[Bug 4165] Review Request: nvidia-video-codec-sdk - Hardware accelerated video encode and decode

https://bugzilla.rpmfusion.org/show_bug.cgi?id=4165 --- Comment #5 from Nicolas Chauvet <kwizart(a)gmail.com&gt; 2016-08-05 14:42:42 CEST --- (In reply to comment #4) .. Sure, it's optional at buildtime and runtime. But then at runtime nvenc enabled ffmpeg will dlopen libnvidia-encode.so (the unversionned SO) That libraries is currently provided by the -cuda subpackage. (despite it's name, libraries doesn't all relate to cuda indeed) If the cuda option is enabled in ffmpeg, it will also dlopen libcuda.so The main purpose of this cuda sub-package was to have all libraries that would only rely on -cuda + nvidia-kmod (and not main + -libs). So cuda developpers would only needs -cuda + nvidia-kmod to install the NVIDIA CUDA toolkit. (at some point, the nvEncodeAPI.h was distributed as part of the toolkit IIRC). This was specially usefull for optimus enabled laptop using the intel adapter for display and using the nvidia GPU for CUDA (that will then miss the cuda/gl interrop, but not as much important). Now the real concern is to have the ffmpeg binary not to crash if users try to use ffmpeg nvenc without the required hw or installed libraries. Okay, so no direct link time dependencies, only dlopen as expected. That ffmpeg build will be provided in the nonfree section and will be allowed to build nonfree dependencies and libraries along with enabling nvenc (and as such dlopen a nonfree library). -- Configure bugmail: https://bugzilla.rpmfusion.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug. You are the assignee for the bug.