Recherche avancée

Sur d’autres sites (4920)

• System.AccessViolationException on avcodec_open2 using FFMpeg.AutoGen

  5 décembre 2024, par forlayo

  I am using FFMpeg.AutoGen (v6.0.0.2) with hwencode support for Nvidia cards, and in few devices (for example on NVIDIA GeForce MX350) when I am doing fffmpeg.avcodec_open2 I got the following ffmpeg message in logs :

  


  [h264_nvenc @ 000001f473fcc480] OpenEncodeSessionEx failed: unsupported device (2): (no details)



  


  And then it crashes with the following crash :

  


  Fatal error. System.AccessViolationException: Attempted to read or write protected memory. This is often an indication that other memory is corrupt.&#xA;   at FFmpeg.AutoGen.DynamicallyLoadedBindings&#x2B;&lt;>c.<initialize>b__2_494(FFmpeg.AutoGen.AVCodecContext*, FFmpeg.AutoGen.AVCodec*, FFmpeg.AutoGen.AVDictionary**)&#xA;   at FFmpeg.AutoGen.ffmpeg.avcodec_open2(FFmpeg.AutoGen.AVCodecContext*, FFmpeg.AutoGen.AVCodec*, FFmpeg.AutoGen.AVDictionary**)&#xA;   at GraphicsCheck.FfmpegChecker.StartEncoder(Omni.Platforms.Windows.Services.AV.CodecInfo, Int32, Int32, Int32)&#xA;   at GraphicsCheck.FfmpegChecker.CheckConfig(CodecConfig)&#xA;   at Program.<main>$(System.String[])&#xA;</main></initialize>

  


  I created a sample app to demo this issue. And following the path of the calls..

  


  

  1. That happens at nvenc_open_session and is called here

  


  2. That would cause going to fail2 routine, which is this one :

  


  


  fail2:
    CHECK_CU(dl_fn->cuda_dl->cuCtxDestroy(ctx->cu_context_internal));
    ctx->cu_context_internal = NULL;


  


  

  3. ctx->cu_context_internal comes from (in same method) then I suppose it was created properly (as otherwise wouldn't continue).

  


  


      ret = CHECK_CU(dl_fn->cuda_dl->cuCtxCreate(&ctx->cu_context_internal, 0, cu_device));
    if (ret < 0)
        goto fail;


  


  

  4. ctx in that method is the priv_data of AVCodecContext as we can see here :

  


  


  static av_cold int nvenc_check_device(AVCodecContext *avctx, int idx)
{
    NvencContext *ctx = avctx->priv_data;


  


  

  5. Also the act of printing the error message, on nvenc_print_error, is interacting with context.

  


  


  static int nvenc_print_error(AVCodecContext *avctx, NVENCSTATUS err,
                             const char *error_string)
{
    const char *desc;
    const char *details = "(no details)";
    int ret = nvenc_map_error(err, &desc);

#ifdef NVENC_HAVE_GETLASTERRORSTRING
    NvencContext *ctx = avctx->priv_data;
    NV_ENCODE_API_FUNCTION_LIST *p_nvenc = &ctx->nvenc_dload_funcs.nvenc_funcs;

    if (p_nvenc && ctx->nvencoder)
        details = p_nvenc->nvEncGetLastErrorString(ctx->nvencoder);
#endif

    av_log(avctx, AV_LOG_ERROR, "%s: %s (%d): %s\n", error_string, desc, err, details);

    return ret;
}


  


  


  6. av_log is executed as we can see the error message on logs then the other lines shouldn't be a problem.

     


  


  7. Seeing the other calls, when nvenc_check_device fails the error is going up to nvenc_check_device -> nvenc_setup_device but we aren't seeing more logs then the issue may be on nvenc_check_device on fail2 routine or on nvenc_print_error

     


  


  


  Any thoughts ?

  


• FFMPEG extracting frames is slow even with CUDA

  5 juin 2023, par Ξένη Γήινος

  I have a bunch of videos at some location that I want to extract frames from.

  


  The videos are of .wmv format, and have a framerate of 25, and I want to extract all frames from them losslessly.

  


  And I do have a dedicated graphics card (NVIDIA Geforce GTX 1050 Ti, I know it's old), so I want to utilize it to accelerate frame extraction. I downloaded FFMPEG with CUDA support from www.gyan.dev for exactly this.

  


  I came up with this PowerShell command :

  


  gci C:\Users\Xeni\Documents -filter *.wmv | %{
 $folder = 'C:\Users\Xeni\Desktop\' + $($_.name -replace '.wmv')
 md $folder;
 D:\ffmpeg\ffmpeg.exe -hwaccel cuda -i $_.fullname -vf fps=25 "$folder\%05d.bmp"
}


  


  But somehow it is slow, its disk write speed is less than 100 MiB/s and it uses less than 1% GPU, and it takes about 2 seconds to process 25 frames.

  


  How can I speed up this command ?

  


• vulkan_decode : do not align the image dimensions

  29 mai 2023, par Lynne

  vulkan_decode : do not align the image dimensions
  According to Dave Airlie :
  > <airlied> but I think ignoring it should be fine, I can't see any
  
  > other way to get the imaeg extents correct for other usage
  
  > <Lynne> what width/height should be used for the images ?
  
  > the final presentable dimensions, or the coded dimensions ?
  
  > <airlied> if you don't want noise I think the presentable dims
  
  > <airlied> the driver should round up the allocations internally,
  
  > but if you are going to sample from the images then w/h have to be
  
  > the bounds of the image you want
  
  > <airlied> since otherwise there's no way to stop the sampler from
  
  > going outside the edges
  Apparently, the alignment values are informative, rather than mandatory,
  
  but the spec's wording makes it sound as if they're mandatory.
  

  • [DH] libavcodec/vulkan_decode.c