoErrorEmptyQueue) > + return AVERROR(EAGAIN); > + > + pkt->size = headerPtr->nFilledLen; > + pkt->pts = headerPtr->pts; > + pkt->dts = headerPtr->dts; > + if (headerPtr->sliceType == IDR_SLICE) > + pkt->flags |= AV_PKT_FLAG_KEY; > + if (headerPtr->sliceType == NON_REF_SLICE) > + pkt->flags |= AV_PKT_FLAG_DISPOSABLE; > + > + ret = (headerPtr->nFlags & EB_BUFFERFLAG_EOS) ? AVERROR_EOF : 0; > + return ret; > +} > + > +static av_cold int eb_enc_close(AVCodecContext *avctx) > +{ > + SvtContext *q = avctx->priv_data; > + SvtEncoder *svt_enc = q->svt_enc; > + > + EbDeinitEncoder(svt_enc->svt_handle); > + EbDeinitHandle(svt_enc->svt_handle); > + > + free_buffer(svt_enc); > + av_freep(&svt_enc); > + > + return 0; > +} > + > +#define OFFSET(x) offsetof(SvtContext, x) > +#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM > +static const AVOption options[] = { > + {"vui", "Enable vui info", OFFSET(svt_param.vui_info), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > + {"hielevel", "Hierarchical Prediction Levels [0,3]", OFFSET(svt_param.hierarchical_level), AV_OPT_TYPE_INT, { .i64 = 3 }, 0, 3, VE }, > + {"la_depth", "Look Ahead Distance [0,256]", OFFSET(svt_param.la_depth), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 256, VE }, > + {"intra_ref_type", "Intra Refresh Type 0: No intra refresh 1: CRA (Open GOP) 2: IDR", OFFSET(svt_param.intra_ref_type), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 2, VE }, > + {"enc_p", "Encoding preset [0,12] (for tune 0 and >=4k resolution), [0,10] (for >= 1080p resolution), [0,9] (for all resolution and modes)", OFFSET(svt_param.enc_mode), AV_OPT_TYPE_INT, { .i64 = 9 }, 0, 12, VE }, > + {"profile", "Profile now support[1,2], Main Still Picture Profile not supported", OFFSET(svt_param.profile), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, 2, VE }, > + {"rc", "RC mode 0: CQP 1: VBR", OFFSET(svt_param.rc_mode), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > + {"q", "QP value for intra frames", OFFSET(svt_param.qp), AV_OPT_TYPE_INT, { .i64 = 32 }, 0, 51, VE }, > + {"scd", "Scene change detection", OFFSET(svt_param.scd), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > + {"tune", "Tune mode: SQ/OQ[0,1]", OFFSET(svt_param.tune), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > + {"bl_mode", "Random Access Prediction Structure Type", OFFSET(svt_param.base_layer_switch_mode), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > + {NULL}, > +}; Need Document. > + > +static const AVClass class = { > + .class_name = "libsvt_hevc", > + .item_name = av_default_item_name, > + .option = options, > + .version = LIBAVUTIL_VERSION_INT, > +}; > + > +static const AVCodecDefault eb_enc_defaults[] = { > + { "b", "7M" }, > + { "refs", "0" }, > + { "g", "64" }, > + { "flags", "+cgop" }, > + { NULL }, > +}; > + > +AVCodec ff_hevc_svt_encoder = { > + .name = "libsvt_hevc", > + .long_name = NULL_IF_CONFIG_SMALL("SVT-HEVC(Scalable Video Technology for HEVC) encoder"), > + .priv_data_size = sizeof(SvtContext), > + .type = AVMEDIA_TYPE_VIDEO, > + .id = AV_CODEC_ID_HEVC, > + .init = eb_enc_init, > + .send_frame = eb_send_frame, > + .receive_packet = eb_receive_packet, > + .close = eb_enc_close, > + .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS, > + .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, > + AV_PIX_FMT_YUV420P10, > + AV_PIX_FMT_NONE }, > + .priv_class = &class, > + .defaults = eb_enc_defaults, > + .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, > + .wrapper_name = "libsvt_hevc", > +}; > -- > 1.7.1 > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel Thanks Steven
On 12/2/2018 7:24 AM, Steven Liu wrote: > > >> On Dec 2, 2018, at 13:38, Jun Zhao <mypopydev@gmail.com> wrote: >> >> base on patch by Huang, Zhengxu from https://github.com/intel/SVT-HEVC >> >> Signed-off-by: Huang, Zhengxu <zhengxu.huang@intel.com> >> Signed-off-by: hassene <hassene.tmar@intel.com> >> Signed-off-by: Jun Zhao <jun.zhao@intel.com> >> --- >> configure | 4 + >> libavcodec/Makefile | 1 + >> libavcodec/allcodecs.c | 1 + >> libavcodec/libsvt_hevc.c | 370 ++++++++++++++++++++++++++++++++++++++++++++++ >> 4 files changed, 376 insertions(+), 0 deletions(-) >> create mode 100644 libavcodec/libsvt_hevc.c >> >> diff --git a/configure b/configure >> index 54b7e11..c3dca73 100755 >> --- a/configure >> +++ b/configure >> @@ -263,6 +263,7 @@ External library support: >> --enable-libspeex enable Speex de/encoding via libspeex [no] >> --enable-libsrt enable Haivision SRT protocol via libsrt [no] >> --enable-libssh enable SFTP protocol via libssh [no] >> + --enable-libsvt enable HEVC encoding via svt [no] >> --enable-libtensorflow enable TensorFlow as a DNN module backend >> for DNN based filters like sr [no] >> --enable-libtesseract enable Tesseract, needed for ocr filter [no] >> @@ -1665,6 +1666,7 @@ EXTERNAL_LIBRARY_GPL_LIST=" >> libcdio >> libdavs2 >> librubberband >> + libsvt >> libvidstab >> libx264 >> libx265 >> @@ -3087,6 +3089,7 @@ pcm_mulaw_at_encoder_select="audio_frame_queue" >> chromaprint_muxer_deps="chromaprint" >> h264_videotoolbox_encoder_deps="pthreads" >> h264_videotoolbox_encoder_select="videotoolbox_encoder" >> +hevc_svt_encoder_deps="libsvt" >> hevc_videotoolbox_encoder_deps="pthreads" >> hevc_videotoolbox_encoder_select="videotoolbox_encoder" >> libaom_av1_decoder_deps="libaom" >> @@ -6147,6 +6150,7 @@ enabled libsoxr && require libsoxr soxr.h soxr_create -lsoxr >> enabled libssh && require_pkg_config libssh libssh libssh/sftp.h sftp_init >> enabled libspeex && require_pkg_config libspeex speex speex/speex.h speex_decoder_init >> enabled libsrt && require_pkg_config libsrt "srt >= 1.3.0" srt/srt.h srt_socket >> +enabled libsvt && require_pkg_config libsvt svt EbApi.h EbInitHandle >> enabled libtensorflow && require libtensorflow tensorflow/c/c_api.h TF_Version -ltensorflow >> enabled libtesseract && require_pkg_config libtesseract tesseract tesseract/capi.h TessBaseAPICreate >> enabled libtheora && require libtheora theora/theoraenc.h th_info_init -ltheoraenc -ltheoradec -logg >> diff --git a/libavcodec/Makefile b/libavcodec/Makefile >> index 5feadac..1a8b547 100644 >> --- a/libavcodec/Makefile >> +++ b/libavcodec/Makefile >> @@ -374,6 +374,7 @@ OBJS-$(CONFIG_HEVC_QSV_DECODER) += qsvdec_h2645.o >> OBJS-$(CONFIG_HEVC_QSV_ENCODER) += qsvenc_hevc.o hevc_ps_enc.o \ >> hevc_data.o >> OBJS-$(CONFIG_HEVC_RKMPP_DECODER) += rkmppdec.o >> +OBJS-$(CONFIG_HEVC_SVT_ENCODER) += libsvt_hevc.o >> OBJS-$(CONFIG_HEVC_VAAPI_ENCODER) += vaapi_encode_h265.o h265_profile_level.o >> OBJS-$(CONFIG_HEVC_V4L2M2M_DECODER) += v4l2_m2m_dec.o >> OBJS-$(CONFIG_HEVC_V4L2M2M_ENCODER) += v4l2_m2m_enc.o >> diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c >> index d70646e..ff8ea94 100644 >> --- a/libavcodec/allcodecs.c >> +++ b/libavcodec/allcodecs.c >> @@ -749,6 +749,7 @@ extern AVCodec ff_hevc_cuvid_decoder; >> extern AVCodec ff_hevc_mediacodec_decoder; >> extern AVCodec ff_hevc_nvenc_encoder; >> extern AVCodec ff_hevc_qsv_encoder; >> +extern AVCodec ff_hevc_svt_encoder; >> extern AVCodec ff_hevc_v4l2m2m_encoder; >> extern AVCodec ff_hevc_vaapi_encoder; >> extern AVCodec ff_hevc_videotoolbox_encoder; >> diff --git a/libavcodec/libsvt_hevc.c b/libavcodec/libsvt_hevc.c >> new file mode 100644 >> index 0000000..1e10111 >> --- /dev/null >> +++ b/libavcodec/libsvt_hevc.c >> @@ -0,0 +1,370 @@ >> +/* >> +* Scalable Video Technology for HEVC encoder library plugin >> +* >> +* Copyright (c) 2018 Intel Corporation >> +* >> +* This program is free software; you can redistribute it and/or >> +* modify it under the terms of the GNU Lesser General Public >> +* License as published by the Free Software Foundation; either >> +* version 2.1 of the License, or (at your option) any later version. >> +* >> +* This program is distributed in the hope that it will be useful, >> +* but WITHOUT ANY WARRANTY; without even the implied warranty of >> +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU >> +* Lesser General Public License for more details. >> +* >> +* You should have received a copy of the GNU Lesser General Public >> +* License along with this program; if not, write to the Free Software >> +* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA >> +*/ >> + >> +#include "EbErrorCodes.h" >> +#include "EbTime.h" >> +#include "EbApi.h" >> + >> +#include "libavutil/common.h" >> +#include "libavutil/frame.h" >> +#include "libavutil/opt.h" >> + >> +#include "internal.h" >> +#include "avcodec.h" >> + >> +typedef struct SvtEncoder { >> + EB_H265_ENC_CONFIGURATION enc_params; >> + EB_COMPONENTTYPE *svt_handle; >> + EB_BUFFERHEADERTYPE *in_buf; >> + EB_BUFFERHEADERTYPE *out_buf; >> + int raw_size; >> +} SvtEncoder; >> + >> +typedef struct SvtParams { >> + int vui_info; >> + int hierarchical_level; >> + int la_depth; >> + int intra_ref_type; >> + int enc_mode; >> + int rc_mode; >> + int scd; >> + int tune; >> + int qp; >> + int profile; >> + int base_layer_switch_mode; >> +}SvtParams; >> + >> +typedef struct SvtContext { >> + AVClass *class; >> + SvtEncoder *svt_enc; >> + SvtParams svt_param; >> + int eos_flag; >> +} SvtContext; >> + >> +static void free_buffer(SvtEncoder *svt_enc) >> +{ >> + if (svt_enc->in_buf) { >> + EB_H265_ENC_INPUT *in_data = (EB_H265_ENC_INPUT *)svt_enc->in_buf->pBuffer; >> + av_freep(&in_data); >> + av_freep(&svt_enc->in_buf); >> + } >> + av_freep(&svt_enc->out_buf); >> +} >> + >> +static EB_ERRORTYPE alloc_buffer(EB_H265_ENC_CONFIGURATION *config, SvtEncoder *svt_enc) >> +{ >> + EB_ERRORTYPE ret = EB_ErrorNone; >> + >> + const int pack_mode_10bit = (config->encoderBitDepth > 8) && (config->compressedTenBitFormat == 0) ? 1 : 0; >> + const size_t luma_size_8bit = config->sourceWidth * config->sourceHeight * (1 << pack_mode_10bit); >> + const size_t luma_size_10bit = (config->encoderBitDepth > 8 && pack_mode_10bit == 0) ? luma_size_8bit : 0; >> + >> + svt_enc->raw_size = (luma_size_8bit + luma_size_10bit) * 3 / 2; >> + >> + // allocate buffer for in and out >> + svt_enc->in_buf = av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); >> + svt_enc->out_buf = av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); >> + if (!svt_enc->in_buf || !svt_enc->out_buf) >> + goto failed; >> + >> + svt_enc->in_buf->pBuffer = av_mallocz(sizeof(EB_H265_ENC_INPUT)); >> + if (!svt_enc->in_buf->pBuffer) >> + goto failed; >> + >> + svt_enc->in_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); >> + svt_enc->in_buf->pAppPrivate = NULL; >> + svt_enc->out_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); >> + svt_enc->out_buf->nAllocLen = svt_enc->raw_size; >> + svt_enc->out_buf->pAppPrivate = NULL; >> + >> + return ret; >> + >> +failed: >> + free_buffer(svt_enc); >> + return AVERROR(ENOMEM); >> +} >> + >> +static int error_mapping(int val) >> +{ >> + if (val == EB_ErrorInsufficientResources) >> + return AVERROR(ENOMEM); >> + if ((val == EB_ErrorUndefined) || (val == EB_ErrorInvalidComponent) || >> + (val == EB_ErrorBadParameter)) >> + return AVERROR(EINVAL); >> + return AVERROR_EXTERNAL; >> +} >> + >> +static EB_ERRORTYPE config_enc_params(EB_H265_ENC_CONFIGURATION *param, AVCodecContext *avctx) >> +{ >> + SvtContext *q = avctx->priv_data; >> + SvtEncoder *svt_enc = q->svt_enc; >> + EB_ERRORTYPE ret = EB_ErrorNone; >> + int tenBits = 0; > Camelcase ? >> + >> + param->sourceWidth = avctx->width; >> + param->sourceHeight = avctx->height; >> + >> + if (avctx->pix_fmt == AV_PIX_FMT_YUV420P10LE) { >> + av_log(avctx, AV_LOG_DEBUG , "Encoder 10 bits depth input\n"); >> + param->compressedTenBitFormat = 0; >> + tenBits = 1; >> + } >> + > maybe need check q->svt_param first. >> >> + // Update param from options >> + param->hierarchicalLevels = q->svt_param.hierarchical_level; >> + param->encMode = q->svt_param.enc_mode; >> + param->intraRefreshType = q->svt_param.intra_ref_type; >> + param->profile = q->svt_param.profile; >> + param->rateControlMode = q->svt_param.rc_mode; >> + param->sceneChangeDetection = q->svt_param.scd; >> + param->tune = q->svt_param.tune; >> + param->baseLayerSwitchMode = q->svt_param.base_layer_switch_mode; >> + param->qp = q->svt_param.qp; >> + >> + param->targetBitRate = avctx->bit_rate; >> + param->intraPeriodLength = avctx->gop_size-1; >> + param->frameRateNumerator = avctx->time_base.den; >> + param->frameRateDenominator = avctx->time_base.num * avctx->ticks_per_frame; >> + >> + param->codeVpsSpsPps = 0; >> + >> + if (q->svt_param.vui_info) >> + param->videoUsabilityInfo = q->svt_param.vui_info; >> + if (q->svt_param.la_depth != -1) >> + param->lookAheadDistance = q->svt_param.la_depth; >> + >> + if (tenBits == 1) { >> + param->encoderBitDepth = 10; >> + param->profile = 2; >> + } >> + >> + ret = alloc_buffer(param, svt_enc); >> + >> + return ret; >> +} >> + >> +static void read_in_data(EB_H265_ENC_CONFIGURATION *config, const AVFrame* frame, EB_BUFFERHEADERTYPE *headerPtr) >> +{ >> + unsigned int is16bit = config->encoderBitDepth > 8; >> + unsigned long long lumaReadSize = (unsigned long long)config->sourceWidth * config->sourceHeight<< is16bit; > what about use uint64_t? >> + EB_H265_ENC_INPUT *in_data = (EB_H265_ENC_INPUT*)headerPtr->pBuffer; >> + >> + // support yuv420p and yuv420p010 >> + in_data->luma = frame->data[0]; >> + in_data->cb = frame->data[1]; >> + in_data->cr = frame->data[2]; >> + >> + // stride info >> + in_data->yStride = frame->linesize[0] >> is16bit; >> + in_data->cbStride = frame->linesize[1] >> is16bit; >> + in_data->crStride = frame->linesize[2] >> is16bit; >> + >> + headerPtr->nFilledLen += lumaReadSize * 3/2u; >> +} >> + >> +static av_cold int eb_enc_init(AVCodecContext *avctx) >> +{ >> + SvtContext *q = avctx->priv_data; >> + SvtEncoder *svt_enc = NULL; >> + EB_ERRORTYPE ret = EB_ErrorNone; >> + >> + q->svt_enc = av_mallocz(sizeof(*q->svt_enc)); >> + if (!q->svt_enc) >> + return AVERROR(ENOMEM); >> + svt_enc = q->svt_enc; >> + >> + q->eos_flag = 0; >> + >> + ret = EbInitHandle(&svt_enc->svt_handle, q, &svt_enc->enc_params); >> + if (ret != EB_ErrorNone) >> + goto failed_init; >> + >> + ret = config_enc_params(&svt_enc->enc_params, avctx); >> + if (ret != EB_ErrorNone) >> + goto failed_init; >> + >> + ret = EbH265EncSetParameter(svt_enc->svt_handle, &svt_enc->enc_params); >> + if (ret != EB_ErrorNone) >> + goto failed_init; >> + >> + ret = EbInitEncoder(svt_enc->svt_handle); >> + if (ret != EB_ErrorNone) >> + goto failed_init; >> + >> + if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) { >> + >> + EB_BUFFERHEADERTYPE headerPtr; >> + headerPtr.nSize = sizeof(EB_BUFFERHEADERTYPE); >> + headerPtr.nFilledLen = 0; >> + headerPtr.pBuffer = av_malloc(10 * 1024 * 1024); >> + headerPtr.nAllocLen = (10 * 1024 * 1024); >> + >> + if (!headerPtr.pBuffer) >> + return AVERROR(ENOMEM); >> + >> + ret = EbH265EncStreamHeader(svt_enc->svt_handle, &headerPtr); >> + if (ret != EB_ErrorNone) { >> + av_freep(&headerPtr.pBuffer); >> + goto failed_init; >> + } >> + avctx->extradata_size = headerPtr.nFilledLen; >> + avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE); >> + if (!avctx->extradata) { >> + av_log(avctx, AV_LOG_ERROR, >> + "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size); >> + return AVERROR(ENOMEM); >> + } >> + memcpy(avctx->extradata, headerPtr.pBuffer, avctx->extradata_size); >> + av_freep(&headerPtr.pBuffer); >> + } >> + return 0; >> + >> +failed_init: > add av_free(q->svt_enc) ? The codec has the FF_CODEC_CAP_INIT_CLEANUP capability, so close() will be called if init() fails. It should be ok as long as both EbDeinitEncoder() and EbDeinitHandle() can be called if either EbInitHandle() or EbInitEncoder() failed. >> + return error_mapping(ret); >> +} >> + >> +static int eb_send_frame(AVCodecContext *avctx, const AVFrame *frame) >> +{ >> + SvtContext *q = avctx->priv_data; >> + SvtEncoder *svt_enc = q->svt_enc; >> + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->in_buf; >> + int ret = 0; >> + >> + if (!frame) { >> + EB_BUFFERHEADERTYPE headerPtrLast; >> + headerPtrLast.nAllocLen = 0; >> + headerPtrLast.nFilledLen = 0; >> + headerPtrLast.nTickCount = 0; >> + headerPtrLast.pAppPrivate = NULL; >> + headerPtrLast.nOffset = 0; >> + headerPtrLast.pBuffer = NULL; >> + headerPtrLast.nFlags = EB_BUFFERFLAG_EOS; >> + EbH265EncSendPicture(svt_enc->svt_handle, &headerPtrLast); >> + q->eos_flag = 1; >> + av_log(avctx, AV_LOG_DEBUG, "Finish sending frames!!!\n"); >> + return ret; >> + } >> + >> + read_in_data(&svt_enc->enc_params, frame, headerPtr); >> + >> + headerPtr->nOffset = 0; >> + headerPtr->nFlags = 0; >> + headerPtr->pAppPrivate = NULL; >> + headerPtr->pts = frame->pts; >> + headerPtr->sliceType = INVALID_SLICE; >> + EbH265EncSendPicture(svt_enc->svt_handle, headerPtr); >> + >> + return ret; >> +} >> + >> +static int eb_receive_packet(AVCodecContext *avctx, AVPacket *pkt) >> +{ >> + SvtContext *q = avctx->priv_data; >> + SvtEncoder *svt_enc = q->svt_enc; >> + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->out_buf; >> + EB_ERRORTYPE stream_status = EB_ErrorNone; >> + int ret = 0; >> + >> + if ((ret = ff_alloc_packet2(avctx, pkt, svt_enc->raw_size, 0)) < 0) { >> + av_log(avctx, AV_LOG_ERROR, "Failed to allocate output packet.\n"); >> + return ret; >> + } >> + headerPtr->pBuffer = pkt->data; >> + stream_status = EbH265GetPacket(svt_enc->svt_handle, headerPtr, q->eos_flag); >> + if (stream_status == EB_NoErrorEmptyQueue) >> + return AVERROR(EAGAIN); >> + >> + pkt->size = headerPtr->nFilledLen; >> + pkt->pts = headerPtr->pts; >> + pkt->dts = headerPtr->dts; >> + if (headerPtr->sliceType == IDR_SLICE) >> + pkt->flags |= AV_PKT_FLAG_KEY; >> + if (headerPtr->sliceType == NON_REF_SLICE) >> + pkt->flags |= AV_PKT_FLAG_DISPOSABLE; >> + >> + ret = (headerPtr->nFlags & EB_BUFFERFLAG_EOS) ? AVERROR_EOF : 0; >> + return ret; >> +} >> + >> +static av_cold int eb_enc_close(AVCodecContext *avctx) >> +{ >> + SvtContext *q = avctx->priv_data; >> + SvtEncoder *svt_enc = q->svt_enc; >> + >> + EbDeinitEncoder(svt_enc->svt_handle); >> + EbDeinitHandle(svt_enc->svt_handle); >> + >> + free_buffer(svt_enc); >> + av_freep(&svt_enc); >> + >> + return 0; >> +} >> + >> +#define OFFSET(x) offsetof(SvtContext, x) >> +#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM >> +static const AVOption options[] = { >> + {"vui", "Enable vui info", OFFSET(svt_param.vui_info), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, >> + {"hielevel", "Hierarchical Prediction Levels [0,3]", OFFSET(svt_param.hierarchical_level), AV_OPT_TYPE_INT, { .i64 = 3 }, 0, 3, VE }, >> + {"la_depth", "Look Ahead Distance [0,256]", OFFSET(svt_param.la_depth), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 256, VE }, >> + {"intra_ref_type", "Intra Refresh Type 0: No intra refresh 1: CRA (Open GOP) 2: IDR", OFFSET(svt_param.intra_ref_type), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 2, VE }, >> + {"enc_p", "Encoding preset [0,12] (for tune 0 and >=4k resolution), [0,10] (for >= 1080p resolution), [0,9] (for all resolution and modes)", OFFSET(svt_param.enc_mode), AV_OPT_TYPE_INT, { .i64 = 9 }, 0, 12, VE }, >> + {"profile", "Profile now support[1,2], Main Still Picture Profile not supported", OFFSET(svt_param.profile), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, 2, VE }, >> + {"rc", "RC mode 0: CQP 1: VBR", OFFSET(svt_param.rc_mode), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, >> + {"q", "QP value for intra frames", OFFSET(svt_param.qp), AV_OPT_TYPE_INT, { .i64 = 32 }, 0, 51, VE }, >> + {"scd", "Scene change detection", OFFSET(svt_param.scd), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, >> + {"tune", "Tune mode: SQ/OQ[0,1]", OFFSET(svt_param.tune), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, >> + {"bl_mode", "Random Access Prediction Structure Type", OFFSET(svt_param.base_layer_switch_mode), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, >> + {NULL}, >> +}; > > Need Document. >> + >> +static const AVClass class = { >> + .class_name = "libsvt_hevc", >> + .item_name = av_default_item_name, >> + .option = options, >> + .version = LIBAVUTIL_VERSION_INT, >> +}; >> + >> +static const AVCodecDefault eb_enc_defaults[] = { >> + { "b", "7M" }, >> + { "refs", "0" }, >> + { "g", "64" }, >> + { "flags", "+cgop" }, >> + { NULL }, >> +}; >> + >> +AVCodec ff_hevc_svt_encoder = { >> + .name = "libsvt_hevc", >> + .long_name = NULL_IF_CONFIG_SMALL("SVT-HEVC(Scalable Video Technology for HEVC) encoder"), >> + .priv_data_size = sizeof(SvtContext), >> + .type = AVMEDIA_TYPE_VIDEO, >> + .id = AV_CODEC_ID_HEVC, >> + .init = eb_enc_init, >> + .send_frame = eb_send_frame, >> + .receive_packet = eb_receive_packet, >> + .close = eb_enc_close, >> + .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS, >> + .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, >> + AV_PIX_FMT_YUV420P10, >> + AV_PIX_FMT_NONE }, >> + .priv_class = &class, >> + .defaults = eb_enc_defaults, >> + .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, >> + .wrapper_name = "libsvt_hevc", >> +}; >> -- >> 1.7.1 >> >> _______________________________________________ >> ffmpeg-devel mailing list >> ffmpeg-devel@ffmpeg.org >> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > Thanks > Steven > > > > > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel >
On 12/2/2018 2:38 AM, Jun Zhao wrote: > base on patch by Huang, Zhengxu from https://github.com/intel/SVT-HEVC > > Signed-off-by: Huang, Zhengxu <zhengxu.huang@intel.com> > Signed-off-by: hassene <hassene.tmar@intel.com> > Signed-off-by: Jun Zhao <jun.zhao@intel.com> > --- > configure | 4 + > libavcodec/Makefile | 1 + > libavcodec/allcodecs.c | 1 + > libavcodec/libsvt_hevc.c | 370 ++++++++++++++++++++++++++++++++++++++++++++++ > 4 files changed, 376 insertions(+), 0 deletions(-) > create mode 100644 libavcodec/libsvt_hevc.c > > diff --git a/configure b/configure > index 54b7e11..c3dca73 100755 > --- a/configure > +++ b/configure > @@ -263,6 +263,7 @@ External library support: > --enable-libspeex enable Speex de/encoding via libspeex [no] > --enable-libsrt enable Haivision SRT protocol via libsrt [no] > --enable-libssh enable SFTP protocol via libssh [no] > + --enable-libsvt enable HEVC encoding via svt [no] > --enable-libtensorflow enable TensorFlow as a DNN module backend > for DNN based filters like sr [no] > --enable-libtesseract enable Tesseract, needed for ocr filter [no] > @@ -1665,6 +1666,7 @@ EXTERNAL_LIBRARY_GPL_LIST=" > libcdio > libdavs2 > librubberband > + libsvt > libvidstab > libx264 > libx265 > @@ -3087,6 +3089,7 @@ pcm_mulaw_at_encoder_select="audio_frame_queue" > chromaprint_muxer_deps="chromaprint" > h264_videotoolbox_encoder_deps="pthreads" > h264_videotoolbox_encoder_select="videotoolbox_encoder" > +hevc_svt_encoder_deps="libsvt" Since this is an external library based encoder, the correct name would be libsvt_encoder. You can also make it libsvt_hevc_encoder instead, which is what you used as AVCodec.name. > hevc_videotoolbox_encoder_deps="pthreads" > hevc_videotoolbox_encoder_select="videotoolbox_encoder" > libaom_av1_decoder_deps="libaom" [...] > +static av_cold int eb_enc_init(AVCodecContext *avctx) > +{ > + SvtContext *q = avctx->priv_data; > + SvtEncoder *svt_enc = NULL; > + EB_ERRORTYPE ret = EB_ErrorNone; > + > + q->svt_enc = av_mallocz(sizeof(*q->svt_enc)); > + if (!q->svt_enc) > + return AVERROR(ENOMEM); > + svt_enc = q->svt_enc; > + > + q->eos_flag = 0; > + > + ret = EbInitHandle(&svt_enc->svt_handle, q, &svt_enc->enc_params); > + if (ret != EB_ErrorNone) > + goto failed_init; > + > + ret = config_enc_params(&svt_enc->enc_params, avctx); > + if (ret != EB_ErrorNone) > + goto failed_init; > + > + ret = EbH265EncSetParameter(svt_enc->svt_handle, &svt_enc->enc_params); > + if (ret != EB_ErrorNone) > + goto failed_init; > + > + ret = EbInitEncoder(svt_enc->svt_handle); > + if (ret != EB_ErrorNone) > + goto failed_init; > + > + if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) { > + > + EB_BUFFERHEADERTYPE headerPtr; > + headerPtr.nSize = sizeof(EB_BUFFERHEADERTYPE); > + headerPtr.nFilledLen = 0; > + headerPtr.pBuffer = av_malloc(10 * 1024 * 1024); > + headerPtr.nAllocLen = (10 * 1024 * 1024); > + > + if (!headerPtr.pBuffer) > + return AVERROR(ENOMEM); > + > + ret = EbH265EncStreamHeader(svt_enc->svt_handle, &headerPtr); > + if (ret != EB_ErrorNone) { > + av_freep(&headerPtr.pBuffer); > + goto failed_init; > + } > + avctx->extradata_size = headerPtr.nFilledLen; > + avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE); > + if (!avctx->extradata) { > + av_log(avctx, AV_LOG_ERROR, > + "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size); > + return AVERROR(ENOMEM); > + } > + memcpy(avctx->extradata, headerPtr.pBuffer, avctx->extradata_size); > + av_freep(&headerPtr.pBuffer); > + } > + return 0; > + > +failed_init: > + return error_mapping(ret); If it's only this line, just use it directly where