1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-07 11:13:41 +02:00
FFmpeg/libavcodec/vaapi_encode_vp9.c
Philip Langdale 109515e16d lavc/vaapi_encode: enable 8bit 4:4:4 encoding for HEVC and VP9
Sufficiently recent Intel hardware is able to do encoding of 8bit 4:4:4
content in HEVC and VP9. The main requirement here is that the frames
must be provided in the AYUV format.

Enabling support is done by adding the appropriate encoding profiles
and noting that AYUV is officially a four channel format with alpha so
we must state that we expect all four channels.
2022-08-09 09:22:49 -07:00

322 lines
11 KiB
C

/*
* 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 <va/va.h>
#include <va/va_enc_vp9.h>
#include "libavutil/avassert.h"
#include "libavutil/common.h"
#include "libavutil/internal.h"
#include "libavutil/opt.h"
#include "libavutil/pixfmt.h"
#include "avcodec.h"
#include "codec_internal.h"
#include "vaapi_encode.h"
#define VP9_MAX_QUANT 255
#define VP9_MAX_TILE_WIDTH 4096
typedef struct VAAPIEncodeVP9Picture {
int slot;
} VAAPIEncodeVP9Picture;
typedef struct VAAPIEncodeVP9Context {
VAAPIEncodeContext common;
// User options.
int loop_filter_level;
int loop_filter_sharpness;
// Derived settings.
int q_idx_idr;
int q_idx_p;
int q_idx_b;
} VAAPIEncodeVP9Context;
static int vaapi_encode_vp9_init_sequence_params(AVCodecContext *avctx)
{
VAAPIEncodeContext *ctx = avctx->priv_data;
VAEncSequenceParameterBufferVP9 *vseq = ctx->codec_sequence_params;
VAEncPictureParameterBufferVP9 *vpic = ctx->codec_picture_params;
vseq->max_frame_width = avctx->width;
vseq->max_frame_height = avctx->height;
vseq->kf_auto = 0;
if (!(ctx->va_rc_mode & VA_RC_CQP)) {
vseq->bits_per_second = ctx->va_bit_rate;
vseq->intra_period = ctx->gop_size;
}
vpic->frame_width_src = avctx->width;
vpic->frame_height_src = avctx->height;
vpic->frame_width_dst = avctx->width;
vpic->frame_height_dst = avctx->height;
return 0;
}
static int vaapi_encode_vp9_init_picture_params(AVCodecContext *avctx,
VAAPIEncodePicture *pic)
{
VAAPIEncodeContext *ctx = avctx->priv_data;
VAAPIEncodeVP9Context *priv = avctx->priv_data;
VAAPIEncodeVP9Picture *hpic = pic->priv_data;
VAEncPictureParameterBufferVP9 *vpic = pic->codec_picture_params;
int i;
int num_tile_columns;
vpic->reconstructed_frame = pic->recon_surface;
vpic->coded_buf = pic->output_buffer;
// Maximum width of a tile in units of superblocks is MAX_TILE_WIDTH_B64(64)
// So the number of tile columns is related to the width of the picture.
// We set the minimum possible number for num_tile_columns as default value.
num_tile_columns = (vpic->frame_width_src + VP9_MAX_TILE_WIDTH - 1) / VP9_MAX_TILE_WIDTH;
vpic->log2_tile_columns = num_tile_columns == 1 ? 0 : av_log2(num_tile_columns - 1) + 1;
switch (pic->type) {
case PICTURE_TYPE_IDR:
av_assert0(pic->nb_refs == 0);
vpic->ref_flags.bits.force_kf = 1;
vpic->refresh_frame_flags = 0xff;
hpic->slot = 0;
break;
case PICTURE_TYPE_P:
av_assert0(pic->nb_refs == 1);
{
VAAPIEncodeVP9Picture *href = pic->refs[0]->priv_data;
av_assert0(href->slot == 0 || href->slot == 1);
if (ctx->max_b_depth > 0) {
hpic->slot = !href->slot;
vpic->refresh_frame_flags = 1 << hpic->slot | 0xfc;
} else {
hpic->slot = 0;
vpic->refresh_frame_flags = 0xff;
}
vpic->ref_flags.bits.ref_frame_ctrl_l0 = 1;
vpic->ref_flags.bits.ref_last_idx = href->slot;
vpic->ref_flags.bits.ref_last_sign_bias = 1;
}
break;
case PICTURE_TYPE_B:
av_assert0(pic->nb_refs == 2);
{
VAAPIEncodeVP9Picture *href0 = pic->refs[0]->priv_data,
*href1 = pic->refs[1]->priv_data;
av_assert0(href0->slot < pic->b_depth + 1 &&
href1->slot < pic->b_depth + 1);
if (pic->b_depth == ctx->max_b_depth) {
// Unreferenced frame.
