2 * Intel MediaSDK QSV based H.264 enccoder
4 * copyright (c) 2013 Yukinori Yamazoe
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 #include <sys/types.h>
27 #include <mfx/mfxvideo.h>
29 #include "libavutil/common.h"
30 #include "libavutil/opt.h"
35 #include "qsv_internal.h"
38 typedef struct QSVH264EncContext {
43 static int qsv_h264_set_encode_ctrl(AVCodecContext *avctx,
44 const AVFrame *frame, mfxEncodeCtrl* enc_ctrl)
46 QSVH264EncContext *qh264 = avctx->priv_data;
47 QSVEncContext *q = &qh264->qsv;
49 if (q->a53_cc && frame) {
55 res = ff_alloc_a53_sei(frame, sizeof(mfxPayload) + 2, (void**)&payload, &sei_size);
56 if (res < 0 || !payload)
59 sei_data = (mfxU8*)(payload + 1);
62 sei_data[1] = (mfxU8)sei_size; // size of SEI data
63 // SEI data filled in by ff_alloc_a53_sei
65 payload->BufSize = sei_size + 2;
66 payload->NumBit = payload->BufSize * 8;
68 payload->Data = sei_data;
70 enc_ctrl->NumExtParam = 0;
71 enc_ctrl->NumPayload = 1;
72 enc_ctrl->Payload[0] = payload;
77 static av_cold int qsv_enc_init(AVCodecContext *avctx)
79 QSVH264EncContext *q = avctx->priv_data;
81 q->qsv.set_encode_ctrl_cb = qsv_h264_set_encode_ctrl;
82 return ff_qsv_enc_init(avctx, &q->qsv);
85 static int qsv_enc_frame(AVCodecContext *avctx, AVPacket *pkt,
86 const AVFrame *frame, int *got_packet)
88 QSVH264EncContext *q = avctx->priv_data;
90 return ff_qsv_encode(avctx, &q->qsv, pkt, frame, got_packet);
93 static av_cold int qsv_enc_close(AVCodecContext *avctx)
95 QSVH264EncContext *q = avctx->priv_data;
97 return ff_qsv_enc_close(avctx, &q->qsv);
100 #define OFFSET(x) offsetof(QSVH264EncContext, x)
101 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
102 static const AVOption options[] = {
105 { "cavlc", "Enable CAVLC", OFFSET(qsv.cavlc), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
107 { "vcm", "Use the video conferencing mode ratecontrol", OFFSET(qsv.vcm), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
109 { "idr_interval", "Distance (in I-frames) between IDR frames", OFFSET(qsv.idr_interval), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
110 { "pic_timing_sei", "Insert picture timing SEI with pic_struct_syntax element", OFFSET(qsv.pic_timing_sei), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
111 { "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 },
112 { "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 },
115 { "look_ahead", "Use VBR algorithm with look ahead", OFFSET(qsv.look_ahead), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
116 { "look_ahead_depth", "Depth of look ahead in number frames", OFFSET(qsv.look_ahead_depth), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 100, VE },
119 { "look_ahead_downsampling", "Downscaling factor for the frames saved for the lookahead analysis", OFFSET(qsv.look_ahead_downsampling),
120 AV_OPT_TYPE_INT, { .i64 = MFX_LOOKAHEAD_DS_UNKNOWN }, MFX_LOOKAHEAD_DS_UNKNOWN, MFX_LOOKAHEAD_DS_4x, VE, "look_ahead_downsampling" },
121 { "unknown" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_UNKNOWN }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
122 { "auto" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_UNKNOWN }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
123 { "off" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_OFF }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
124 { "2x" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_2x }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
125 { "4x" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_4x }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
128 { "int_ref_type", "Intra refresh type", OFFSET(qsv.int_ref_type), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, UINT16_MAX, VE, "int_ref_type" },
129 { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, .flags = VE, "int_ref_type" },
130 { "vertical", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, .flags = VE, "int_ref_type" },
131 { "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 },
132 { "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 },
133 { "recovery_point_sei", "Insert recovery point SEI messages", OFFSET(qsv.recovery_point_sei), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 1, VE },
135 { "trellis", "Trellis quantization", OFFSET(qsv.trellis), AV_OPT_TYPE_FLAGS, { .i64 = 0 }, 0, UINT_MAX, VE, "trellis" },
136 { "off", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_TRELLIS_OFF }, .flags = VE, "trellis" },
137 { "I", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_TRELLIS_I }, .flags = VE, "trellis" },
138 { "P", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_TRELLIS_P }, .flags = VE, "trellis" },
139 { "B", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_TRELLIS_B }, .flags = VE, "trellis" },
141 { "profile", NULL, OFFSET(qsv.profile), AV_OPT_TYPE_INT, { .i64 = MFX_PROFILE_UNKNOWN }, 0, INT_MAX, VE, "profile" },
142 { "unknown" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_UNKNOWN }, INT_MIN, INT_MAX, VE, "profile" },
143 { "baseline", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_AVC_BASELINE }, INT_MIN, INT_MAX, VE, "profile" },
144 { "main" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_AVC_MAIN }, INT_MIN, INT_MAX, VE, "profile" },
145 { "high" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_AVC_HIGH }, INT_MIN, INT_MAX, VE, "profile" },
147 { "a53cc" , "Use A53 Closed Captions (if available)", OFFSET(qsv.a53_cc), AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, VE},
149 { "aud", "Insert the Access Unit Delimiter NAL", OFFSET(qsv.aud), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE},
152 { "mfmode", "Multi-Frame Mode", OFFSET(qsv.mfmode), AV_OPT_TYPE_INT, { .i64 = MFX_MF_AUTO }, MFX_MF_DEFAULT, MFX_MF_AUTO, VE, "mfmode"},
153 { "off" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_MF_DISABLED }, INT_MIN, INT_MAX, VE, "mfmode" },
154 { "auto" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_MF_AUTO }, INT_MIN, INT_MAX, VE, "mfmode" },
160 static const AVClass class = {
161 .class_name = "h264_qsv encoder",
162 .item_name = av_default_item_name,
164 .version = LIBAVUTIL_VERSION_INT,
167 static const AVCodecDefault qsv_enc_defaults[] = {
170 // same as the x264 default
173 #if FF_API_CODER_TYPE
177 { "flags", "+cgop" },
178 #if FF_API_PRIVATE_OPT
179 { "b_strategy", "-1" },
184 AVCodec ff_h264_qsv_encoder = {
186 .long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (Intel Quick Sync Video acceleration)"),
187 .priv_data_size = sizeof(QSVH264EncContext),
188 .type = AVMEDIA_TYPE_VIDEO,
189 .id = AV_CODEC_ID_H264,
190 .init = qsv_enc_init,
191 .encode2 = qsv_enc_frame,
192 .close = qsv_enc_close,
193 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_HYBRID,
194 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_NV12,
198 .priv_class = &class,
199 .defaults = qsv_enc_defaults,
200 .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
201 .wrapper_name = "qsv",