Recherche avancée

Sur d’autres sites (9321)

• How get video output on hardware acceleration with VAAPI on FFMPEG ?

  14 mai 2019, par HellBlood

  For my project I have to use hardware acceleration with VAAPI using FFMPEG, the problem is that with VAAPI we have to use the "vaapi_retrieve_data" function to download the GPU data. This is time consuming and I would like to be able to directly view the video via the GPU. The program is in C language and aims to decode videos. Do you have a solution to recover the output data without going through the function ?
  Thank you in advance :)
  Have a good day

• How to use AMD hardware acceleration VAAPI when using ffmpeg in qtcreator ? [closed]

  28 décembre 2024, par Ling Yun

  ffmpeg version:4.2.2
env:QT creator

  


  I passed this command in terminal :

  


  ffmpeg -hwaccel vaapi -vaapi_device /dev/dri/renderD128 -i input.mp4 -vf format=yuv420p,hwupload -c:v h264_vaapi -b:v 1000k output.mp4


  


  This can be accelerated, the CPU usage is very low
But when I want to use code to implement hardware accelerated encoding, it prompts :

  


  cannot allocate memory(-12)


  


  my c++ code :

  


  av_hwdevice_ctx_create = (&hw_device_ctx,AV_HWDEVICE_TYPE_VAAPI,"/dev/dri/renderD128");


  


  print :-12

  


  I can be sure that I compiled ffmpeg correctly because I can enable vaapi hardware acceleration through the command and my graphics card has enough video memory.

  


  Possible causes :
I introduced -lavcodec -lavformat -lavutil -lswscale in my qt pro file libs, but did not add -lva -lva-drm

  


  Is it correct to add -lva -lva-drm after libs ?
What should I do ?

  


• Libavcodec (FFMPEG) D3D11 hardware acceleration not outputting hardware frames

  12 juin 2024, par luckybroma

  The Problem

  


  I'm trying to get FFMPEG to output D3D11 2D textures directly as Frames. As I understand, this is infact possible. The reasoning for this, is I can then decode the YUV format directly on the GPU and then display it without having to Marshall it back to the CPU.

  


  My problem is that I can't get the decoder to actually output frames in the desired format, they are always software YUV420J.

  


  Is this because I'm not initializing the decoder properly ? Or perhaps there is some other configuration I'm missing.

  


  ---

  


  The Details

  


  I've been following the example for the most part, with a few minor differences :

  


  

  1. I only have to deal with 1 format, raw h264, so I'm manually configuring the Decoder to use AV_CODEC_ID_H264

  


  2. I've all ready initialized the D3DDevice and Context and so am using av_hwdevice_ctx_init over av_hwdevice_ctx_create. (I have tried just using av_hwdevice_ctx_create and got the same results)

  


  3. As mentioned earlier, I don't want to use av_hwframe_transfer_data because I want to keep the data on the GPU as a texture and immediately display it.

  


  


  Here is my initialization code :

  


      AVHWDeviceType type = AV_HWDEVICE_TYPE_D3D11VA;

    hw_pix_fmt = AV_PIX_FMT_D3D11;

    _codec = avcodec_find_decoder(AV_CODEC_ID_H264); // NOT SURE IF THIS IS THE HW SUPPORTED FMT
    
    int i;
    for (i = 0;; i++) {
        const AVCodecHWConfig* config = avcodec_get_hw_config(_codec, i);
        if (!config) {
            fprintf(stderr, "Decoder %s does not support device type %s.\n",
                _codec->name, av_hwdevice_get_type_name(type));
            return;
        }
        if (config->methods & AV_CODEC_HW_CONFIG_METHOD_HW_DEVICE_CTX &&
            config->device_type == type) {
            hw_pix_fmt = config->pix_fmt;
            break;
        }
    }

    if (!(_codecCtx = avcodec_alloc_context3(_codec))) {
        // ERROR!
        return;
    }

    _hw_device_ctx = av_hwdevice_ctx_alloc(type);

    _device_ctx = reinterpret_cast(_hw_device_ctx->data);
    _d3d11va_device_ctx = reinterpret_cast(_device_ctx->hwctx);
    _d3d11va_device_ctx->device = dev;

    _codecCtx->hw_device_ctx = av_buffer_ref(_hw_device_ctx);


    if (av_hwdevice_ctx_init(_codecCtx->hw_device_ctx) < 0) {
        return;
    }


    if (avcodec_open2(_codecCtx, _codec, NULL) < 0) {
        return;
    }

    _parser = av_parser_init(_codec->id);
    _packet = av_packet_alloc();


  


  I'm doing the standard things to decode, and know it works because I've used software decoding just fine, but here it is anyway :

  


  void Decoder::Load(std::vector data) {
    int data_size = data.size();

    int bytesRead = 0;
    int bytesParsed = 0;

    while (data_size > 0) {
        bytesParsed = av_parser_parse2(
            _parser,
            _codecCtx,
            &_packet->data,
            &_packet->size,
            (uint8_t*)&data[bytesRead],
            data.size() - bytesRead,
            AV_NOPTS_VALUE,
            AV_NOPTS_VALUE,
            0);

        bytesRead += bytesParsed;
        data_size -= bytesParsed;

        if (bytesParsed < 0) {
            //std::cout << "Error while parsing packet" << std::endl;
            return;
        }

        if (_packet->size) {
            int ret = avcodec_send_packet(_codecCtx, _packet);
            if (ret < 0) {
                //std::cout << "Error Sending Packet for Decoding" << std::endl;
                return;
            }

            while (ret >= 0) {
                AVFrame* frame = av_frame_alloc();
                ret = avcodec_receive_frame(_codecCtx, frame);

                if (frame->format != AV_PIX_FMT_D3D11) { // <--- THIS NEVER EVALUATES AS FALSE, IT'S ALWAYS THE YUV420J FORMAT, NOT THE HARDWARE FORMAT
                    break;
                }

                if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF) {
                    break;
                }
                else if (ret < 0) {
                    //std::cout << "Error During Decoding" << std::endl;
                    return;
                }

                _frameQueue.push(frame);
            }
        }
    }
}


  


  ---

  


  How I built FFMPEG :

  


  ./../../sources/ffmpeg/configure \
             --prefix=./../../installed \
             --toolchain=msvc \
             --arch=x86_64 \
             --enable-yasm  \
             --enable-asm \
             --enable-shared \
             --enable-dxva2 \
             --enable-d3d11va \
             --disable-static \
             --enable-gpl \
             --extra-ldflags="-LIBPATH:./../../installed/lib/" \
             --extra-cflags="-I./../../installed/include/"