Merging of the Nvidia repository into Multimedia
The whole multimedia repository has been rebased with recent releases, and it now features FFmpeg 3.2 as the foundation. Most of the programs that suppport some Nvidia integration are now enabled and compiled with support for CUDA/NVENC/CUVID; leveraging the previous reorganization of CUDA 8 in the various subpackages.
This means that all the Nvidia packages are now included in the repository as well, so if you have an Nvidia card and you are interested in both repositories, you can just have the multimedia repository enabled. If you still just want the Nvidia stuff (as enabled in Fedora 25) then it’s still available as a separate repository; and that will not change.
Why all of this? Because I can’t keep them separated anymore. The Nvidia repository can exist on its own, but the multimedia one can’t, due to the dependencies and the constant rebases (also of main Fedora and CentOS/RHEL packages). You can use the Nvidia repository alone, if you just need that, or use the multimedia one if you need everything else.
The repository is now exposed also at this URL, and contains Delta RPM support:
https://negativo17.org/repos/multimedia
All repository files and configurations have been updated, so this means that in the future this would be the place where the metadata and repository information will be placed and any new installation will get the repository from there. If you are reading this blog post, you can switch now. I will add a negativo17-release package soon, along with a few mirrors; I’m sorting out the details now with the mirror owners.
FFmpeg and other CUDA enablements
To make proper use of the Nvidia hardware encode features (NVENC/CUVID) and CUDA kernel support (i.e. Blender GPU rendering) in the various programs you need the Nvidia driver installed (nvidia-driver-cuda), and for Nvidia Performance Primitives you require the CUDA driver and the NPP library package (cuda-npp).
This means that for most people NOT requiring CUDA support or not using an Nvidia video card, the following 2 packages will be installed anyway:
$ ls -alghs nvidia-driver-cuda-libs*.rpm cuda-npp*.rpm
92M -rw-r--r-T. 1 mock 92M Nov 16 12:35 cuda-npp-8.0.44-6.fc25.x86_64.rpm
22M -rw-r--r-T. 1 mock 22M Nov 19 15:00 nvidia-driver-cuda-libs-375.20-1.fc25.x86_64.rpm
Both packages contain just libraries, and they will be on your system as much as other libraries for multimedia codecs you don’t actually need. Example, with most multimedia programs you will get Xvid libraries for opening Xvid files, even though the format is pretty much abandoned. Having them installed does not enable any unwanted feature in your system. Also, NPP libraries should decrease 50% in size in one of the next CUDA updates, being the monolithic version of the library being deprecated in favor of split functionality.
There are some patches being evaluated to make those libraries loadable at runtime, but they have not been merged yet and there’s no guarantee that they ever will. Also, they are available for FFmpeg but not for all the other programs where support has been enabled for; so depending on your installation, you might get them anyway.
As of today, from the Multimedia repository the following programs have been enabled with some Nvidia hardware enablement:
- MPV (video decoding through CUVID)
- FFmpeg (encoding through NVENC, decoding through CUVID and filtering through CUDA NPP)
- Avidemux (encoding, through NVENC)
- GStreamer (NVENC plugin)
- Blender (GPU rendering)
VDPAU for decoding was already enabled where possible.
Of course anything that is using FFmpeg (like the GStreamer plugins) could theoretically benefit from the same enablements as in FFMpeg:
$ for i in encoders decoders filters; do
echo $i:; ffmpeg -hide_banner -${i} | egrep -i "npp|cuvid|nvenc|cuda"
done
encoders:
V..... h264_nvenc NVIDIA NVENC H.264 encoder (codec h264)
V..... nvenc NVIDIA NVENC H.264 encoder (codec h264)
V..... nvenc_h264 NVIDIA NVENC H.264 encoder (codec h264)
V..... nvenc_hevc NVIDIA NVENC hevc encoder (codec hevc)
V..... hevc_nvenc NVIDIA NVENC hevc encoder (codec hevc)
decoders:
V..... h263_cuvid Nvidia CUVID H263 decoder (codec h263)
V..... h264_cuvid Nvidia CUVID H264 decoder (codec h264)
V..... hevc_cuvid Nvidia CUVID HEVC decoder (codec hevc)
V..... mjpeg_cuvid Nvidia CUVID MJPEG decoder (codec mjpeg)
V..... mpeg1_cuvid Nvidia CUVID MPEG1VIDEO decoder (codec mpeg1video)
V..... mpeg2_cuvid Nvidia CUVID MPEG2VIDEO decoder (codec mpeg2video)
V..... mpeg4_cuvid Nvidia CUVID MPEG4 decoder (codec mpeg4)
V..... vc1_cuvid Nvidia CUVID VC1 decoder (codec vc1)
V..... vp8_cuvid Nvidia CUVID VP8 decoder (codec vp8)
V..... vp9_cuvid Nvidia CUVID VP9 decoder (codec vp9)
filters:
... hwupload_cuda V->V Upload a system memory frame to a CUDA device.
... scale_npp V->V NVIDIA Performance Primitives video scaling and format conversion
I think this will be much appreciated for you users out there that are already using CUDA for deep learning and FFMpeg to process data 🙂
Rebases: FFmpeg, HandBrake, VLC, OpenH264, WebP (CentOS/RHEL), MPV.
A note on Blender: Blender with CUDA support is still at 2.78 built with CUDA 7.5, and not 2.78a built with CUDA 8; so no Nvidia Pascal GPU support. I’m working on it.
GNOME Software integration
Most of the graphical software is now enabled in GNOME software for Fedora 25, meaning that you can search stuff with a keyword and that if you have the repository enabled it will just pop-up:
There is still some packages that need AppStream metadata, but that will come.
As usual, feedback, bugs and comments are welcome.
Recent Comments