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/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 AVFrameSideData *side_data = NULL;
47 QSVH264EncContext *qh264 = avctx->priv_data;
48 QSVEncContext *q = &qh264->qsv;
50 if (q->a53_cc && frame) {
51 side_data = av_frame_get_side_data(frame, AV_FRAME_DATA_A53_CC);
54 int sei_payload_size = 0;
55 mfxU8* sei_data = NULL;
56 mfxPayload* payload = NULL;
58 sei_payload_size = side_data->size + 13;
60 sei_data = av_mallocz(sei_payload_size);
62 av_log(avctx, AV_LOG_ERROR, "No memory for CC, skipping...\n");
63 return AVERROR(ENOMEM);
68 sei_data[1] = sei_payload_size - 2; // size of SEI data
75 // ATSC_identifier - using 'GA94' only
77 MKTAG('G', 'A', '9', '4'));
80 ((side_data->size/3) & 0x1f) | 0xC0;
82 sei_data[11] = 0xFF; // reserved
84 memcpy(sei_data + 12, side_data->data, side_data->size);
86 sei_data[side_data->size+12] = 255;
88 payload = av_mallocz(sizeof(mfxPayload));
90 av_log(avctx, AV_LOG_ERROR, "No memory, skipping captions\n");
92 return AVERROR(ENOMEM);
94 payload->BufSize = side_data->size + 13;
95 payload->NumBit = payload->BufSize * 8;
97 payload->Data = sei_data;
99 enc_ctrl->NumExtParam = 0;
100 enc_ctrl->NumPayload = 1;
101 enc_ctrl->Payload[0] = payload;
107 static av_cold int qsv_enc_init(AVCodecContext *avctx)
109 QSVH264EncContext *q = avctx->priv_data;
111 q->qsv.set_encode_ctrl_cb = qsv_h264_set_encode_ctrl;
112 return ff_qsv_enc_init(avctx, &q->qsv);
115 static int qsv_enc_frame(AVCodecContext *avctx, AVPacket *pkt,
116 const AVFrame *frame, int *got_packet)
118 QSVH264EncContext *q = avctx->priv_data;
120 return ff_qsv_encode(avctx, &q->qsv, pkt, frame, got_packet);
123 static av_cold int qsv_enc_close(AVCodecContext *avctx)
125 QSVH264EncContext *q = avctx->priv_data;
127 return ff_qsv_enc_close(avctx, &q->qsv);
130 #define OFFSET(x) offsetof(QSVH264EncContext, x)
131 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
132 static const AVOption options[] = {
135 { "idr_interval", "Distance (in I-frames) between IDR frames", OFFSET(qsv.idr_interval), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
136 { "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 },
137 { "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 },
138 { "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 },
141 { "look_ahead", "Use VBR algorithm with look ahead", OFFSET(qsv.look_ahead), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
142 { "look_ahead_depth", "Depth of look ahead in number frames", OFFSET(qsv.look_ahead_depth), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 100, VE },
145 #if QSV_VERSION_ATLEAST(1,8)
146 { "look_ahead_downsampling", NULL, OFFSET(qsv.look_ahead_downsampling), AV_OPT_TYPE_INT, { .i64 = MFX_LOOKAHEAD_DS_UNKNOWN }, MFX_LOOKAHEAD_DS_UNKNOWN, MFX_LOOKAHEAD_DS_2x, VE, "look_ahead_downsampling" },
147 { "unknown" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_UNKNOWN }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
148 { "off" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_OFF }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
149 { "2x" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_LOOKAHEAD_DS_2x }, INT_MIN, INT_MAX, VE, "look_ahead_downsampling" },
152 { "int_ref_type", "Intra refresh type", OFFSET(qsv.int_ref_type), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, UINT16_MAX, VE, "int_ref_type" },
153 { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, .flags = VE, "int_ref_type" },
154 { "vertical", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, .flags = VE, "int_ref_type" },
155 { "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 },
156 { "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 },
157 { "recovery_point_sei", "Insert recovery point SEI messages", OFFSET(qsv.recovery_point_sei), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 1, VE },
159 { "trellis", "Trellis quantization", OFFSET(qsv.trellis), AV_OPT_TYPE_FLAGS, { .i64 = 0 }, 0, UINT_MAX, VE, "trellis" },
160 { "off", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_TRELLIS_OFF }, .flags = VE, "trellis" },
161 { "I", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_TRELLIS_I }, .flags = VE, "trellis" },
162 { "P", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_TRELLIS_P }, .flags = VE, "trellis" },
163 { "B", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_TRELLIS_B }, .flags = VE, "trellis" },
165 { "profile", NULL, OFFSET(qsv.profile), AV_OPT_TYPE_INT, { .i64 = MFX_PROFILE_UNKNOWN }, 0, INT_MAX, VE, "profile" },
166 { "unknown" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_UNKNOWN }, INT_MIN, INT_MAX, VE, "profile" },
167 { "baseline", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_AVC_BASELINE }, INT_MIN, INT_MAX, VE, "profile" },
168 { "main" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_AVC_MAIN }, INT_MIN, INT_MAX, VE, "profile" },
169 { "high" , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MFX_PROFILE_AVC_HIGH }, INT_MIN, INT_MAX, VE, "profile" },
171 { "a53cc" , "Use A53 Closed Captions (if available)", OFFSET(qsv.a53_cc), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, VE},
175 static const AVClass class = {
176 .class_name = "h264_qsv encoder",
177 .item_name = av_default_item_name,
179 .version = LIBAVUTIL_VERSION_INT,
182 static const AVCodecDefault qsv_enc_defaults[] = {
185 // same as the x264 default
190 { "flags", "+cgop" },
191 #if FF_API_PRIVATE_OPT
192 { "b_strategy", "-1" },
197 AVCodec ff_h264_qsv_encoder = {
199 .long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (Intel Quick Sync Video acceleration)"),
200 .priv_data_size = sizeof(QSVH264EncContext),
201 .type = AVMEDIA_TYPE_VIDEO,
202 .id = AV_CODEC_ID_H264,
203 .init = qsv_enc_init,
204 .encode2 = qsv_enc_frame,
205 .close = qsv_enc_close,
206 .capabilities = AV_CODEC_CAP_DELAY,
207 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_NV12,
210 .priv_class = &class,
211 .defaults = qsv_enc_defaults,
212 .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,