needed instead of a goto. [...] > +#define OFFSET(x) offsetof(SvtContext, x) > +#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM > +static const AVOption options[] = { > + {"vui", "Enable vui info", OFFSET(svt_param.vui_info), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > + {"hielevel", "Hierarchical Prediction Levels [0,3]", OFFSET(svt_param.hierarchical_level), AV_OPT_TYPE_INT, { .i64 = 3 }, 0, 3, VE }, > + {"la_depth", "Look Ahead Distance [0,256]", OFFSET(svt_param.la_depth), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 256, VE }, > + {"intra_ref_type", "Intra Refresh Type 0: No intra refresh 1: CRA (Open GOP) 2: IDR", OFFSET(svt_param.intra_ref_type), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 2, VE }, > + {"enc_p", "Encoding preset [0,12] (for tune 0 and >=4k resolution), [0,10] (for >= 1080p resolution), [0,9] (for all resolution and modes)", OFFSET(svt_param.enc_mode), AV_OPT_TYPE_INT, { .i64 = 9 }, 0, 12, VE }, Use preset instead, which is more in line with similar encoders. > + {"profile", "Profile now support[1,2], Main Still Picture Profile not supported", OFFSET(svt_param.profile), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, 2, VE }, > + {"rc", "RC mode 0: CQP 1: VBR", OFFSET(svt_param.rc_mode), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > + {"q", "QP value for intra frames", OFFSET(svt_param.qp), AV_OPT_TYPE_INT, { .i64 = 32 }, 0, 51, VE }, qp. > + {"scd", "Scene change detection", OFFSET(svt_param.scd), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > + {"tune", "Tune mode: SQ/OQ[0,1]", OFFSET(svt_param.tune), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > + {"bl_mode", "Random Access Prediction Structure Type", OFFSET(svt_param.base_layer_switch_mode), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > + {NULL}, > +}; > + > +static const AVClass class = { > + .class_name = "libsvt_hevc", > + .item_name = av_default_item_name, > + .option = options, > + .version = LIBAVUTIL_VERSION_INT, > +}; > + > +static const AVCodecDefault eb_enc_defaults[] = { > + { "b", "7M" }, > + { "refs", "0" }, > + { "g", "64" }, > + { "flags", "+cgop" }, > + { NULL }, > +}; > + > +AVCodec ff_hevc_svt_encoder = { > + .name = "libsvt_hevc", > + .long_name = NULL_IF_CONFIG_SMALL("SVT-HEVC(Scalable Video Technology for HEVC) encoder"), > + .priv_data_size = sizeof(SvtContext), > + .type = AVMEDIA_TYPE_VIDEO, > + .id = AV_CODEC_ID_HEVC, > + .init = eb_enc_init, > + .send_frame = eb_send_frame, > + .receive_packet = eb_receive_packet, Keep in mind that, at least for now and unlike with the decoupled input-output decode API, if you only provide a send/receive callback then the encoder will not work with avcodec_encode_video2(), only with avcodec_send_frame() and avcodec_receive_packet(). > + .close = eb_enc_close, > + .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS, > + .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, > + AV_PIX_FMT_YUV420P10, > + AV_PIX_FMT_NONE }, > + .priv_class = &class, > + .defaults = eb_enc_defaults, > + .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, > + .wrapper_name = "libsvt_hevc", > +}; >
On Sun, Dec 02, 2018 at 13:38:48 +0800, Jun Zhao wrote: > + {"intra_ref_type", "Intra Refresh Type 0: No intra refresh 1: CRA (Open GOP) 2: IDR", OFFSET(svt_param.intra_ref_type), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 2, VE }, [...] > + {"profile", "Profile now support[1,2], Main Still Picture Profile not supported", OFFSET(svt_param.profile), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, 2, VE }, > + {"rc", "RC mode 0: CQP 1: VBR", OFFSET(svt_param.rc_mode), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > + {"tune", "Tune mode: SQ/OQ[0,1]", OFFSET(svt_param.tune), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, [...] Four these four, named parameters would be welcome. Moritz
On Sun, Dec 2, 2018 at 10:19 PM James Almer <jamrial@gmail.com> wrote: > > On 12/2/2018 2:38 AM, Jun Zhao wrote: > > base on patch by Huang, Zhengxu from https://github.com/intel/SVT-HEVC > > > > Signed-off-by: Huang, Zhengxu <zhengxu.huang@intel.com> > > Signed-off-by: hassene <hassene.tmar@intel.com> > > Signed-off-by: Jun Zhao <jun.zhao@intel.com> > > --- > > configure | 4 + > > libavcodec/Makefile | 1 + > > libavcodec/allcodecs.c | 1 + > > libavcodec/libsvt_hevc.c | 370 ++++++++++++++++++++++++++++++++++++++++++++++ > > 4 files changed, 376 insertions(+), 0 deletions(-) > > create mode 100644 libavcodec/libsvt_hevc.c > > > > diff --git a/configure b/configure > > index 54b7e11..c3dca73 100755 > > --- a/configure > > +++ b/configure > > @@ -263,6 +263,7 @@ External library support: > > --enable-libspeex enable Speex de/encoding via libspeex [no] > > --enable-libsrt enable Haivision SRT protocol via libsrt [no] > > --enable-libssh enable SFTP protocol via libssh [no] > > + --enable-libsvt enable HEVC encoding via svt [no] > > --enable-libtensorflow enable TensorFlow as a DNN module backend > > for DNN based filters like sr [no] > > --enable-libtesseract enable Tesseract, needed for ocr filter [no] > > @@ -1665,6 +1666,7 @@ EXTERNAL_LIBRARY_GPL_LIST=" > > libcdio > > libdavs2 > > librubberband > > + libsvt > > libvidstab > > libx264 > > libx265 > > @@ -3087,6 +3089,7 @@ pcm_mulaw_at_encoder_select="audio_frame_queue" > > chromaprint_muxer_deps="chromaprint" > > h264_videotoolbox_encoder_deps="pthreads" > > h264_videotoolbox_encoder_select="videotoolbox_encoder" > > +hevc_svt_encoder_deps="libsvt" > > Since this is an external library based encoder, the correct name would > be libsvt_encoder. > You can also make it libsvt_hevc_encoder instead, which is what you used > as AVCodec.name. > I perfer libsvt_hevc_encoder than libsvt_encoder, because maybe we will have libsvt_{av1/vp9/xxx}_encoder in the future. > > hevc_videotoolbox_encoder_deps="pthreads" > > hevc_videotoolbox_encoder_select="videotoolbox_encoder" > > libaom_av1_decoder_deps="libaom" > > > [...] > > > +static av_cold int eb_enc_init(AVCodecContext *avctx) > > +{ > > + SvtContext *q = avctx->priv_data; > > + SvtEncoder *svt_enc = NULL; > > + EB_ERRORTYPE ret = EB_ErrorNone; > > + > > + q->svt_enc = av_mallocz(sizeof(*q->svt_enc)); > > + if (!q->svt_enc) > > + return AVERROR(ENOMEM); > > + svt_enc = q->svt_enc; > > + > > + q->eos_flag = 0; > > + > > + ret = EbInitHandle(&svt_enc->svt_handle, q, &svt_enc->enc_params); > > + if (ret != EB_ErrorNone) > > + goto failed_init; > > + > > + ret = config_enc_params(&svt_enc->enc_params, avctx); > > + if (ret != EB_ErrorNone) > > + goto failed_init; > > + > > + ret = EbH265EncSetParameter(svt_enc->svt_handle, &svt_enc->enc_params); > > + if (ret != EB_ErrorNone) > > + goto failed_init; > > + > > + ret = EbInitEncoder(svt_enc->svt_handle); > > + if (ret != EB_ErrorNone) > > + goto failed_init; > > + > > + if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) { > > + > > + EB_BUFFERHEADERTYPE headerPtr; > > + headerPtr.nSize = sizeof(EB_BUFFERHEADERTYPE); > > + headerPtr.nFilledLen = 0; > > + headerPtr.pBuffer = av_malloc(10 * 1024 * 1024); > > + headerPtr.nAllocLen = (10 * 1024 * 1024); > > + > > + if (!headerPtr.pBuffer) > > + return AVERROR(ENOMEM); > > + > > + ret = EbH265EncStreamHeader(svt_enc->svt_handle, &headerPtr); > > + if (ret != EB_ErrorNone) { > > + av_freep(&headerPtr.pBuffer); > > + goto failed_init; > > + } > > + avctx->extradata_size = headerPtr.nFilledLen; > > + avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE); > > + if (!avctx->extradata) { > > + av_log(avctx, AV_LOG_ERROR, > > + "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size); > > + return AVERROR(ENOMEM); > > + } > > + memcpy(avctx->extradata, headerPtr.pBuffer, avctx->extradata_size); > > + av_freep(&headerPtr.pBuffer); > > + } > > + return 0; > > + > > +failed_init: > > + return error_mapping(ret); > > If it's only this line, just use it directly where needed instead of a goto. Will double-check the fail handle, tks. > > [...] > > > +#define OFFSET(x) offsetof(SvtContext, x) > > +#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM > > +static const AVOption options[] = { > > + {"vui", "Enable vui info", OFFSET(svt_param.vui_info), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > > + {"hielevel", "Hierarchical Prediction Levels [0,3]", OFFSET(svt_param.hierarchical_level), AV_OPT_TYPE_INT, { .i64 = 3 }, 0, 3, VE }, > > + {"la_depth", "Look Ahead Distance [0,256]", OFFSET(svt_param.la_depth), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 256, VE }, > > + {"intra_ref_type", "Intra Refresh Type 0: No intra refresh 1: CRA (Open GOP) 2: IDR", OFFSET(svt_param.intra_ref_type), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 2, VE }, > > + {"enc_p", "Encoding preset [0,12] (for tune 0 and >=4k resolution), [0,10] (for >= 1080p resolution), [0,9] (for all resolution and modes)", OFFSET(svt_param.enc_mode), AV_OPT_TYPE_INT, { .i64 = 9 }, 0, 12, VE }, > > Use preset instead, which is more in line with similar encoders. Ok, will follow > > > + {"profile", "Profile now support[1,2], Main Still Picture Profile not supported", OFFSET(svt_param.profile), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, 2, VE }, > > + {"rc", "RC mode 0: CQP 1: VBR", OFFSET(svt_param.rc_mode), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > > + {"q", "QP value for intra frames", OFFSET(svt_param.qp), AV_OPT_TYPE_INT, { .i64 = 32 }, 0, 51, VE }, > > qp. Will follow the name to avoid surprise > > > + {"scd", "Scene change detection", OFFSET(svt_param.scd), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > > + {"tune", "Tune mode: SQ/OQ[0,1]", OFFSET(svt_param.tune), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > > + {"bl_mode", "Random Access Prediction Structure Type", OFFSET(svt_param.base_layer_switch_mode), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > > + {NULL}, > > +}; > > + > > +static const AVClass class = { > > + .class_name = "libsvt_hevc", > > + .item_name = av_default_item_name, > > + .option = options, > > + .version = LIBAVUTIL_VERSION_INT, > > +}; > > + > > +static const AVCodecDefault eb_enc_defaults[] = { > > + { "b", "7M" }, > > + { "refs", "0" }, > > + { "g", "64" }, > > + { "flags", "+cgop" }, > > + { NULL }, > > +}; > > + > > +AVCodec ff_hevc_svt_encoder = { > > + .name = "libsvt_hevc", > > + .long_name = NULL_IF_CONFIG_SMALL("SVT-HEVC(Scalable Video Technology for HEVC) encoder"), > > + .priv_data_size = sizeof(SvtContext), > > + .type = AVMEDIA_TYPE_VIDEO, > > + .id = AV_CODEC_ID_HEVC, > > + .init = eb_enc_init, > > + .send_frame = eb_send_frame, > > + .receive_packet = eb_receive_packet, > > Keep in mind that, at least for now and unlike with the decoupled > input-output decode API, if you only provide a send/receive callback > then the encoder will not work with avcodec_encode_video2(), only with > avcodec_send_frame() and avcodec_receive_packet(). > Because avcodec_encode_video2() flags with deprecated, so I think is Ok if we don't implement the encode2 callback. BTW: What's the reason we don't use avcodec_send_frame() and avcodec_receive_packet() in avcodec_encode_video2() like use avcodec_send_packet() and avcodec_receive_frame() in avcodec_decode_video2() ? > > + .close = eb_enc_close, > > + .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS, > > + .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, > > + AV_PIX_FMT_YUV420P10, > > + AV_PIX_FMT_NONE }, > > + .priv_class = &class, > > + .defaults = eb_enc_defaults, > > + .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, > > + .wrapper_name = "libsvt_hevc", > > +}; > > >
On Mon, Dec 3, 2018 at 1:58 AM Moritz Barsnick <barsnick@gmx.net> wrote: > On Sun, Dec 02, 2018 at 13:38:48 +0800, Jun Zhao wrote: > > + {"intra_ref_type", "Intra Refresh Type 0: No intra refresh 1: CRA > (Open GOP) 2: IDR", OFFSET(svt_param.intra_ref_type), AV_OPT_TYPE_INT, { > .i64 = 1 }, 1, 2, VE }, > [...] > > + {"profile", "Profile now support[1,2], Main Still Picture Profile > not supported", OFFSET(svt_param.profile), AV_OPT_TYPE_INT, { .i64 = 2 }, > 1, 2, VE }, > > + {"rc", "RC mode 0: CQP 1: VBR", OFFSET(svt_param.rc_mode), > AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > > + {"tune", "Tune mode: SQ/OQ[0,1]", OFFSET(svt_param.tune), > AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > [...] > > Four these four, named parameters would be welcome. > > > Will improve and follow the comments, Tks.
