Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-28 Thread Timo Rothenpieler
applied ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".

Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-20 Thread Timo Rothenpieler
On 20.03.2020 12:30, Yaroslav Pogrebnyak wrote: On 20.03.20 00:47, Timo Rothenpieler wrote: I'm looking into adding hardware-frame support to make_writable, so modifications might not be needed. Yep it seems to be more consistent if av_frame_make_writable could support hardware frames.

Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-20 Thread Yaroslav Pogrebnyak
On 20.03.20 00:47, Timo Rothenpieler wrote: I'm looking into adding hardware-frame support to make_writable, so modifications might not be needed. Yep it seems to be more consistent if av_frame_make_writable could support hardware frames. Please let me know if you are going to do it, or if

Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-19 Thread Timo Rothenpieler
On 19.03.2020 15:59, Yaroslav Pogrebnyak wrote: Got it, thanks! I'll re-do it and submit updated patch soon. I'm looking into adding hardware-frame support to make_writable, so modifications might not be needed. ___ ffmpeg-devel mailing list

Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-19 Thread Yaroslav Pogrebnyak
On 19.03.20 22:41, Timo Rothenpieler wrote: h264_cuvid copies frames back to normal VRAM, and does not pass around mapped nvdec surfaces, like nvdec does. Writing around in these is documented as disallowed. You can call av_frame_is_writable() on the frame. If it returns true, it's safe to

Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-19 Thread Timo Rothenpieler
On 19.03.2020 15:11, Yaroslav Pogrebnyak wrote: On 19.03.20 21:40, Timo Rothenpieler wrote: For what I'm aware, make_writable does not work on hardware frames. And the nvdec hwaccel returns frames that are mapped device memory, and thus hard read-only. You will need to manually allocate

Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-19 Thread Yaroslav Pogrebnyak
On 19.03.20 21:40, Timo Rothenpieler wrote: For what I'm aware, make_writable does not work on hardware frames. And the nvdec hwaccel returns frames that are mapped device memory, and thus hard read-only. You will need to manually allocate output frames from the hw_frames_ctx. Yes I see.

Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-19 Thread Timo Rothenpieler
On 19.03.2020 14:35, Yaroslav Pogrebnyak wrote: Oh, I didn't noticed that h264_cuvid is legacy. It seems the problem in this line: ret = av_frame_make_writable(input_main); If removed, it starts to work with -hwaccel cuda. I'll take a closed look why and what happens but any advice would be

Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-19 Thread Yaroslav Pogrebnyak
Oh, I didn't noticed that h264_cuvid is legacy. It seems the problem in this line: ret = av_frame_make_writable(input_main); If removed, it starts to work with -hwaccel cuda. I'll take a closed look why and what happens but any advice would be helpful. Thanks! On 19.03.20 21:15, Timo

Re: [FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-19 Thread Timo Rothenpieler
I'm currently trying to get this to work with nvdec, but seemingly can't: ./ffmpeg_g.exe -v verbose -hwaccel_output_format cuda -hwaccel cuda -i test_h264.mp4 -hwaccel_output_format cuda -hwaccel cuda -i test2_h264.mp4 -filter_complex

[FFmpeg-devel] [PATCH v2 2/2] avfilter: add vf_overlay_cuda

2020-03-18 Thread Yaroslav Pogrebnyak
Signed-off-by: Yaroslav Pogrebnyak --- Changes in v2: - Fixed switch() indentation style configure | 2 + libavfilter/Makefile | 1 + libavfilter/allfilters.c | 1 + libavfilter/vf_overlay_cuda.c | 446 +