[FFmpeg-devel] nvenc/cuda force IDR

Grant Brownewell gbrownewell at InXpo.com
Wed Oct 12 20:45:20 EEST 2016

>> Hello,
>> I've been experimenting with encoding with a  cuda card, and I noticed that setting the pict_type member of the AVFrame structure to AV_PICTURE_TYPE_I >does not trigger it to encode the next frame as an IDR frame as libx264 does. I looked at the nvidia docs and it appears there is a mechanism for this behavior.

>Is that behaviour documented anywhere?

I found documentation on the nvidia site (https://developer.nvidia.com/nvencode), here is the section of the document that I referred to in my previous post...

4.2.3 Forcing current frame to be used as an IDR frame 
To force the current frame to be encoded as IDR frame, set  

>There should not be any technical issue with triggering a forced iframe, the encoder might just choose to ignore it.

Looking at the latest nvenc.c and the function ff_nvenc_encode_frame, I see that the encodePicFlags member is just always being set to zero...

        pic_params.encodePicFlags = 0;

...before being passed along in this line...

nv_status = p_nvenc->nvEncEncodePicture(ctx->nvencoder, &pic_params);

I have not contributed any patches to ffmpeg before, but this seems like a simple and small patch to look at the incoming AVFrame's pict_type member and when it's value is AV_PICTURE_TYPE_I, set the corresponding NV_ENC_PIC_PARAMS::encodePicFlags  member to NV_ENC_PIC_FLAG_FORCEIDR.

Would this be a welcome patch, or would I just be duplicating effort because there is already someone maintaining this code? It seems to me there is quite a bit of activity with nvenc.c, it works well and I don't want to interfere. Also, my situation is somewhat narrow and perhaps a patch should be more complete than what I describe?

I appreciate the guidance.

More information about the ffmpeg-devel mailing list