On Sun, Dec 2, 2018 at 10:20 PM James Almer <jamrial@gmail.com> wrote: > On 12/2/2018 7:24 AM, Steven Liu wrote: > > > > > >> On Dec 2, 2018, at 13:38, Jun Zhao <mypopydev@gmail.com> wrote: > >> > >> base on patch by Huang, Zhengxu from https://github.com/intel/SVT-HEVC > >> > >> Signed-off-by: Huang, Zhengxu <zhengxu.huang@intel.com> > >> Signed-off-by: hassene <hassene.tmar@intel.com> > >> Signed-off-by: Jun Zhao <jun.zhao@intel.com> > >> --- > >> configure | 4 + > >> libavcodec/Makefile | 1 + > >> libavcodec/allcodecs.c | 1 + > >> libavcodec/libsvt_hevc.c | 370 > ++++++++++++++++++++++++++++++++++++++++++++++ > >> 4 files changed, 376 insertions(+), 0 deletions(-) > >> create mode 100644 libavcodec/libsvt_hevc.c > >> > >> diff --git a/configure b/configure > >> index 54b7e11..c3dca73 100755 > >> --- a/configure > >> +++ b/configure > >> @@ -263,6 +263,7 @@ External library support: > >> --enable-libspeex enable Speex de/encoding via libspeex [no] > >> --enable-libsrt enable Haivision SRT protocol via libsrt [no] > >> --enable-libssh enable SFTP protocol via libssh [no] > >> + --enable-libsvt enable HEVC encoding via svt [no] > >> --enable-libtensorflow enable TensorFlow as a DNN module backend > >> for DNN based filters like sr [no] > >> --enable-libtesseract enable Tesseract, needed for ocr filter [no] > >> @@ -1665,6 +1666,7 @@ EXTERNAL_LIBRARY_GPL_LIST=" > >> libcdio > >> libdavs2 > >> librubberband > >> + libsvt > >> libvidstab > >> libx264 > >> libx265 > >> @@ -3087,6 +3089,7 @@ pcm_mulaw_at_encoder_select="audio_frame_queue" > >> chromaprint_muxer_deps="chromaprint" > >> h264_videotoolbox_encoder_deps="pthreads" > >> h264_videotoolbox_encoder_select="videotoolbox_encoder" > >> +hevc_svt_encoder_deps="libsvt" > >> hevc_videotoolbox_encoder_deps="pthreads" > >> hevc_videotoolbox_encoder_select="videotoolbox_encoder" > >> libaom_av1_decoder_deps="libaom" > >> @@ -6147,6 +6150,7 @@ enabled libsoxr && require libsoxr > soxr.h soxr_create -lsoxr > >> enabled libssh && require_pkg_config libssh libssh > libssh/sftp.h sftp_init > >> enabled libspeex && require_pkg_config libspeex speex > speex/speex.h speex_decoder_init > >> enabled libsrt && require_pkg_config libsrt "srt >= 1.3.0" > srt/srt.h srt_socket > >> +enabled libsvt && require_pkg_config libsvt svt EbApi.h > EbInitHandle > >> enabled libtensorflow && require libtensorflow tensorflow/c/c_api.h > TF_Version -ltensorflow > >> enabled libtesseract && require_pkg_config libtesseract tesseract > tesseract/capi.h TessBaseAPICreate > >> enabled libtheora && require libtheora theora/theoraenc.h > th_info_init -ltheoraenc -ltheoradec -logg > >> diff --git a/libavcodec/Makefile b/libavcodec/Makefile > >> index 5feadac..1a8b547 100644 > >> --- a/libavcodec/Makefile > >> +++ b/libavcodec/Makefile > >> @@ -374,6 +374,7 @@ OBJS-$(CONFIG_HEVC_QSV_DECODER) += > qsvdec_h2645.o > >> OBJS-$(CONFIG_HEVC_QSV_ENCODER) += qsvenc_hevc.o hevc_ps_enc.o > \ > >> hevc_data.o > >> OBJS-$(CONFIG_HEVC_RKMPP_DECODER) += rkmppdec.o > >> +OBJS-$(CONFIG_HEVC_SVT_ENCODER) += libsvt_hevc.o > >> OBJS-$(CONFIG_HEVC_VAAPI_ENCODER) += vaapi_encode_h265.o > h265_profile_level.o > >> OBJS-$(CONFIG_HEVC_V4L2M2M_DECODER) += v4l2_m2m_dec.o > >> OBJS-$(CONFIG_HEVC_V4L2M2M_ENCODER) += v4l2_m2m_enc.o > >> diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c > >> index d70646e..ff8ea94 100644 > >> --- a/libavcodec/allcodecs.c > >> +++ b/libavcodec/allcodecs.c > >> @@ -749,6 +749,7 @@ extern AVCodec ff_hevc_cuvid_decoder; > >> extern AVCodec ff_hevc_mediacodec_decoder; > >> extern AVCodec ff_hevc_nvenc_encoder; > >> extern AVCodec ff_hevc_qsv_encoder; > >> +extern AVCodec ff_hevc_svt_encoder; > >> extern AVCodec ff_hevc_v4l2m2m_encoder; > >> extern AVCodec ff_hevc_vaapi_encoder; > >> extern AVCodec ff_hevc_videotoolbox_encoder; > >> diff --git a/libavcodec/libsvt_hevc.c b/libavcodec/libsvt_hevc.c > >> new file mode 100644 > >> index 0000000..1e10111 > >> --- /dev/null > >> +++ b/libavcodec/libsvt_hevc.c > >> @@ -0,0 +1,370 @@ > >> +/* > >> +* Scalable Video Technology for HEVC encoder library plugin > >> +* > >> +* Copyright (c) 2018 Intel Corporation > >> +* > >> +* This program is free software; you can redistribute it and/or > >> +* modify it under the terms of the GNU Lesser General Public > >> +* License as published by the Free Software Foundation; either > >> +* version 2.1 of the License, or (at your option) any later version. > >> +* > >> +* This program is distributed in the hope that it will be useful, > >> +* but WITHOUT ANY WARRANTY; without even the implied warranty of > >> +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > >> +* Lesser General Public License for more details. > >> +* > >> +* You should have received a copy of the GNU Lesser General Public > >> +* License along with this program; if not, write to the Free Software > >> +* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA > 02110-1301 USA > >> +*/ > >> + > >> +#include "EbErrorCodes.h" > >> +#include "EbTime.h" > >> +#include "EbApi.h" > >> + > >> +#include "libavutil/common.h" > >> +#include "libavutil/frame.h" > >> +#include "libavutil/opt.h" > >> + > >> +#include "internal.h" > >> +#include "avcodec.h" > >> + > >> +typedef struct SvtEncoder { > >> + EB_H265_ENC_CONFIGURATION enc_params; > >> + EB_COMPONENTTYPE *svt_handle; > >> + EB_BUFFERHEADERTYPE *in_buf; > >> + EB_BUFFERHEADERTYPE *out_buf; > >> + int raw_size; > >> +} SvtEncoder; > >> + > >> +typedef struct SvtParams { > >> + int vui_info; > >> + int hierarchical_level; > >> + int la_depth; > >> + int intra_ref_type; > >> + int enc_mode; > >> + int rc_mode; > >> + int scd; > >> + int tune; > >> + int qp; > >> + int profile; > >> + int base_layer_switch_mode; > >> +}SvtParams; > >> + > >> +typedef struct SvtContext { > >> + AVClass *class; > >> + SvtEncoder *svt_enc; > >> + SvtParams svt_param; > >> + int eos_flag; > >> +} SvtContext; > >> + > >> +static void free_buffer(SvtEncoder *svt_enc) > >> +{ > >> + if (svt_enc->in_buf) { > >> + EB_H265_ENC_INPUT *in_data = (EB_H265_ENC_INPUT > *)svt_enc->in_buf->pBuffer; > >> + av_freep(&in_data); > >> + av_freep(&svt_enc->in_buf); > >> + } > >> + av_freep(&svt_enc->out_buf); > >> +} > >> + > >> +static EB_ERRORTYPE alloc_buffer(EB_H265_ENC_CONFIGURATION *config, > SvtEncoder *svt_enc) > >> +{ > >> + EB_ERRORTYPE ret = EB_ErrorNone; > >> + > >> + const int pack_mode_10bit = (config->encoderBitDepth > 8) && > (config->compressedTenBitFormat == 0) ? 1 : 0; > >> + const size_t luma_size_8bit = config->sourceWidth * > config->sourceHeight * (1 << pack_mode_10bit); > >> + const size_t luma_size_10bit = (config->encoderBitDepth > 8 && > pack_mode_10bit == 0) ? luma_size_8bit : 0; > >> + > >> + svt_enc->raw_size = (luma_size_8bit + luma_size_10bit) * 3 / 2; > >> + > >> + // allocate buffer for in and out > >> + svt_enc->in_buf = > av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); > >> + svt_enc->out_buf = > av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); > >> + if (!svt_enc->in_buf || !svt_enc->out_buf) > >> + goto failed; > >> + > >> + svt_enc->in_buf->pBuffer = av_mallocz(sizeof(EB_H265_ENC_INPUT)); > >> + if (!svt_enc->in_buf->pBuffer) > >> + goto failed; > >> + > >> + svt_enc->in_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); > >> + svt_enc->in_buf->pAppPrivate = NULL; > >> + svt_enc->out_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); > >> + svt_enc->out_buf->nAllocLen = svt_enc->raw_size; > >> + svt_enc->out_buf->pAppPrivate = NULL; > >> + > >> + return ret; > >> + > >> +failed: > >> + free_buffer(svt_enc); > >> + return AVERROR(ENOMEM); > >> +} > >> + > >> +static int error_mapping(int val) > >> +{ > >> + if (val == EB_ErrorInsufficientResources) > >> + return AVERROR(ENOMEM); > >> + if ((val == EB_ErrorUndefined) || (val == > EB_ErrorInvalidComponent) || > >> + (val == EB_ErrorBadParameter)) > >> + return AVERROR(EINVAL); > >> + return AVERROR_EXTERNAL; > >> +} > >> + > >> +static EB_ERRORTYPE config_enc_params(EB_H265_ENC_CONFIGURATION > *param, AVCodecContext *avctx) > >> +{ > >> + SvtContext *q = avctx->priv_data; > >> + SvtEncoder *svt_enc = q->svt_enc; > >> + EB_ERRORTYPE ret = EB_ErrorNone; > >> + int tenBits = 0; > > Camelcase ? > >> + > >> + param->sourceWidth = avctx->width; > >> + param->sourceHeight = avctx->height; > >> + > >> + if (avctx->pix_fmt == AV_PIX_FMT_YUV420P10LE) { > >> + av_log(avctx, AV_LOG_DEBUG , "Encoder 10 bits depth input\n"); > >> + param->compressedTenBitFormat = 0; > >> + tenBits = 1; > >> + } > >> + > > maybe need check q->svt_param first. > >> > >> + // Update param from options > >> + param->hierarchicalLevels = q->svt_param.hierarchical_level; > >> + param->encMode = q->svt_param.enc_mode; > >> + param->intraRefreshType = q->svt_param.intra_ref_type; > >> + param->profile = q->svt_param.profile; > >> + param->rateControlMode = q->svt_param.rc_mode; > >> + param->sceneChangeDetection = q->svt_param.scd; > >> + param->tune = q->svt_param.tune; > >> + param->baseLayerSwitchMode = > q->svt_param.base_layer_switch_mode; > >> + param->qp = q->svt_param.qp; > >> + > >> + param->targetBitRate = avctx->bit_rate; > >> + param->intraPeriodLength = avctx->gop_size-1; > >> + param->frameRateNumerator = avctx->time_base.den; > >> + param->frameRateDenominator = avctx->time_base.num * > avctx->ticks_per_frame; > >> + > >> + param->codeVpsSpsPps = 0; > >> + > >> + if (q->svt_param.vui_info) > >> + param->videoUsabilityInfo = q->svt_param.vui_info; > >> + if (q->svt_param.la_depth != -1) > >> + param->lookAheadDistance = q->svt_param.la_depth; > >> + > >> + if (tenBits == 1) { > >> + param->encoderBitDepth = 10; > >> + param->profile = 2; > >> + } > >> + > >> + ret = alloc_buffer(param, svt_enc); > >> + > >> + return ret; > >> +} > >> + > >> +static void read_in_data(EB_H265_ENC_CONFIGURATION *config, const > AVFrame* frame, EB_BUFFERHEADERTYPE *headerPtr) > >> +{ > >> + unsigned int is16bit = config->encoderBitDepth > 8; > >> + unsigned long long lumaReadSize = (unsigned long > long)config->sourceWidth * config->sourceHeight<< is16bit; > > what about use uint64_t? > >> + EB_H265_ENC_INPUT *in_data = > (EB_H265_ENC_INPUT*)headerPtr->pBuffer; > >> + > >> + // support yuv420p and yuv420p010 > >> + in_data->luma = frame->data[0]; > >> + in_data->cb = frame->data[1]; > >> + in_data->cr = frame->data[2]; > >> + > >> + // stride info > >> + in_data->yStride = frame->linesize[0] >> is16bit; > >> + in_data->cbStride = frame->linesize[1] >> is16bit; > >> + in_data->crStride = frame->linesize[2] >> is16bit; > >> + > >> + headerPtr->nFilledLen += lumaReadSize * 3/2u; > >> +} > >> + > >> +static av_cold int eb_enc_init(AVCodecContext *avctx) > >> +{ > >> + SvtContext *q = avctx->priv_data; > >> + SvtEncoder *svt_enc = NULL; > >> + EB_ERRORTYPE ret = EB_ErrorNone; > >> + > >> + q->svt_enc = av_mallocz(sizeof(*q->svt_enc)); > >> + if (!q->svt_enc) > >> + return AVERROR(ENOMEM); > >> + svt_enc = q->svt_enc; > >> + > >> + q->eos_flag = 0; > >> + > >> + ret = EbInitHandle(&svt_enc->svt_handle, q, &svt_enc->enc_params); > >> + if (ret != EB_ErrorNone) > >> + goto failed_init; > >> + > >> + ret = config_enc_params(&svt_enc->enc_params, avctx); > >> + if (ret != EB_ErrorNone) > >> + goto failed_init; > >> + > >> + ret = EbH265EncSetParameter(svt_enc->svt_handle, > &svt_enc->enc_params); > >> + if (ret != EB_ErrorNone) > >> + goto failed_init; > >> + > >> + ret = EbInitEncoder(svt_enc->svt_handle); > >> + if (ret != EB_ErrorNone) > >> + goto failed_init; > >> + > >> + if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) { > >> + > >> + EB_BUFFERHEADERTYPE headerPtr; > >> + headerPtr.nSize = sizeof(EB_BUFFERHEADERTYPE); > >> + headerPtr.nFilledLen = 0; > >> + headerPtr.pBuffer = av_malloc(10 * 1024 * 1024); > >> + headerPtr.nAllocLen = (10 * 1024 * 1024); > >> + > >> + if (!headerPtr.pBuffer) > >> + return AVERROR(ENOMEM); > >> + > >> + ret = EbH265EncStreamHeader(svt_enc->svt_handle, &headerPtr); > >> + if (ret != EB_ErrorNone) { > >> + av_freep(&headerPtr.pBuffer); > >> + goto failed_init; > >> + } > >> + avctx->extradata_size = headerPtr.nFilledLen; > >> + avctx->extradata = av_malloc(avctx->extradata_size + > AV_INPUT_BUFFER_PADDING_SIZE); > >> + if (!avctx->extradata) { > >> + av_log(avctx, AV_LOG_ERROR, > >> + "Cannot allocate HEVC header of size %d.