I’m working on a video transcoding project using FFmpeg with VAAPI hardware acceleration, and I’m encountering an issue where the gop_size parameter is not being respected during encoding. Despite setting the gop_size to 150 in the encoder context, the output video does not have the expected GOP structure. This problem only occurs when using VAAPI for hardware acceleration—when I switch to software decoding or CUDA, the GOP size is applied correctly.
Here’s a simplified version of the relevant code where I set the gop_size:
encoder_ctx->gop_size = 150;
I’m using FFmpeg’s VAAPI for hardware-accelerated encoding, and everything else works as expected, except for this GOP size issue. Has anyone encountered this problem, or is there something specific to VAAPI that I’m missing?
Environment:
- FFmpeg version: (6.0)
- VAAPI (Intel GPU)
- Encoder: H.264
#include <stdio.h>
#include <errno.h>
extern "C"
{
#include <libavutil/hwcontext.h>
#include <libavcodec/avcodec.h>
#include <libavformat/avformat.h>
}
static AVFormatContext* ifmt_ctx = NULL, * ofmt_ctx = NULL;
static AVBufferRef* hw_device_ctx = NULL;
static AVCodecContext* decoder_ctx = NULL, * encoder_ctx = NULL;
static int video_stream = -1;
static AVStream* ost;
static int initialized = 0;
static enum AVPixelFormat get_vaapi_format(AVCodecContext* ctx,
const enum AVPixelFormat* pix_fmts)
{
const enum AVPixelFormat* p;
for (p = pix_fmts; *p != AV_PIX_FMT_NONE; p++) {
if (*p == AV_PIX_FMT_VAAPI)
return *p;
}
fprintf(stderr, "Unable to decode this file using VA-API.\n");
return AV_PIX_FMT_NONE;
}
static int open_input_file(const char* filename)
{
int ret;
const AVCodec* decoder = NULL;
AVStream* video = NULL;
char err_buf[AV_ERROR_MAX_STRING_SIZE] = { 0, };
if ((ret = avformat_open_input(&ifmt_ctx, filename, NULL, NULL)) < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Cannot open input file '%s', Error code: %s\n", filename, err_buf);
return ret;
}
if ((ret = avformat_find_stream_info(ifmt_ctx, NULL)) < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Cannot find input stream information. Error code: %s\n", err_buf);
return ret;
}
ret = av_find_best_stream(ifmt_ctx, AVMEDIA_TYPE_VIDEO, -1, -1, &decoder, 0);
if (ret < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Cannot find a video stream in the input file. Error code: %s\n", err_buf);
return ret;
}
video_stream = ret;
if (!(decoder_ctx = avcodec_alloc_context3(decoder)))
return AVERROR(ENOMEM);
video = ifmt_ctx->streams[video_stream];
if ((ret = avcodec_parameters_to_context(decoder_ctx, video->codecpar)) < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "avcodec_parameters_to_context error. Error code: %s\n", err_buf);
return ret;
}
decoder_ctx->hw_device_ctx = av_buffer_ref(hw_device_ctx);
if (!decoder_ctx->hw_device_ctx) {
fprintf(stderr, "A hardware device reference create failed.\n");
return AVERROR(ENOMEM);
}
decoder_ctx->get_format = get_vaapi_format;
if ((ret = avcodec_open2(decoder_ctx, decoder, NULL)) < 0){
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Failed to open codec for decoding.Error code: %s\n", err_buf);
}
return ret;
}
static int encode_write(AVPacket* enc_pkt, AVFrame* frame)
{
int ret = 0;
char err_buf[AV_ERROR_MAX_STRING_SIZE] = { 0, };
av_packet_unref(enc_pkt);
if ((ret = avcodec_send_frame(encoder_ctx, frame)) < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Error during encoding. Error code: %s\n", err_buf);
goto end;
}
while (1) {
ret = avcodec_receive_packet(encoder_ctx, enc_pkt);
if (ret)
break;
enc_pkt->stream_index = 0;
av_packet_rescale_ts(enc_pkt, ifmt_ctx->streams[video_stream]->time_base,
ofmt_ctx->streams[0]->time_base);
ret = av_interleaved_write_frame(ofmt_ctx, enc_pkt);
if (ret < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Error during writing data to output file. Error code: %s\n", err_buf);
return -1;
}
}
end:
if (ret == AVERROR_EOF)
return 0;
ret = ((ret == AVERROR(EAGAIN)) ? 0 : -1);
return ret;
}
static int dec_enc(AVPacket* pkt, const AVCodec* enc_codec)
{
AVFrame* frame;
AVDictionary* opts = NULL;
int ret = 0;
char err_buf[AV_ERROR_MAX_STRING_SIZE] = { 0, };
ret = avcodec_send_packet(decoder_ctx, pkt);
if (ret < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Error during decoding.Error code : %s\n", err_buf);
return ret;
}
while (ret >= 0) {
if (!(frame = av_frame_alloc()))
return AVERROR(ENOMEM);
ret = avcodec_receive_frame(decoder_ctx, frame);
if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF) {
av_frame_free(&frame);
return 0;
}
else if (ret < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Error while decoding.Error code : %s\n", err_buf);
goto fail;
}
if (!initialized) {
/* we need to ref hw_frames_ctx of decoder to initialize encoder's codec.
