ffmpeg/libavcodec/qsvenc_h264.c
galinart 5002829416 libavcodec/qsvenc: enable Hyper Encode
Hyper Encode uses Intel integrated and discrete graphics on one system
to accelerate encoding of a single video stream.
Depending on the selected parameters and codecs, performance gain on AlderLake iGPU + ARC Gfx up to 1.6x.

More information: https://www.intel.co.uk/content/www/uk/en/architecture-and-technology/adaptix/deep-link.html
Developer guide: https://github.com/oneapi-src/oneVPL-intel-gpu/blob/main/doc/HyperEncode_FeatureDeveloperGuide.md

Hyper Encode is supported only on Windows and requires D3D11 and oneVPL.

To enable Hyper Encode need to specify:
-Hyper Encode mode (-dual_gfx on or dual_gfx adaptive)
-Encoder: h264_qsv or hevc_qsv
-BRC: VBR, CQP or ICQ
-Lowpower mode (-low_power 1)
-Closed GOP for AVC or strict GOP for HEVC -idr_interval = 0 used by default

Depending on the encoding parameters, the following parameters may need
to be adjusted:
-g recommended >= 30 for better performance
-async_depth recommended >= 30 for better performance
-extra_hw_frames recommended equal to async_depth value
-bf recommended = 0 for better performance

In the cases with fast encoding (-preset veryfast) there may be no
performance gain due to the fact that the decode is slower than the encode.

Command line examples:

ffmpeg.exe -init_hw_device qsv:hw,child_device_type=d3d11va,child_device=0 -v verbose -y -hwaccel qsv -extra_hw_frames 60 -async_depth 60 -c:v h264_qsv -i bbb_sunflower_2160p_60fps_normal.mp4
-async_depth 60 -c:v h264_qsv -preset medium -g 60 -low_power 1 -bf 0 -dual_gfx on output.h265