\n", > avctx->extradata_size); > >> + return AVERROR(ENOMEM); > >> + } > >> + memcpy(avctx->extradata, headerPtr.pBuffer, > avctx->extradata_size); > >> + av_freep(&headerPtr.pBuffer); > >> + } > >> + return 0; > >> + > >> +failed_init: > > add av_free(q->svt_enc) ? > > The codec has the FF_CODEC_CAP_INIT_CLEANUP capability, so close() will > be called if init() fails. > > It should be ok as long as both EbDeinitEncoder() and EbDeinitHandle() > can be called if either EbInitHandle() or EbInitEncoder() failed. > > Yes, Steven Liu given the same commnets for this part. > >> + return error_mapping(ret); > >> +} > >> + > >> +static int eb_send_frame(AVCodecContext *avctx, const AVFrame *frame) > >> +{ > >> + SvtContext *q = avctx->priv_data; > >> + SvtEncoder *svt_enc = q->svt_enc; > >> + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->in_buf; > >> + int ret = 0; > >> + > >> + if (!frame) { > >> + EB_BUFFERHEADERTYPE headerPtrLast; > >> + headerPtrLast.nAllocLen = 0; > >> + headerPtrLast.nFilledLen = 0; > >> + headerPtrLast.nTickCount = 0; > >> + headerPtrLast.pAppPrivate = NULL; > >> + headerPtrLast.nOffset = 0; > >> + headerPtrLast.pBuffer = NULL; > >> + headerPtrLast.nFlags = EB_BUFFERFLAG_EOS; > >> + EbH265EncSendPicture(svt_enc->svt_handle, &headerPtrLast); > >> + q->eos_flag = 1; > >> + av_log(avctx, AV_LOG_DEBUG, "Finish sending frames!!!\n"); > >> + return ret; > >> + } > >> + > >> + read_in_data(&svt_enc->enc_params, frame, headerPtr); > >> + > >> + headerPtr->nOffset = 0; > >> + headerPtr->nFlags = 0; > >> + headerPtr->pAppPrivate = NULL; > >> + headerPtr->pts = frame->pts; > >> + headerPtr->sliceType = INVALID_SLICE; > >> + EbH265EncSendPicture(svt_enc->svt_handle, headerPtr); > >> + > >> + return ret; > >> +} > >> + > >> +static int eb_receive_packet(AVCodecContext *avctx, AVPacket *pkt) > >> +{ > >> + SvtContext *q = avctx->priv_data; > >> + SvtEncoder *svt_enc = q->svt_enc; > >> + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->out_buf; > >> + EB_ERRORTYPE stream_status = EB_ErrorNone; > >> + int ret = 0; > >> + > >> + if ((ret = ff_alloc_packet2(avctx, pkt, svt_enc->raw_size, 0)) < > 0) { > >> + av_log(avctx, AV_LOG_ERROR, "Failed to allocate output > packet.\n"); > >> + return ret; > >> + } > >> + headerPtr->pBuffer = pkt->data; > >> + stream_status = EbH265GetPacket(svt_enc->svt_handle, headerPtr, > q->eos_flag); > >> + if (stream_status == EB_NoErrorEmptyQueue) > >> + return AVERROR(EAGAIN); > >> + > >> + pkt->size = headerPtr->nFilledLen; > >> + pkt->pts = headerPtr->pts; > >> + pkt->dts = headerPtr->dts; > >> + if (headerPtr->sliceType == IDR_SLICE) > >> + pkt->flags |= AV_PKT_FLAG_KEY; > >> + if (headerPtr->sliceType == NON_REF_SLICE) > >> + pkt->flags |= AV_PKT_FLAG_DISPOSABLE; > >> + > >> + ret = (headerPtr->nFlags & EB_BUFFERFLAG_EOS) ? AVERROR_EOF : 0; > >> + return ret; > >> +} > >> + > >> +static av_cold int eb_enc_close(AVCodecContext *avctx) > >> +{ > >> + SvtContext *q = avctx->priv_data; > >> + SvtEncoder *svt_enc = q->svt_enc; > >> + > >> + EbDeinitEncoder(svt_enc->svt_handle); > >> + EbDeinitHandle(svt_enc->svt_handle); > >> + > >> + free_buffer(svt_enc); > >> + av_freep(&svt_enc); > >> + > >> + return 0; > >> +} > >> + > >> +#define OFFSET(x) offsetof(SvtContext, x) > >> +#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM > >> +static const AVOption options[] = { > >> + {"vui", "Enable vui info", OFFSET(svt_param.vui_info), > AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > >> + {"hielevel", "Hierarchical Prediction Levels [0,3]", > OFFSET(svt_param.hierarchical_level), AV_OPT_TYPE_INT, { .i64 = 3 }, 0, 3, > VE }, > >> + {"la_depth", "Look Ahead Distance [0,256]", > OFFSET(svt_param.la_depth), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 256, VE }, > >> + {"intra_ref_type", "Intra Refresh Type 0: No intra refresh 1: CRA > (Open GOP) 2: IDR", OFFSET(svt_param.intra_ref_type), AV_OPT_TYPE_INT, { > .i64 = 1 }, 1, 2, VE }, > >> + {"enc_p", "Encoding preset [0,12] (for tune 0 and >=4k > resolution), [0,10] (for >= 1080p resolution), [0,9] (for all resolution > and modes)", OFFSET(svt_param.enc_mode), AV_OPT_TYPE_INT, { .i64 = 9 }, 0, > 12, VE }, > >> + {"profile", "Profile now support[1,2], Main Still Picture Profile > not supported", OFFSET(svt_param.profile), AV_OPT_TYPE_INT, { .i64 = 2 }, > 1, 2, VE }, > >> + {"rc", "RC mode 0: CQP 1: VBR", OFFSET(svt_param.rc_mode), > AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > >> + {"q", "QP value for intra frames", OFFSET(svt_param.qp), > AV_OPT_TYPE_INT, { .i64 = 32 }, 0, 51, VE }, > >> + {"scd", "Scene change detection", OFFSET(svt_param.scd), > AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > >> + {"tune", "Tune mode: SQ/OQ[0,1]", OFFSET(svt_param.tune), > AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > >> + {"bl_mode", "Random Access Prediction Structure Type", > OFFSET(svt_param.base_layer_switch_mode), AV_OPT_TYPE_BOOL, { .i64 = 0 }, > 0, 1, VE }, > >> + {NULL}, > >> +}; > > > > Need Document. > >> + > >> +static const AVClass class = { > >> + .class_name = "libsvt_hevc", > >> + .item_name = av_default_item_name, > >> + .option = options, > >> + .version = LIBAVUTIL_VERSION_INT, > >> +}; > >> + > >> +static const AVCodecDefault eb_enc_defaults[] = { > >> + { "b", "7M" }, > >> + { "refs", "0" }, > >> + { "g", "64" }, > >> + { "flags", "+cgop" }, > >> + { NULL }, > >> +}; > >> + > >> +AVCodec ff_hevc_svt_encoder = { > >> + .name = "libsvt_hevc", > >> + .long_name = NULL_IF_CONFIG_SMALL("SVT-HEVC(Scalable Video > Technology for HEVC) encoder"), > >> + .priv_data_size = sizeof(SvtContext), > >> + .type = AVMEDIA_TYPE_VIDEO, > >> + .id = AV_CODEC_ID_HEVC, > >> + .init = eb_enc_init, > >> + .send_frame = eb_send_frame, > >> + .receive_packet = eb_receive_packet, > >> + .close = eb_enc_close, > >> + .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS, > >> + .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, > >> + > AV_PIX_FMT_YUV420P10, > >> + AV_PIX_FMT_NONE }, > >> + .priv_class = &class, > >> + .defaults = eb_enc_defaults, > >> + .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, > >> + .wrapper_name = "libsvt_hevc", > >> +}; > >> -- > >
On Sun, Dec 2, 2018 at 6:25 PM Steven Liu <lq@chinaffmpeg.org> wrote: > > > > > On Dec 2, 2018, at 13:38, Jun Zhao <mypopydev@gmail.com> wrote: > > > > base on patch by Huang, Zhengxu from https://github.com/intel/SVT-HEVC > > > > Signed-off-by: Huang, Zhengxu <zhengxu.huang@intel.com> > > Signed-off-by: hassene <hassene.tmar@intel.com> > > Signed-off-by: Jun Zhao <jun.zhao@intel.com> > > --- > > configure | 4 + > > libavcodec/Makefile | 1 + > > libavcodec/allcodecs.c | 1 + > > libavcodec/libsvt_hevc.c | 370 ++++++++++++++++++++++++++++++++++++++++++++++ > > 4 files changed, 376 insertions(+), 0 deletions(-) > > create mode 100644 libavcodec/libsvt_hevc.c > > > > diff --git a/configure b/configure > > index 54b7e11..c3dca73 100755 > > --- a/configure > > +++ b/configure > > @@ -263,6 +263,7 @@ External library support: > > --enable-libspeex enable Speex de/encoding via libspeex [no] > > --enable-libsrt enable Haivision SRT protocol via libsrt [no] > > --enable-libssh enable SFTP protocol via libssh [no] > > + --enable-libsvt enable HEVC encoding via svt [no] > > --enable-libtensorflow enable TensorFlow as a DNN module backend > > for DNN based filters like sr [no] > > --enable-libtesseract enable Tesseract, needed for ocr filter [no] > > @@ -1665,6 +1666,7 @@ EXTERNAL_LIBRARY_GPL_LIST=" > > libcdio > > libdavs2 > > librubberband > > + libsvt > > libvidstab > > libx264 > > libx265 > > @@ -3087,6 +3089,7 @@ pcm_mulaw_at_encoder_select="audio_frame_queue" > > chromaprint_muxer_deps="chromaprint" > > h264_videotoolbox_encoder_deps="pthreads" > > h264_videotoolbox_encoder_select="videotoolbox_encoder" > > +hevc_svt_encoder_deps="libsvt" > > hevc_videotoolbox_encoder_deps="pthreads" > > hevc_videotoolbox_encoder_select="videotoolbox_encoder" > > libaom_av1_decoder_deps="libaom" > > @@ -6147,6 +6150,7 @@ enabled libsoxr && require libsoxr soxr.h soxr_create -lsoxr > > enabled libssh && require_pkg_config libssh libssh libssh/sftp.h sftp_init > > enabled libspeex && require_pkg_config libspeex speex speex/speex.h speex_decoder_init > > enabled libsrt && require_pkg_config libsrt "srt >= 1.3.0" srt/srt.h srt_socket > > +enabled libsvt && require_pkg_config libsvt svt EbApi.h EbInitHandle > > enabled libtensorflow && require libtensorflow tensorflow/c/c_api.h TF_Version -ltensorflow > > enabled libtesseract && require_pkg_config libtesseract tesseract tesseract/capi.h TessBaseAPICreate > > enabled libtheora && require libtheora theora/theoraenc.h th_info_init -ltheoraenc -ltheoradec -logg > > diff --git a/libavcodec/Makefile b/libavcodec/Makefile > > index 5feadac..1a8b547 100644 > > --- a/libavcodec/Makefile > > +++ b/libavcodec/Makefile > > @@ -374,6 +374,7 @@ OBJS-$(CONFIG_HEVC_QSV_DECODER) += qsvdec_h2645.o > > OBJS-$(CONFIG_HEVC_QSV_ENCODER) += qsvenc_hevc.o hevc_ps_enc.o \ > > hevc_data.o > > OBJS-$(CONFIG_HEVC_RKMPP_DECODER) += rkmppdec.o > > +OBJS-$(CONFIG_HEVC_SVT_ENCODER) += libsvt_hevc.o > > OBJS-$(CONFIG_HEVC_VAAPI_ENCODER) += vaapi_encode_h265.o h265_profile_level.o > > OBJS-$(CONFIG_HEVC_V4L2M2M_DECODER) += v4l2_m2m_dec.o > > OBJS-$(CONFIG_HEVC_V4L2M2M_ENCODER) += v4l2_m2m_enc.o > > diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c > > index d70646e..ff8ea94 100644 > > --- a/libavcodec/allcodecs.c > > +++ b/libavcodec/allcodecs.c > > @@ -749,6 +749,7 @@ extern AVCodec ff_hevc_cuvid_decoder; > > extern AVCodec