Only after we get a decoded frame, can we obtain its hw_frames_ctx */
encoder_ctx->hw_frames_ctx = av_buffer_ref(decoder_ctx->hw_frames_ctx);
if (!encoder_ctx->hw_frames_ctx) {
ret = AVERROR(ENOMEM);
goto fail;
}
/* set AVCodecContext Parameters for encoder, here we keep them stay
* the same as decoder.
* xxx: now the sample can't handle resolution change case.
*/
encoder_ctx->time_base = av_inv_q(decoder_ctx->framerate);
encoder_ctx->pix_fmt = AV_PIX_FMT_VAAPI;
encoder_ctx->width = decoder_ctx->width;
encoder_ctx->height = decoder_ctx->height;
encoder_ctx->gop_size = 150;
av_dict_set(&opts, "g", "150", 0); // gop_size 설정
if ((ret = avcodec_open2(encoder_ctx, enc_codec, &opts)) < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Failed to open encode codec. Error code : %s\n", err_buf);
goto fail;
}
if (!(ost = avformat_new_stream(ofmt_ctx, enc_codec))) {
fprintf(stderr, "Failed to allocate stream for output format.\n");
ret = AVERROR(ENOMEM);
goto fail;
}
ost->time_base = encoder_ctx->time_base;
ret = avcodec_parameters_from_context(ost->codecpar, encoder_ctx);
if (ret < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Failed to copy the stream parameters. Error code : %s\n", err_buf);
goto fail;
}
/* write the stream header */
if ((ret = avformat_write_header(ofmt_ctx, NULL)) < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Error while writing stream header. Error code : %s\n", err_buf);
goto fail;
}
initialized = 1;
}
if ((ret = encode_write(pkt, frame)) < 0)
fprintf(stderr, "Error during encoding and writing.\n");
fail:
av_frame_free(&frame);
if (ret < 0)
return ret;
}
return 0;
}
int main(int argc, char** argv)
{
const AVCodec* enc_codec;
int ret = 0;
AVPacket* dec_pkt;
char err_buf[AV_ERROR_MAX_STRING_SIZE] = { 0, };
if (argc != 4) {
fprintf(stderr, "Usage: %s <input file> <encode codec> <output file>\n"
"The output format is guessed according to the file extension.\n"
"\n", argv[0]);
return -1;
}
ret = av_hwdevice_ctx_create(&hw_device_ctx, AV_HWDEVICE_TYPE_VAAPI, NULL, NULL, 0);
if (ret < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Failed to create a VAAPI device. Error code : %s\n", err_buf);
return -1;
}
dec_pkt = av_packet_alloc();
if (!dec_pkt) {
fprintf(stderr, "Failed to allocate decode packet\n");
goto end;
}
if ((ret = open_input_file(argv[1])) < 0)
goto end;
if (!(enc_codec = avcodec_find_encoder_by_name(argv[2]))) {
fprintf(stderr, "Could not find encoder '%s'\n", argv[2]);
ret = -1;
goto end;
}
if ((ret = (avformat_alloc_output_context2(&ofmt_ctx, NULL, NULL, argv[3]))) < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Failed to deduce output format from file extension. Error code : %s\n", err_buf);
goto end;
}
if (!(encoder_ctx = avcodec_alloc_context3(enc_codec))) {
ret = AVERROR(ENOMEM);
goto end;
}
ret = avio_open(&ofmt_ctx->pb, argv[3], AVIO_FLAG_WRITE);
if (ret < 0) {
av_strerror(ret, err_buf, AV_ERROR_MAX_STRING_SIZE);
fprintf(stderr, "Cannot open output file. Error code : %s\n", err_buf);
goto end;
}
/* read all packets and only transcoding video */
while (ret >= 0) {
if ((ret = av_read_frame(ifmt_ctx, dec_pkt)) < 0)
break;
if (video_stream == dec_pkt->stream_index)
ret = dec_enc(dec_pkt, enc_codec);
av_packet_unref(dec_pkt);
}
/* flush decoder */
av_packet_unref(dec_pkt);
ret = dec_enc(dec_pkt, enc_codec);
/* flush encoder */
ret = encode_write(dec_pkt, NULL);
/* write the trailer for output stream */
av_write_trailer(ofmt_ctx);
end:
avformat_close_input(&ifmt_ctx);
avformat_close_input(&ofmt_ctx);
avcodec_free_context(&decoder_ctx);
avcodec_free_context(&encoder_ctx);
av_buffer_unref(&hw_device_ctx);
av_packet_free(&dec_pkt);
return ret;
}
You need to sign in to view this answers
Leave feedback about this