Am 02.06.2017 um 19:47 schrieb Ganapathy Raman Kasi:
Hi,
HW accelerated transcode (h264_cuvid -> h264_nvenc with -hwaccel cuvid) was
broken after the filtergraph initialization was changed to intialize decoder first
followed by encoder (commit af1761f7b5b1b72197dc40934953b775c2d951cc). During
initialzing encoder with bframes, local buffers are allocated internally in
encoder which fails since no cuda context is bound to current cpu thread. Now
pushing the correct cuda context before encoder initialization fixes the issue.
Also adding push/pop cuda ctx during create/destroy/map/unmap resources and
destroy encoder session.
Regards
Ganapathy
applied, thanks!
Will also backport to affected release branches.
What I'm still wondering is, how did this ever happen to work?
I suspected something like this, and verified that we were not leaking a
bound context somewhere, and thus gave up on that trail.
Changing the order of initialization, or rather throwing initializing
cuvid before nvenc, seems to be what causes this.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