ff_hevc_mediacodec_decoder; > > extern AVCodec ff_hevc_nvenc_encoder; > > extern AVCodec ff_hevc_qsv_encoder; > > +extern AVCodec ff_hevc_svt_encoder; > > extern AVCodec ff_hevc_v4l2m2m_encoder; > > extern AVCodec ff_hevc_vaapi_encoder; > > extern AVCodec ff_hevc_videotoolbox_encoder; > > diff --git a/libavcodec/libsvt_hevc.c b/libavcodec/libsvt_hevc.c > > new file mode 100644 > > index 0000000..1e10111 > > --- /dev/null > > +++ b/libavcodec/libsvt_hevc.c > > @@ -0,0 +1,370 @@ > > +/* > > +* Scalable Video Technology for HEVC encoder library plugin > > +* > > +* Copyright (c) 2018 Intel Corporation > > +* > > +* This program is free software; you can redistribute it and/or > > +* modify it under the terms of the GNU Lesser General Public > > +* License as published by the Free Software Foundation; either > > +* version 2.1 of the License, or (at your option) any later version. > > +* > > +* This program is distributed in the hope that it will be useful, > > +* but WITHOUT ANY WARRANTY; without even the implied warranty of > > +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > > +* Lesser General Public License for more details. > > +* > > +* You should have received a copy of the GNU Lesser General Public > > +* License along with this program; if not, write to the Free Software > > +* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA > > +*/ > > + > > +#include "EbErrorCodes.h" > > +#include "EbTime.h" > > +#include "EbApi.h" > > + > > +#include "libavutil/common.h" > > +#include "libavutil/frame.h" > > +#include "libavutil/opt.h" > > + > > +#include "internal.h" > > +#include "avcodec.h" > > + > > +typedef struct SvtEncoder { > > + EB_H265_ENC_CONFIGURATION enc_params; > > + EB_COMPONENTTYPE *svt_handle; > > + EB_BUFFERHEADERTYPE *in_buf; > > + EB_BUFFERHEADERTYPE *out_buf; > > + int raw_size; > > +} SvtEncoder; > > + > > +typedef struct SvtParams { > > + int vui_info; > > + int hierarchical_level; > > + int la_depth; > > + int intra_ref_type; > > + int enc_mode; > > + int rc_mode; > > + int scd; > > + int tune; > > + int qp; > > + int profile; > > + int base_layer_switch_mode; > > +}SvtParams; > > + > > +typedef struct SvtContext { > > + AVClass *class; > > + SvtEncoder *svt_enc; > > + SvtParams svt_param; > > + int eos_flag; > > +} SvtContext; > > + > > +static void free_buffer(SvtEncoder *svt_enc) > > +{ > > + if (svt_enc->in_buf) { > > + EB_H265_ENC_INPUT *in_data = (EB_H265_ENC_INPUT *)svt_enc->in_buf->pBuffer; > > + av_freep(&in_data); > > + av_freep(&svt_enc->in_buf); > > + } > > + av_freep(&svt_enc->out_buf); > > +} > > + > > +static EB_ERRORTYPE alloc_buffer(EB_H265_ENC_CONFIGURATION *config, SvtEncoder *svt_enc) > > +{ > > + EB_ERRORTYPE ret = EB_ErrorNone; > > + > > + const int pack_mode_10bit = (config->encoderBitDepth > 8) && (config->compressedTenBitFormat == 0) ? 1 : 0; > > + const size_t luma_size_8bit = config->sourceWidth * config->sourceHeight * (1 << pack_mode_10bit); > > + const size_t luma_size_10bit = (config->encoderBitDepth > 8 && pack_mode_10bit == 0) ? luma_size_8bit : 0; > > + > > + svt_enc->raw_size = (luma_size_8bit + luma_size_10bit) * 3 / 2; > > + > > + // allocate buffer for in and out > > + svt_enc->in_buf = av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); > > + svt_enc->out_buf = av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); > > + if (!svt_enc->in_buf || !svt_enc->out_buf) > > + goto failed; > > + > > + svt_enc->in_buf->pBuffer = av_mallocz(sizeof(EB_H265_ENC_INPUT)); > > + if (!svt_enc->in_buf->pBuffer) > > + goto failed; > > + > > + svt_enc->in_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); > > + svt_enc->in_buf->pAppPrivate = NULL; > > + svt_enc->out_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); > > + svt_enc->out_buf->nAllocLen = svt_enc->raw_size; > > + svt_enc->out_buf->pAppPrivate = NULL; > > + > > + return ret; > > + > > +failed: > > + free_buffer(svt_enc); > > + return AVERROR(ENOMEM); > > +} > > + > > +static int error_mapping(int val) > > +{ > > + if (val == EB_ErrorInsufficientResources) > > + return AVERROR(ENOMEM); > > + if ((val == EB_ErrorUndefined) || (val == EB_ErrorInvalidComponent) || > > + (val == EB_ErrorBadParameter)) > > + return AVERROR(EINVAL); > > + return AVERROR_EXTERNAL; > > +} > > + > > +static EB_ERRORTYPE config_enc_params(EB_H265_ENC_CONFIGURATION *param, AVCodecContext *avctx) > > +{ > > + SvtContext *q = avctx->priv_data; > > + SvtEncoder *svt_enc = q->svt_enc; > > + EB_ERRORTYPE ret = EB_ErrorNone; > > + int tenBits = 0; > Camelcase ? Will change and follow FFmpeg's coding style > > > + > > + param->sourceWidth = avctx->width; > > + param->sourceHeight = avctx->height; > > + > > + if (avctx->pix_fmt == AV_PIX_FMT_YUV420P10LE) { > > + av_log(avctx, AV_LOG_DEBUG , "Encoder 10 bits depth input\n"); > > + param->compressedTenBitFormat = 0; > > + tenBits = 1; > > + } > > + > maybe need check q->svt_param first. Yes, good catch. > > > > > + // Update param from options > > + param->hierarchicalLevels = q->svt_param.hierarchical_level; > > + param->encMode = q->svt_param.enc_mode; > > + param->intraRefreshType = q->svt_param.intra_ref_type; > > + param->profile = q->svt_param.profile; > > + param->rateControlMode = q->svt_param.rc_mode; > > + param->sceneChangeDetection = q->svt_param.scd; > > + param->tune = q->svt_param.tune; > > + param->baseLayerSwitchMode = q->svt_param.base_layer_switch_mode; > > + param->qp = q->svt_param.qp; > > + > > + param->targetBitRate = avctx->bit_rate; > > + param->intraPeriodLength = avctx->gop_size-1; > > + param->frameRateNumerator = avctx->time_base.den; > > + param->frameRateDenominator = avctx->time_base.num * avctx->ticks_per_frame; > > + > > + param->codeVpsSpsPps = 0; > > + > > + if (q->svt_param.vui_info) > > + param->videoUsabilityInfo = q->svt_param.vui_info; > > + if (q->svt_param.la_depth != -1) > > + param->lookAheadDistance = q->svt_param.la_depth; > > + > > + if (tenBits == 1) { > > + param->encoderBitDepth = 10; > > + param->profile = 2; > > + } > > + > > + ret = alloc_buffer(param, svt_enc); > > + > > + return ret; > > +} > > + > > +static void read_in_data(EB_H265_ENC_CONFIGURATION *config, const AVFrame* frame, EB_BUFFERHEADERTYPE *headerPtr) > > +{ > > + unsigned int is16bit = config->encoderBitDepth > 8; > > + unsigned long long lumaReadSize = (unsigned long long)config->sourceWidth * config->sourceHeight<< is16bit; > what about use uint64_t? > > + EB_H265_ENC_INPUT *in_data = (EB_H265_ENC_INPUT*)headerPtr->pBuffer; > > + > > + // support yuv420p and yuv420p010 > > + in_data->luma = frame->data[0]; > > + in_data->cb = frame->data[1]; > > + in_data->cr = frame->data[2]; > > + > > + // stride info > > + in_data->yStride = frame->linesize[0] >> is16bit; > > + in_data->cbStride = frame->linesize[1] >> is16bit; > > + in_data->crStride = frame->linesize[2] >> is16bit; > > + > > + headerPtr->nFilledLen += lumaReadSize * 3/2u; > > +} > > + > > +static av_cold int eb_enc_init(AVCodecContext *avctx) > > +{ > > + SvtContext *q = avctx->priv_data; > > + SvtEncoder *svt_enc = NULL; > > + EB_ERRORTYPE ret = EB_ErrorNone; > > + > > + q->svt_enc = av_mallocz(sizeof(*q->svt_enc)); > > + if (!q->svt_enc) > > + return AVERROR(ENOMEM); > > + svt_enc = q->svt_enc; > > + > > + q->eos_flag = 0; > > + > > + ret = EbInitHandle(&svt_enc->svt_handle, q, &svt_enc->enc_params); > > + if (ret != EB_ErrorNone) > > + goto failed_init; > > + > > + ret = config_enc_params(&svt_enc->enc_params, avctx); > > + if (ret != EB_ErrorNone) > > + goto failed_init; > > + > > + ret = EbH265EncSetParameter(svt_enc->svt_handle, &svt_enc->enc_params); > > + if (ret != EB_ErrorNone) > > + goto failed_init; > > + > > + ret = EbInitEncoder(svt_enc->svt_handle); > > + if (ret != EB_ErrorNone) > > + goto failed_init; > > + > > + if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) { > > + > > + EB_BUFFERHEADERTYPE headerPtr; > > + headerPtr.nSize = sizeof(EB_BUFFERHEADERTYPE); > > + headerPtr.nFilledLen = 0; > > + headerPtr.pBuffer = av_malloc(10 * 1024 * 1024); > > + headerPtr.nAllocLen = (10 * 1024 * 1024); > > + > > + if (!headerPtr.pBuffer) > > + return AVERROR(ENOMEM); > > + > > + ret = EbH265EncStreamHeader(svt_enc->svt_handle, &headerPtr); > > + if (ret != EB_ErrorNone) { > > + av_freep(&headerPtr.pBuffer); > > + goto failed_init; > > + } > > + avctx->extradata_size = headerPtr.nFilledLen; > > + avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE); > > + if (!avctx->extradata) { > > + av_log(avctx, AV_LOG_ERROR, > > + "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size); > > + return AVERROR(ENOMEM); > > + } > > + memcpy(avctx->extradata, headerPtr.pBuffer, avctx->extradata_size); > > + av_freep(&headerPtr.pBuffer); > > + } > > + return 0; > > + > > +failed_init: > add av_free(q->svt_enc) ? Will double check the error handle (James given the same comments for this part) > > > + return error_mapping(ret); > > +} > > + > > +static int eb_send_frame(AVCodecContext *avctx, const AVFrame *frame) > > +{ > > + SvtContext *q = avctx->priv_data; > > + SvtEncoder *svt_enc = q->svt_enc; > > + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->in_buf; > > + int ret = 0; > > + > > + if (!frame) { > > + EB_BUFFERHEADERTYPE headerPtrLast; > > + headerPtrLast.nAllocLen = 0; > > + headerPtrLast.nFilledLen = 0; > > + headerPtrLast.nTickCount = 0; > > + headerPtrLast.pAppPrivate = NULL; > > + headerPtrLast.nOffset = 0; > > + headerPtrLast.pBuffer = NULL; > > + headerPtrLast.nFlags = EB_BUFFERFLAG_EOS; > > + EbH265EncSendPicture(svt_enc->svt_handle, &headerPtrLast); > > + q->eos_flag = 1; > > + av_log(avctx, AV_LOG_DEBUG, "Finish sending frames!!!\n"); > > + return ret; > > + } > > + > > + read_in_data(&svt_enc->enc_params, frame, headerPtr); > > + > > + headerPtr->nOffset = 0; > > + headerPtr->nFlags = 0; > > + headerPtr->pAppPrivate = NULL; > > + headerPtr->pts = frame->pts; > > + headerPtr->sliceType = INVALID_SLICE; > > + EbH265EncSendPicture(svt_enc->svt_handle, headerPtr); > > + > > + return ret; > > +} > > + > > +static int eb_receive_packet(AVCodecContext *avctx, AVPacket *pkt) > > +{ > > + SvtContext *q = avctx->priv_data; > > + SvtEncoder *svt_enc = q->svt_enc; > > + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->out_buf; > > + EB_ERRORTYPE stream_status = EB_ErrorNone; > > + int ret = 0; > > + > > + if ((ret = ff_alloc_packet2(avctx, pkt, svt_enc->raw_size, 0)) < 0) { > > + av_log(avctx, AV_LOG_ERROR, "Failed to allocate output packet.