Recherche avancée

Sur d’autres sites (8437)

• lavf/audiotoolboxdec : only provide block alignment for ILBC

  8 avril 2016, par Rodger Combs

  lavf/audiotoolboxdec : only provide block alignment for ILBC
  Fixes decode errors for some AVI files
  

  • [DH] libavcodec/audiotoolboxdec.c

• cuda invalid resource handle when using h264_cuvid decoder in my C program [closed]

  29 juillet 2021, par ChrisFisher

  I try to use GPU acceleration in my ffmpeg decode program, and I check the codecs of ffmpeg by command :ffmpeg -codecs | grep nv, and it shows that I can use h264_cuvid decoder(In fact, I have already used ffmpeg command line to encode and decode a test video with hardware acceleration and it turned out everything was all fine), but when I use the decoder in my program

  


  AVCodec *pCodec = avcodec_find_decoder_by_name("h264_cuvid");

  


  here is part of my program

  


  void FFMPEGCodec::initDecoder()
{
    AVCodec *pCodec = avcodec_find_decoder(AV_CODEC_ID_H264);
    if (!pCodec) {
        LOG("Codec decoder not found\n");
        exit(1);
    }

    pCodecCtx = avcodec_alloc_context3(pCodec);
    if (!pCodecCtx) {
        LOG("Could not allocate video codec context\n");
        exit(1);
    }

    pCodecCtx->width = gConfig->totalWidth;
    pCodecCtx->height = gConfig->totalHeight;
    pCodecCtx->has_b_frames = 0;

    if (avcodec_open2(pCodecCtx, pCodec, NULL) < 0) {
        LOG("Could not open codec\n");
        exit(1);
    }

    if(_x264rgb){
        //used to convert GBRP frame to RGB image.
        convertCtx = sws_getContext(gConfig->totalWidth, gConfig->totalHeight, AV_PIX_FMT_GBRP, 
        gConfig->totalWidth, gConfig->totalHeight, AV_PIX_FMT_RGB24, SWS_FAST_BILINEAR, NULL, NULL, NULL); 
    } else {
        //used to convert YUV frame to RGB image.
        convertCtx = sws_getContext(gConfig->totalWidth, gConfig->totalHeight, AV_PIX_FMT_YUV420P, 
        gConfig->totalWidth, gConfig->totalHeight, AV_PIX_FMT_RGB24, SWS_FAST_BILINEAR, NULL, NULL, NULL); 
    }

    if(convertCtx == NULL){
        LOG("Failed to get SwsContext\n");
        exit(1);
    }

    //when using x264rgb, it's actually GBRP frame, 
    //just don't want to define another variable
    yuvFrame = av_frame_alloc();
    if (!yuvFrame) {
        LOG("Failed to allocate yuv frame\n");
        exit(1);
    }

    rgbFrame = av_frame_alloc();
    if (!rgbFrame) {
        LOG("Failed to allocate rgb frame\n");
        exit(1);
    }

    rgbFrame->format = AV_PIX_FMT_RGB24;
    rgbFrame->width  = pCodecCtx->width;
    rgbFrame->height = pCodecCtx->height;
 
    int ret = av_image_alloc(rgbFrame->data, rgbFrame->linesize, rgbFrame->width, rgbFrame->height,
                         AV_PIX_FMT_RGB24, 32);
    if (ret < 0) {
        LOG("Failed to allocate raw picture buffer\n");
        exit(1);
    }
}   


  


  and

  


  int FFMPEGCodec::decode(byte* pktData, int pktSize, byte* imgData)
{
    int ret = 0, got_packet = 0;
    AVPacket pkt;
    av_init_packet(&pkt);
    pkt.data = pktData;
    pkt.size = pktSize;

    // decode video frame
    ret = avcodec_decode_video2(pCodecCtx, yuvFrame, &got_packet, &pkt);
    if (ret < 0) {
        LOG("Error decoding frame\n");
        return -1;
    }

    sws_scale(convertCtx, yuvFrame->data, yuvFrame->linesize, 0, pCodecCtx->height, rgbFrame->data, rgbFrame->linesize);
    
    if (got_packet) {
        int width = pCodecCtx->width, height = pCodecCtx->height;
        int fsize = rgbFrame->linesize[0] * rgbFrame->height;
        int size = 54 + fsize;

        byte bmp_file_header[14] = { 'B', 'M', 0, 0, 0, 0, 0, 0, 0, 0, 54, 0, 0, 0, };
        byte bmp_info_header[40] = { 40, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 24, 0, };
        byte bmp_pad[3] = { 0, 0, 0 };

        bmp_file_header[2] = (unsigned char)size;
        bmp_file_header[3] = (unsigned char)(size >> 8);
        bmp_file_header[4] = (unsigned char)(size >> 16);
        bmp_file_header[5] = (unsigned char)(size >> 24);

        bmp_info_header[4] = (unsigned char)(width);
        bmp_info_header[5] = (unsigned char)(width >> 8);
        bmp_info_header[6] = (unsigned char)(width >> 16);
        bmp_info_header[7] = (unsigned char)(width >> 24);
        bmp_info_header[8] = (unsigned char)(height);
        bmp_info_header[9] = (unsigned char)(height >> 8);
        bmp_info_header[10] = (unsigned char)(height >> 16);
        bmp_info_header[11] = (unsigned char)(height >> 24);

        memcpy(imgData, bmp_file_header, 14);
        memcpy(imgData + 14, bmp_info_header, 40);
        memcpy(imgData + 54, rgbFrame->data[0], fsize);
        ret = size;
    }
    av_free_packet(&pkt);

    return ret;
}



  


  after compiling, I ran the program, and the decoder throw me a error :

  


  ctx->cvdl->cuvidDecodePicture(ctx->cudecoder, picparams) failed -> CUDA_ERROR_INVALID_HANDLE: invalid resource handle
when calling the function avcodec_decode_video2

  


  I don't know why this error occurred, by the way, I use a GTX1060 6G(Sorry I'm not a native English speaker)

  


• tests/checkasm/diracdsp : fix alignment for src and ombc_weight buffers

  19 novembre 2024, par James Almer

  tests/checkasm/diracdsp : fix alignment for src and ombc_weight buffers
  They are supposed to be 16 byte aligned, not 8.
  
  Should fix crashes in some systems.
  Signed-off-by : James Almer <jamrial@gmail.com>
  

  • [DH] tests/checkasm/diracdsp.c