vpic->refresh_frame_flags = 0x00;
hpic->slot = 8;
} else {
vpic->refresh_frame_flags = 0xfe << pic->b_depth & 0xff;
hpic->slot = 1 + pic->b_depth;
}
vpic->ref_flags.bits.ref_frame_ctrl_l0 = 1;
vpic->ref_flags.bits.ref_frame_ctrl_l1 = 2;
vpic->ref_flags.bits.ref_last_idx = href0->slot;
vpic->ref_flags.bits.ref_last_sign_bias = 1;
vpic->ref_flags.bits.ref_gf_idx = href1->slot;
vpic->ref_flags.bits.ref_gf_sign_bias = 0;
}
break;
default:
av_assert0(0 && "invalid picture type");
}
if (vpic->refresh_frame_flags == 0x00) {
av_log(avctx, AV_LOG_DEBUG, "Pic %"PRId64" not stored.\n",
pic->display_order);
} else {
av_log(avctx, AV_LOG_DEBUG, "Pic %"PRId64" stored in slot %d.\n",
pic->display_order, hpic->slot);
}
for (i = 0; i < FF_ARRAY_ELEMS(vpic->reference_frames); i++)
vpic->reference_frames[i] = VA_INVALID_SURFACE;
for (i = 0; i < pic->nb_refs; i++) {
VAAPIEncodePicture *ref_pic = pic->refs[i];
int slot;
slot = ((VAAPIEncodeVP9Picture*)ref_pic->priv_data)->slot;
av_assert0(vpic->reference_frames[slot] == VA_INVALID_SURFACE);
vpic->reference_frames[slot] = ref_pic->recon_surface;
}
vpic->pic_flags.bits.frame_type = (pic->type != PICTURE_TYPE_IDR);
vpic->pic_flags.bits.show_frame = pic->display_order <= pic->encode_order;
if (pic->type == PICTURE_TYPE_IDR)
vpic->luma_ac_qindex = priv->q_idx_idr;
else if (pic->type == PICTURE_TYPE_P)
vpic->luma_ac_qindex = priv->q_idx_p;
else
vpic->luma_ac_qindex = priv->q_idx_b;
vpic->luma_dc_qindex_delta = 0;
vpic->chroma_ac_qindex_delta = 0;
vpic->chroma_dc_qindex_delta = 0;
vpic->filter_level = priv->loop_filter_level;
vpic->sharpness_level = priv->loop_filter_sharpness;
return 0;
}
static av_cold int vaapi_encode_vp9_get_encoder_caps(AVCodecContext *avctx)
{
VAAPIEncodeContext *ctx = avctx->priv_data;
// Surfaces must be aligned to 64x64 superblock boundaries.
ctx->surface_width = FFALIGN(avctx->width, 64);
ctx->surface_height = FFALIGN(avctx->height, 64);
return 0;
}
static av_cold int vaapi_encode_vp9_configure(AVCodecContext *avctx)
{
VAAPIEncodeContext *ctx = avctx->priv_data;
VAAPIEncodeVP9Context *priv = avctx->priv_data;
if (ctx->rc_mode->quality) {
priv->q_idx_p = av_clip(ctx->rc_quality, 0, VP9_MAX_QUANT);
if (avctx->i_quant_factor > 0.0)
priv->q_idx_idr =
av_clip((avctx->i_quant_factor * priv->q_idx_p +
avctx->i_quant_offset) + 0.5,
0, VP9_MAX_QUANT);
else
priv->q_idx_idr = priv->q_idx_p;
if (avctx->b_quant_factor > 0.0)
priv->q_idx_b =
av_clip((avctx->b_quant_factor * priv->q_idx_p +
avctx->b_quant_offset) + 0.5,
0, VP9_MAX_QUANT);
else
priv->q_idx_b = priv->q_idx_p;
} else {
// Arbitrary value.