\n"); > > + return ret; > > + } > > + headerPtr->pBuffer = pkt->data; > > + stream_status = EbH265GetPacket(svt_enc->svt_handle, headerPtr, q->eos_flag); > > + if (stream_status == EB_NoErrorEmptyQueue) > > + return AVERROR(EAGAIN); > > + > > + pkt->size = headerPtr->nFilledLen; > > + pkt->pts = headerPtr->pts; > > + pkt->dts = headerPtr->dts; > > + if (headerPtr->sliceType == IDR_SLICE) > > + pkt->flags |= AV_PKT_FLAG_KEY; > > + if (headerPtr->sliceType == NON_REF_SLICE) > > + pkt->flags |= AV_PKT_FLAG_DISPOSABLE; > > + > > + ret = (headerPtr->nFlags & EB_BUFFERFLAG_EOS) ? AVERROR_EOF : 0; > > + return ret; > > +} > > + > > +static av_cold int eb_enc_close(AVCodecContext *avctx) > > +{ > > + SvtContext *q = avctx->priv_data; > > + SvtEncoder *svt_enc = q->svt_enc; > > + > > + EbDeinitEncoder(svt_enc->svt_handle); > > + EbDeinitHandle(svt_enc->svt_handle); > > + > > + free_buffer(svt_enc); > > + av_freep(&svt_enc); > > + > > + return 0; > > +} > > + > > +#define OFFSET(x) offsetof(SvtContext, x) > > +#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM > > +static const AVOption options[] = { > > + {"vui", "Enable vui info", OFFSET(svt_param.vui_info), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > > + {"hielevel", "Hierarchical Prediction Levels [0,3]", OFFSET(svt_param.hierarchical_level), AV_OPT_TYPE_INT, { .i64 = 3 }, 0, 3, VE }, > > + {"la_depth", "Look Ahead Distance [0,256]", OFFSET(svt_param.la_depth), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 256, VE }, > > + {"intra_ref_type", "Intra Refresh Type 0: No intra refresh 1: CRA (Open GOP) 2: IDR", OFFSET(svt_param.intra_ref_type), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 2, VE }, > > + {"enc_p", "Encoding preset [0,12] (for tune 0 and >=4k resolution), [0,10] (for >= 1080p resolution), [0,9] (for all resolution and modes)", OFFSET(svt_param.enc_mode), AV_OPT_TYPE_INT, { .i64 = 9 }, 0, 12, VE }, > > + {"profile", "Profile now support[1,2], Main Still Picture Profile not supported", OFFSET(svt_param.profile), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, 2, VE }, > > + {"rc", "RC mode 0: CQP 1: VBR", OFFSET(svt_param.rc_mode), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > > + {"q", "QP value for intra frames", OFFSET(svt_param.qp), AV_OPT_TYPE_INT, { .i64 = 32 }, 0, 51, VE }, > > + {"scd", "Scene change detection", OFFSET(svt_param.scd), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > > + {"tune", "Tune mode: SQ/OQ[0,1]", OFFSET(svt_param.tune), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, > > + {"bl_mode", "Random Access Prediction Structure Type", OFFSET(svt_param.base_layer_switch_mode), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, > > + {NULL}, > > +}; > > Need Document. Will supply a other patch for docs/man-page part. Tks. > > > + > > +static const AVClass class = { > > + .class_name = "libsvt_hevc", > > + .item_name = av_default_item_name, > > + .option = options, > > + .version = LIBAVUTIL_VERSION_INT, > > +}; > > + > > +static const AVCodecDefault eb_enc_defaults[] = { > > + { "b", "7M" }, > > + { "refs", "0" }, > > + { "g", "64" }, > > + { "flags", "+cgop" }, > > + { NULL }, > > +}; > > + > > +AVCodec ff_hevc_svt_encoder = { > > + .name = "libsvt_hevc", > > + .long_name = NULL_IF_CONFIG_SMALL("SVT-HEVC(Scalable Video Technology for HEVC) encoder"), > > + .priv_data_size = sizeof(SvtContext), > > + .type = AVMEDIA_TYPE_VIDEO, > > + .id = AV_CODEC_ID_HEVC, > > + .init = eb_enc_init, > > + .send_frame = eb_send_frame, > > + .receive_packet = eb_receive_packet, > > + .close = eb_enc_close, > > + .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS, > > + .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, > > + AV_PIX_FMT_YUV420P10, > > + AV_PIX_FMT_NONE }, > > + .priv_class = &class, > > + .defaults = eb_enc_defaults, > > + .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, > > + .wrapper_name = "libsvt_hevc", > > +}; > > -- > > 1.7.1 > > > >
On 12/2/2018 10:59 PM, mypopy@gmail.com wrote: > On Sun, Dec 2, 2018 at 10:19 PM James Almer <jamrial@gmail.com> wrote: >> >> On 12/2/2018 2:38 AM, Jun Zhao wrote: >>> base on patch by Huang, Zhengxu from https://github.com/intel/SVT-HEVC >>> >>> Signed-off-by: Huang, Zhengxu <zhengxu.huang@intel.com> >>> Signed-off-by: hassene <hassene.tmar@intel.com> >>> Signed-off-by: Jun Zhao <jun.zhao@intel.com> >>> --- >>> configure | 4 + >>> libavcodec/Makefile | 1 + >>> libavcodec/allcodecs.c | 1 + >>> libavcodec/libsvt_hevc.c | 370 > ++++++++++++++++++++++++++++++++++++++++++++++ >>> 4 files changed, 376 insertions(+), 0 deletions(-) >>> create mode 100644 libavcodec/libsvt_hevc.c >>> >>> diff --git a/configure b/configure >>> index 54b7e11..c3dca73 100755 >>> --- a/configure >>> +++ b/configure >>> @@ -263,6 +263,7 @@ External library support: >>> --enable-libspeex enable Speex de/encoding via libspeex [no] >>> --enable-libsrt enable Haivision SRT protocol via libsrt > [no] >>> --enable-libssh enable SFTP protocol via libssh [no] >>> + --enable-libsvt enable HEVC encoding via svt [no] >>> --enable-libtensorflow enable TensorFlow as a DNN module backend >>> for DNN based filters like sr [no] >>> --enable-libtesseract enable Tesseract, needed for ocr filter [no] >>> @@ -1665,6 +1666,7 @@ EXTERNAL_LIBRARY_GPL_LIST=" >>> libcdio >>> libdavs2 >>> librubberband >>> + libsvt >>> libvidstab >>> libx264 >>> libx265 >>> @@ -3087,6 +3089,7 @@ pcm_mulaw_at_encoder_select="audio_frame_queue" >>> chromaprint_muxer_deps="chromaprint" >>> h264_videotoolbox_encoder_deps="pthreads" >>> h264_videotoolbox_encoder_select="videotoolbox_encoder" >>> +hevc_svt_encoder_deps="libsvt" >> >> Since this is an external library based encoder, the correct name would >> be libsvt_encoder. >> You can also make it libsvt_hevc_encoder instead, which is what you used >> as AVCodec.name. >> > I perfer libsvt_hevc_encoder than libsvt_encoder, because maybe we will > have libsvt_{av1/vp9/xxx}_encoder in the future. Sounds good. [...] >>> +AVCodec ff_hevc_svt_encoder = { >>> + .name = "libsvt_hevc", >>> + .long_name = NULL_IF_CONFIG_SMALL("SVT-HEVC(Scalable Video > Technology for HEVC) encoder"), >>> + .priv_data_size = sizeof(SvtContext), >>> + .type = AVMEDIA_TYPE_VIDEO, >>> + .id = AV_CODEC_ID_HEVC, >>> + .init = eb_enc_init, >>> + .send_frame = eb_send_frame, >>> + .receive_packet = eb_receive_packet, >> >> Keep in mind that, at least for now and unlike with the decoupled >> input-output decode API, if you only provide a send/receive callback >> then the encoder will not work with avcodec_encode_video2(), only with >> avcodec_send_frame() and avcodec_receive_packet(). >> > Because avcodec_encode_video2() flags with deprecated, so I think is Ok if > we don't implement the encode2 callback. > > BTW: What's the reason we don't use avcodec_send_frame() and > avcodec_receive_packet() in avcodec_encode_video2() like use > avcodec_send_packet() and avcodec_receive_frame() in avcodec_decode_video2() > ? Because nobody implemented it yet, simple as that. The encoder equivalent of 061a0c14bb needs to be written. > > >>> + .close = eb_enc_close, >>> + .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS, >>> + .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, >>> + > AV_PIX_FMT_YUV420P10, >>> + AV_PIX_FMT_NONE }, >>> + .priv_class = &class, >>> + .defaults = eb_enc_defaults, >>> + .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, >>> + .wrapper_name = "libsvt_hevc", >>> +}; >>> >> > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel >
diff --git a/configure b/configure index 54b7e11..c3dca73 100755 --- a/configure +++ b/configure @@ -263,6 +263,7 @@ External library support: --enable-libspeex enable Speex de/encoding via libspeex [no] --enable-libsrt enable Haivision SRT protocol via libsrt [no] --enable-libssh enable SFTP protocol via libssh [no] + --enable-libsvt enable HEVC encoding via svt [no] --enable-libtensorflow enable TensorFlow as a DNN module backend for DNN based filters like sr [no] --enable-libtesseract enable Tesseract, needed for ocr filter [no] @@ -1665,6 +1666,7 @@ EXTERNAL_LIBRARY_GPL_LIST=" libcdio libdavs2 librubberband + libsvt libvidstab libx264 libx265 @@ -3087,6 +3089,7 @@ pcm_mulaw_at_encoder_select="audio_frame_queue" chromaprint_muxer_deps="chromaprint" h264_videotoolbox_encoder_deps="pthreads" h264_videotoolbox_encoder_select="videotoolbox_encoder" +hevc_svt_encoder_deps="libsvt" hevc_videotoolbox_encoder_deps="pthreads" hevc_videotoolbox_encoder_select="videotoolbox_encoder" libaom_av1_decoder_deps="libaom" @@ -6147,6 +6150,7 @@ enabled libsoxr && require libsoxr soxr.h soxr_create -lsoxr enabled libssh && require_pkg_config libssh libssh libssh/sftp.h sftp_init enabled libspeex && require_pkg_config libspeex speex speex/speex.h speex_decoder_init enabled libsrt && require_pkg_config libsrt "srt >= 1.3.0" srt/srt.h srt_socket +enabled libsvt && require_pkg_config libsvt svt EbApi.h EbInitHandle enabled libtensorflow && require libtensorflow tensorflow/c/c_api.h TF_Version -ltensorflow enabled libtesseract && require_pkg_config libtesseract tesseract tesseract/capi.h TessBaseAPICreate enabled libtheora && require libtheora theora/theoraenc.h th_info_init -ltheoraenc -ltheoradec -logg diff --git a/libavcodec/Makefile b/libavcodec/Makefile index 5feadac..1a8b547 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -374,6 +374,7 @@ OBJS-$(CONFIG_HEVC_QSV_DECODER) += qsvdec_h2645.o OBJS-$(CONFIG_HEVC_QSV_ENCODER) += qsvenc_hevc.o hevc_ps_enc.o \ hevc_data.o OBJS-$(CONFIG_HEVC_RKMPP_DECODER) += rkmppdec.o +OBJS-$(CONFIG_HEVC_SVT_ENCODER) += libsvt_hevc.o OBJS-$(CONFIG_HEVC_VAAPI_ENCODER) += vaapi_encode_h265.o h265_profile_level.o OBJS-$(CONFIG_HEVC_V4L2M2M_DECODER) += v4l2_m2m_dec.o OBJS-$(CONFIG_HEVC_V4L2M2M_ENCODER) += v4l2_m2m_enc.o diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index d70646e..ff8ea94 100644 --- a/libavcodec/allcodecs.c +++ b/libavcodec/allcodecs.c @@ -749,6 +749,7 @@ extern AVCodec ff_hevc_cuvid_decoder; extern AVCodec ff_hevc_mediacodec_decoder; extern AVCodec ff_hevc_nvenc_encoder; extern AVCodec ff_hevc_qsv_encoder; +extern AVCodec ff_hevc_svt_encoder; extern AVCodec ff_hevc_v4l2m2m_encoder; extern AVCodec ff_hevc_vaapi_encoder; extern AVCodec ff_hevc_videotoolbox_encoder; diff --git a/libavcodec/libsvt_hevc.c b/libavcodec/libsvt_hevc.c new file mode 100644 index 0000000..1e10111 --- /dev/null +++ b/libavcodec/libsvt_hevc.c @@ -0,0 +1,370 @@ +/* +* Scalable Video Technology for HEVC encoder library plugin +* +* Copyright (c) 2018 Intel Corporation +* +* This program is free software; you can redistribute it and/or +* modify it under the terms of the GNU Lesser General Public +* License as published by the Free Software Foundation; either +* version 2.1 of the License, or (at your option) any later version. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +* Lesser General Public License for more details. +* +* You should have received a copy of the GNU Lesser General Public +* License along with this program; if not, write to the Free Software +* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA +*/ + +#include "EbErrorCodes.h" +#include "EbTime.h" +#include "EbApi.h" + +#include "libavutil/common.h" +#include "libavutil/frame.h" +#include "libavutil/opt.h" + +#include "internal.h" +#include "avcodec.h" + +typedef struct SvtEncoder { + EB_H265_ENC_CONFIGURATION enc_params; + EB_COMPONENTTYPE *svt_handle; + EB_BUFFERHEADERTYPE *in_buf; + EB_BUFFERHEADERTYPE *out_buf; + int raw_size; +} SvtEncoder; + +typedef struct SvtParams { + int vui_info; + int hierarchical_level; + int la_depth; + int intra_ref_type; + int enc_mode; + int rc_mode; + int scd; + int tune; + int qp; + int profile; + int base_layer_switch_mode; +}SvtParams; + +typedef struct SvtContext { + AVClass *class; + SvtEncoder *svt_enc; + SvtParams svt_param; + int eos_flag; +} SvtContext; + +static void free_buffer(SvtEncoder *svt_enc) +{ + if (svt_enc->in_buf) { + EB_H265_ENC_INPUT *in_data = (EB_H265_ENC_INPUT *)svt_enc->in_buf->pBuffer; + av_freep(&in_data); + av_freep(&svt_enc->in_buf); + } + av_freep(&svt_enc->out_buf); +} + +static EB_ERRORTYPE alloc_buffer(EB_H265_ENC_CONFIGURATION *config, SvtEncoder *svt_enc) +{ + EB_ERRORTYPE ret = EB_ErrorNone; + + const int pack_mode_10bit = (config->encoderBitDepth > 8) && (config->compressedTenBitFormat == 0) ? 1 : 0; + const size_t luma_size_8bit = config->sourceWidth * config->sourceHeight * (1 << pack_mode_10bit); + const size_t luma_size_10bit = (config->encoderBitDepth > 8 && pack_mode_10bit == 0) ? luma_size_8bit : 0; + + svt_enc->raw_size = (luma_size_8bit + luma_size_10bit) * 3 / 2; + + // allocate buffer for in and out + svt_enc->in_buf = av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); + svt_enc->out_buf = av_mallocz(sizeof(EB_BUFFERHEADERTYPE)); + if (!svt_enc->in_buf || !svt_enc->out_buf) + goto failed; + + svt_enc->in_buf->pBuffer = av_mallocz(sizeof(EB_H265_ENC_INPUT)); + if (!svt_enc->in_buf->pBuffer) + goto failed; + + svt_enc->in_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); + svt_enc->in_buf->pAppPrivate = NULL; + svt_enc->out_buf->nSize = sizeof(EB_BUFFERHEADERTYPE); + svt_enc->out_buf->nAllocLen = svt_enc->raw_size; + svt_enc->out_buf->pAppPrivate = NULL; + + return ret; + +failed: + free_buffer(svt_enc); + return AVERROR(ENOMEM); +} + +static int error_mapping(int val) +{ + if (val == EB_ErrorInsufficientResources) + return AVERROR(ENOMEM); + if ((val == EB_ErrorUndefined) || (val == EB_ErrorInvalidComponent) || + (val == EB_ErrorBadParameter)) + return AVERROR(EINVAL); + return AVERROR_EXTERNAL; +} + +static EB_ERRORTYPE config_enc_params(EB_H265_ENC_CONFIGURATION *param, AVCodecContext *avctx) +{ + SvtContext *q = avctx->priv_data; + SvtEncoder *svt_enc = q->svt_enc; + EB_ERRORTYPE ret = EB_ErrorNone; + int tenBits = 0; + + param->sourceWidth = avctx->width; + param->sourceHeight = avctx->height; + + if (avctx->pix_fmt == AV_PIX_FMT_YUV420P10LE) { + av_log(avctx, AV_LOG_DEBUG , "Encoder 10 bits depth input\n"); + param->compressedTenBitFormat = 0; + tenBits = 1; + } + + // Update param from options + param->hierarchicalLevels = q->svt_param.hierarchical_level; + param->encMode = q->svt_param.enc_mode; + param->intraRefreshType = q->svt_param.intra_ref_type; + param->profile = q->svt_param.profile; + param->rateControlMode = q->svt_param.rc_mode; + param->sceneChangeDetection = q->svt_param.scd; + param->tune = q->svt_param.tune; + param->baseLayerSwitchMode = q->svt_param.base_layer_switch_mode; + param->qp = q->svt_param.qp; + + param->targetBitRate = avctx->bit_rate; + param->intraPeriodLength = avctx->gop_size-1; + param->frameRateNumerator = avctx->time_base.den; + param->frameRateDenominator = avctx->time_base.num * avctx->ticks_per_frame; + + param->codeVpsSpsPps = 0; + + if (q->svt_param.vui_info) + param->videoUsabilityInfo = q->svt_param.vui_info; + if (q->svt_param.la_depth != -1) + param->lookAheadDistance = q->svt_param.la_depth; + + if (tenBits == 1) { + param->encoderBitDepth = 10; + param->profile = 2; + } + + ret = alloc_buffer(param, svt_enc); + + return ret; +} + +static void read_in_data(EB_H265_ENC_CONFIGURATION *config, const AVFrame* frame, EB_BUFFERHEADERTYPE *headerPtr) +{ + unsigned int is16bit = config->encoderBitDepth > 8; + unsigned long long lumaReadSize = (unsigned long long)config->sourceWidth * config->sourceHeight<< is16bit; + EB_H265_ENC_INPUT *in_data = (EB_H265_ENC_INPUT*)headerPtr->pBuffer; + + // support yuv420p and yuv420p010 + in_data->luma = frame->data[0]; + in_data->cb = frame->data[1]; + in_data->cr = frame->data[2]; + + // stride info + in_data->yStride = frame->linesize[0] >> is16bit; + in_data->cbStride = frame->linesize[1] >> is16bit; + in_data->crStride = frame->linesize[2] >> is16bit; + + headerPtr->nFilledLen += lumaReadSize * 3/2u; +} + +static av_cold int eb_enc_init(AVCodecContext *avctx) +{ + SvtContext *q = avctx->priv_data; + SvtEncoder *svt_enc = NULL; + EB_ERRORTYPE ret = EB_ErrorNone; + + q->svt_enc = av_mallocz(sizeof(*q->svt_enc)); + if (!q->svt_enc) + return AVERROR(ENOMEM); + svt_enc = q->svt_enc; + + q->eos_flag = 0; + + ret = EbInitHandle(&svt_enc->svt_handle, q, &svt_enc->enc_params); + if (ret != EB_ErrorNone) + goto failed_init; + + ret = config_enc_params(&svt_enc->enc_params, avctx); + if (ret != EB_ErrorNone) + goto failed_init; + + ret = EbH265EncSetParameter(svt_enc->svt_handle, &svt_enc->enc_params); + if (ret != EB_ErrorNone) + goto failed_init; + + ret = EbInitEncoder(svt_enc->svt_handle); + if (ret != EB_ErrorNone) + goto failed_init; + + if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) { + + EB_BUFFERHEADERTYPE headerPtr; + headerPtr.nSize = sizeof(EB_BUFFERHEADERTYPE); + headerPtr.nFilledLen = 0; + headerPtr.pBuffer = av_malloc(10 * 1024 * 1024); + headerPtr.nAllocLen = (10 * 1024 * 1024); + + if (!headerPtr.pBuffer) + return AVERROR(ENOMEM); + + ret = EbH265EncStreamHeader(svt_enc->svt_handle, &headerPtr); + if (ret != EB_ErrorNone) { + av_freep(&headerPtr.pBuffer); + goto failed_init; + } + avctx->extradata_size = headerPtr.nFilledLen; + avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE); + if (!avctx->extradata) { + av_log(avctx, AV_LOG_ERROR, + "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size); + return AVERROR(ENOMEM); + } + memcpy(avctx->extradata, headerPtr.pBuffer, avctx->extradata_size); + av_freep(&headerPtr.pBuffer); + } + return 0; + +failed_init: + return error_mapping(ret); +} + +static int eb_send_frame(AVCodecContext *avctx, const AVFrame *frame) +{ + SvtContext *q = avctx->priv_data; + SvtEncoder *svt_enc = q->svt_enc; + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->in_buf; + int ret = 0; + + if (!frame) { + EB_BUFFERHEADERTYPE headerPtrLast; + headerPtrLast.nAllocLen = 0; + headerPtrLast.nFilledLen = 0; + headerPtrLast.nTickCount = 0; + headerPtrLast.pAppPrivate = NULL; + headerPtrLast.nOffset = 0; + headerPtrLast.pBuffer = NULL; + headerPtrLast.nFlags = EB_BUFFERFLAG_EOS; + EbH265EncSendPicture(svt_enc->svt_handle, &headerPtrLast); + q->eos_flag = 1; + av_log(avctx, AV_LOG_DEBUG, "Finish sending frames!!!\n"); + return ret; + } + + read_in_data(&svt_enc->enc_params, frame, headerPtr); + + headerPtr->nOffset = 0; + headerPtr->nFlags = 0; + headerPtr->pAppPrivate = NULL; + headerPtr->pts = frame->pts; + headerPtr->sliceType = INVALID_SLICE; + EbH265EncSendPicture(svt_enc->svt_handle, headerPtr); + + return ret; +} + +static int eb_receive_packet(AVCodecContext *avctx, AVPacket *pkt) +{ + SvtContext *q = avctx->priv_data; + SvtEncoder *svt_enc = q->svt_enc; + EB_BUFFERHEADERTYPE *headerPtr = svt_enc->out_buf; + EB_ERRORTYPE stream_status = EB_ErrorNone; + int ret = 0; + + if ((ret = ff_alloc_packet2(avctx, pkt, svt_enc->raw_size, 0)) < 0) { + av_log(avctx, AV_LOG_ERROR, "Failed to allocate output packet.\n"); + return ret; + } + headerPtr->pBuffer = pkt->data; + stream_status = EbH265GetPacket(svt_enc->svt_handle, headerPtr, q->eos_flag); + if (stream_status == EB_NoErrorEmptyQueue) + return AVERROR(EAGAIN); + + pkt->size = headerPtr->nFilledLen; + pkt->pts = headerPtr->pts; + pkt->dts = headerPtr->dts; + if (headerPtr->sliceType == IDR_SLICE) + pkt->flags |= AV_PKT_FLAG_KEY; + if (headerPtr->sliceType == NON_REF_SLICE) + pkt->flags |= AV_PKT_FLAG_DISPOSABLE; + + ret = (headerPtr->nFlags & EB_BUFFERFLAG_EOS) ? AVERROR_EOF : 0; + return ret; +} + +static av_cold int eb_enc_close(AVCodecContext *avctx) +{ + SvtContext *q = avctx->priv_data; + SvtEncoder *svt_enc = q->svt_enc; + + EbDeinitEncoder(svt_enc->svt_handle); + EbDeinitHandle(svt_enc->svt_handle); + + free_buffer(svt_enc); + av_freep(&svt_enc); + + return 0; +} + +#define OFFSET(x) offsetof(SvtContext, x) +#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM +static const AVOption options[] = { + {"vui", "Enable vui info", OFFSET(svt_param.vui_info), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, + {"hielevel", "Hierarchical Prediction Levels [0,3]", OFFSET(svt_param.hierarchical_level), AV_OPT_TYPE_INT, { .i64 = 3 }, 0, 3, VE }, + {"la_depth", "Look Ahead Distance [0,256]", OFFSET(svt_param.la_depth), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 256, VE }, + {"intra_ref_type", "Intra Refresh Type 0: No intra refresh 1: CRA (Open GOP) 2: IDR", OFFSET(svt_param.intra_ref_type), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 2, VE }, + {"enc_p", "Encoding preset [0,12] (for tune 0 and >=4k resolution), [0,10] (for >= 1080p resolution), [0,9] (for all resolution and modes)", OFFSET(svt_param.enc_mode), AV_OPT_TYPE_INT, { .i64 = 9 }, 0, 12, VE }, + {"profile", "Profile now support[1,2], Main Still Picture Profile not supported", OFFSET(svt_param.profile), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, 2, VE }, + {"rc", "RC mode 0: CQP 1: VBR", OFFSET(svt_param.rc_mode), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, + {"q", "QP value for intra frames", OFFSET(svt_param.qp), AV_OPT_TYPE_INT, { .i64 = 32 }, 0, 51, VE }, + {"scd", "Scene change detection", OFFSET(svt_param.scd), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, + {"tune", "Tune mode: SQ/OQ[0,1]", OFFSET(svt_param.tune), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE }, + {"bl_mode", "Random Access Prediction Structure Type", OFFSET(svt_param.base_layer_switch_mode), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, + {NULL}, +}; + +static const AVClass class = { + .class_name = "libsvt_hevc", + .item_name = av_default_item_name, + .option = options, + .version = LIBAVUTIL_VERSION_INT, +}; + +static const AVCodecDefault eb_enc_defaults[] = { + { "b", "7M" }, + { "refs", "0" }, + { "g", "64" }, + { "flags", "+cgop" }, + { NULL }, +}; + +AVCodec ff_hevc_svt_encoder = { + .name = "libsvt_hevc", + .long_name = NULL_IF_CONFIG_SMALL("SVT-HEVC(Scalable Video Technology for HEVC) encoder"), + .priv_data_size = sizeof(SvtContext), + .type = AVMEDIA_TYPE_VIDEO, + .id = AV_CODEC_ID_HEVC, + .init = eb_enc_init, + .send_frame = eb_send_frame, + .receive_packet = eb_receive_packet, + .close = eb_enc_close, + .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS, + .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, + AV_PIX_FMT_YUV420P10, + AV_PIX_FMT_NONE }, + .priv_class = &class, + .defaults = eb_enc_defaults, + .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, + .wrapper_name = "libsvt_hevc", +};