Signed-off-by: galinart <artem.galin@intel.com>
2023-01-30 08:48:58 +08:00

212 lines
9.6 KiB
C

/*
* Intel MediaSDK QSV based H.264 encoder
*
* copyright (c) 2013 Yukinori Yamazoe
*
* This file is part of FFmpeg.
*
* FFmpeg 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.
*
* FFmpeg 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 FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include <stdint.h>
#include <sys/types.h>
#include <mfxvideo.h>
#include "libavutil/common.h"
#include "libavutil/opt.h"
#include "avcodec.h"
#include "codec_internal.h"
#include "qsv.h"
#include "qsvenc.h"
#include "atsc_a53.h"
typedef struct QSVH264EncContext {
AVClass *class;
QSVEncContext qsv;
} QSVH264EncContext;
static int qsv_h264_set_encode_ctrl(AVCodecContext *avctx,
const AVFrame *frame, mfxEncodeCtrl* enc_ctrl)
{
QSVH264EncContext *qh264 = avctx->priv_data;
QSVEncContext *q = &qh264->qsv;
if (q->a53_cc && frame) {
mfxPayload* payload;
mfxU8* sei_data;
size_t sei_size;
int res;
res = ff_alloc_a53_sei(frame, sizeof(mfxPayload) + 2, (void**)&payload, &sei_size);
if (res < 0 || !payload)
return res;
sei_data = (mfxU8*)(payload + 1);
// SEI header
sei_data[0] = 4;
sei_data[1] = (mfxU8)sei_size; // size of SEI data
// SEI data filled in by ff_alloc_a53_sei
payload->BufSize = sei_size + 2;
payload->NumBit = payload->BufSize * 8;
payload->Type = 4;
payload->Data = sei_data;
enc_ctrl->NumExtParam = 0;
enc_ctrl->NumPayload = 1;
enc_ctrl->Payload[0] = payload;
}
return 0;
}
static av_cold int qsv_enc_init(AVCodecContext *avctx)
{
QSVH264EncContext *q = avctx->priv_data;
q->qsv.set_encode_ctrl_cb = qsv_h264_set_encode_ctrl;
return ff_qsv_enc_init(avctx, &q->qsv);
}
static int qsv_enc_frame(AVCodecContext *avctx, AVPacket *pkt,
const AVFrame *frame, int *got_packet)
{
QSVH264EncContext *q = avctx->priv_data;
return ff_qsv_encode(avctx, &q->qsv, pkt, frame, got_packet);
}
static av_cold int qsv_enc_close(AVCodecContext *avctx)
{
QSVH264EncContext *q = avctx->priv_data;
return ff_qsv_enc_close(avctx, &q->qsv);
}
#define OFFSET(x) offsetof(QSVH264EncContext, x)
#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
static const AVOption options[] = {
QSV_COMMON_OPTS
QSV_OPTION_RDO
QSV_OPTION_MAX_FRAME_SIZE
QSV_OPTION_MAX_SLICE_SIZE
QSV_OPTION_BITRATE_LIMIT
QSV_OPTION_MBBRC
QSV_OPTION_EXTBRC
QSV_OPTION_ADAPTIVE_I
QSV_OPTION_ADAPTIVE_B
QSV_OPTION_P_STRATEGY
QSV_OPTION_B_STRATEGY
QSV_OPTION_DBLK_IDC
QSV_OPTION_LOW_DELAY_BRC
QSV_OPTION_MAX_MIN_QP
QSV_OPTION_SCENARIO
QSV_OPTION_AVBR
QSV_OPTION_SKIP_FRAME
#if QSV_HAVE_HE
QSV_HE_OPTIONS
#endif
{ "cavlc", "Enable CAVLC", OFFSET(qsv.cavlc), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
#if QSV_HAVE_VCM
{ "vcm", "Use the video conferencing mode ratecontrol", OFFSET(qsv.vcm), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
#endif
{ "idr_interval", "Distance (in I-frames) between IDR frames", OFFSET(qsv.idr_interval), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
{ "pic_timing_sei", "Insert picture timing SEI with pic_struct_syntax element", OFFSET(qsv.pic_timing_sei), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, VE },
{ "single_sei_nal_unit", "Put all the SEI messages into one NALU", OFFSET(qsv.single_sei_nal_unit), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 1, VE },
{ "max_dec_frame_buffering", "Maximum number of frames buffered in the DPB", OFFSET(qsv.max_dec_frame_buffering), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, UINT16_MAX, VE },
{ "look_ahead", "Use VBR algorithm with look ahead", OFFSET(qsv.look_ahead), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
{ "look_ahead_depth", "Depth of look ahead in number frames", OFFSET(qsv.look_ahead_depth), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 100, VE },
{ "look_ahead_downsampling", "Downscaling factor for the frames saved for the lookahead analysis", OFFSET(qsv.look_ahead_downsampling),
AV_OPT_TYPE_INT, { .i64 = MFX_LOOKAHEAD_DS_UNKNOWN }, MFX_LOOKAHEAD_DS_UNKNOWN, MFX_LOOKAHEAD_DS_4x, VE, "look_ahead_downsampling" },
{ "unknown" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_UNKNOWN }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
{ "auto" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_UNKNOWN }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
{ "off" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_OFF }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
{ "2x" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_2x }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
{ "4x" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_4x }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
{ "int_ref_type", "Intra refresh type. B frames should be set to 0.", OFFSET(qsv.int_ref_type), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, UINT16_MAX, VE, "int_ref_type" },
{ "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, .flags = VE, "int_ref_type" },
{ "vertical", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, .flags = VE, "int_ref_type" },
{ "horizontal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 2 }, .flags = VE, "int_ref_type" },
{ "slice" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 3 }, .flags = VE, "int_ref_type" },
{ "int_ref_cycle_size", "Number of frames in the intra refresh cycle", OFFSET(qsv.int_ref_cycle_size), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, UINT16_MAX, VE },
{ "int_ref_qp_delta", "QP difference for the refresh MBs", OFFSET(qsv.int_ref_qp_delta), AV_OPT_TYPE_INT, { .i64 = INT16_MIN }, INT16_MIN, INT16_MAX, VE },
{ "recovery_point_sei", "Insert recovery point SEI messages", OFFSET(qsv.recovery_point_sei), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 1, VE },
{ "int_ref_cycle_dist", "Distance between the beginnings of the intra-refresh cycles in frames", OFFSET(qsv.int_ref_cycle_dist), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT16_MAX, VE },
{ "profile", NULL, OFFSET(qsv.profile), AV_OPT_TYPE_INT, { .i64 = MFX_PROFILE_UNKNOWN }, 0, INT_MAX, VE, "profile" },
{ "unknown" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_UNKNOWN }, INT_MIN, INT_MAX, VE, "profile" },
{ "baseline", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_AVC_BASELINE }, INT_MIN, INT_MAX, VE, "profile" },
{ "main" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_AVC_MAIN }, INT_MIN, INT_MAX, VE, "profile" },
{ "high" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_AVC_HIGH }, INT_MIN, INT_MAX, VE, "profile" },
{ "a53cc" , "Use A53 Closed Captions (if available)", OFFSET(qsv.a53_cc), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, VE},
{ "aud", "Insert the Access Unit Delimiter NAL", OFFSET(qsv.aud), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE},
#if QSV_HAVE_MF
{ "mfmode", "Multi-Frame Mode", OFFSET(qsv.mfmode), AV_OPT_TYPE_INT, { .i64 = MFX_MF_AUTO }, MFX_MF_DEFAULT, MFX_MF_AUTO, VE, "mfmode"},
{ "off" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_MF_DISABLED }, INT_MIN, INT_MAX, VE, "mfmode" },
{ "auto" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_MF_AUTO }, INT_MIN, INT_MAX, VE, "mfmode" },
#endif
{ "repeat_pps", "repeat pps for every frame", OFFSET(qsv.repeat_pps), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
{ NULL },
};
static const AVClass class = {
.class_name = "h264_qsv encoder",
.item_name = av_default_item_name,
.option = options,
.version = LIBAVUTIL_VERSION_INT,
};
static const FFCodecDefault qsv_enc_defaults[] = {
{ "b", "1M" },
{ "refs", "0" },
{ "g", "-1" },
{ "bf", "-1" },
{ "qmin", "-1" },
{ "qmax", "-1" },
{ "trellis", "-1" },
{ "flags", "+cgop" },
{ NULL },
};
const FFCodec ff_h264_qsv_encoder = {
.p.name = "h264_qsv",
CODEC_LONG_NAME("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (Intel Quick Sync Video acceleration)"),
.priv_data_size = sizeof(QSVH264EncContext),
.p.type = AVMEDIA_TYPE_VIDEO,
.p.id = AV_CODEC_ID_H264,
.init = qsv_enc_init,
FF_CODEC_ENCODE_CB(qsv_enc_frame),
.close = qsv_enc_close,
.p.capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_HYBRID,
.p.pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_NV12,
AV_PIX_FMT_QSV,
AV_PIX_FMT_NONE },
.p.priv_class = &class,
.defaults = qsv_enc_defaults,
.caps_internal = FF_CODEC_CAP_NOT_INIT_THREADSAFE |
FF_CODEC_CAP_INIT_CLEANUP,
.p.wrapper_name = "qsv",
.hw_configs = ff_qsv_enc_hw_configs,
};