priv->q_idx_idr = priv->q_idx_p = priv->q_idx_b = 100;
}
ctx->roi_quant_range = VP9_MAX_QUANT;
return 0;
}
static const VAAPIEncodeProfile vaapi_encode_vp9_profiles[] = {
{ FF_PROFILE_VP9_0, 8, 3, 1, 1, VAProfileVP9Profile0 },
// Four channels because this uses the AYUV format which has Alpha
{ FF_PROFILE_VP9_1, 8, 4, 0, 0, VAProfileVP9Profile1 },
{ FF_PROFILE_VP9_2, 10, 3, 1, 1, VAProfileVP9Profile2 },
{ FF_PROFILE_UNKNOWN }
};
static const VAAPIEncodeType vaapi_encode_type_vp9 = {
.profiles = vaapi_encode_vp9_profiles,
.flags = FLAG_B_PICTURES |
FLAG_B_PICTURE_REFERENCES,
.default_quality = 100,
.picture_priv_data_size = sizeof(VAAPIEncodeVP9Picture),
.get_encoder_caps = &vaapi_encode_vp9_get_encoder_caps,
.configure = &vaapi_encode_vp9_configure,
.sequence_params_size = sizeof(VAEncSequenceParameterBufferVP9),
.init_sequence_params = &vaapi_encode_vp9_init_sequence_params,
.picture_params_size = sizeof(VAEncPictureParameterBufferVP9),
.init_picture_params = &vaapi_encode_vp9_init_picture_params,
};
static av_cold int vaapi_encode_vp9_init(AVCodecContext *avctx)
{
VAAPIEncodeContext *ctx = avctx->priv_data;
ctx->codec = &vaapi_encode_type_vp9;
// No packed headers are currently desired. They could be written,
// but there isn't any reason to do so - the one usable driver (i965)
// can write its own headers and there is no metadata to include.
ctx->desired_packed_headers = 0;
return ff_vaapi_encode_init(avctx);
}
#define OFFSET(x) offsetof(VAAPIEncodeVP9Context, x)
#define FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
static const AVOption vaapi_encode_vp9_options[] = {
VAAPI_ENCODE_COMMON_OPTIONS,
VAAPI_ENCODE_RC_OPTIONS,
{ "loop_filter_level", "Loop filter level",
OFFSET(loop_filter_level), AV_OPT_TYPE_INT, { .i64 = 16 }, 0, 63, FLAGS },
{ "loop_filter_sharpness", "Loop filter sharpness",
OFFSET(loop_filter_sharpness), AV_OPT_TYPE_INT, { .i64 = 4 }, 0, 15, FLAGS },
{ NULL },
};
static const FFCodecDefault vaapi_encode_vp9_defaults[] = {
{ "b", "0" },
{ "bf", "0" },
{ "g", "250" },
{ "qmin", "-1" },
{ "qmax", "-1" },
{ NULL },
};
static const AVClass vaapi_encode_vp9_class = {
.class_name = "vp9_vaapi",
.item_name = av_default_item_name,
.option = vaapi_encode_vp9_options,
.version = LIBAVUTIL_VERSION_INT,
};
const FFCodec ff_vp9_vaapi_encoder = {
.p.name = "vp9_vaapi",
.p.long_name = NULL_IF_CONFIG_SMALL("VP9 (VAAPI)"),
.p.type = AVMEDIA_TYPE_VIDEO,
.p.id = AV_CODEC_ID_VP9,
.priv_data_size = sizeof(VAAPIEncodeVP9Context),
.init = &vaapi_encode_vp9_init,
FF_CODEC_RECEIVE_PACKET_CB(&ff_vaapi_encode_receive_packet),
.close = &ff_vaapi_encode_close,
.p.priv_class = &vaapi_encode_vp9_class,
.p.capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_HARDWARE |
AV_CODEC_CAP_DR1,
.caps_internal = FF_CODEC_CAP_NOT_INIT_THREADSAFE |
FF_CODEC_CAP_INIT_CLEANUP,
.defaults = vaapi_encode_vp9_defaults,
.p.pix_fmts = (const enum AVPixelFormat[]) {
AV_PIX_FMT_VAAPI,
AV_PIX_FMT_NONE,
},
.hw_configs = ff_vaapi_encode_hw_configs,
.p.wrapper_name = "vaapi",
};