Heterogeneous Computing Special Interest Group
Mission
To encourage the packaging and accessibility of heterogeneous computing projects in Fedora and EPEL. This includes machine learning, OpenCL, and scientific computing.
Communication
While separate from the AI/ML SIG, we do share communication channels for now. If you're interested, please join us on:
- #ai-ml@fedoraproject.org Matrix room
- #fedora-ai-ml on libera.chat IRC (NOTE: With the recent Matrix/IRC bridge issues, there may not be much activity on IRC. More conversation happens in the Matrix room)
- #ai-ml-sig on Fedora's Discourse instance
Members/Packagers/Reviewers
- Jeremy Newton (ROCm related)
- Felix Schwarz (currently busy with certbot, borgbackup, WeasyPrint)
- Benson Muite
- Morgan Hough (Intel related while testing SODECL on Intel NUCs)
- Tom Rix (Misc)
- Tim Flink
- Brian Lu
- Jonathan Steffan
- David DUncan (currently working on Fedora Cloud Edition and want to build Deep Learning Images for public cloud)
Please add your name above if you want to help.
Tasks
- Package and make HC related projects more accessible to users (such as OpenCL, AMD's ROCm HIP, Intel oneAPI, SYCL, Vulkan, OpenGL, etc.)
- Document installation, packages, and general use cases
- Translation of wiki pages and documentation (if applicable)
AMD's ROCm (OpenCL/HIP)
Package status
Below is a list of ROCm packages with their status. If something isn't packaged in Fedora or EPEL, feel free to take it!
For reference, pytorch still requires: rocblas, miopen, rocfft, hipfft, hipsparse, rccl, rocprim, hipcub, rocthrust
HW Support
Some ROCm packages in Fedora are built for all 64bit little endian systems (x86_64, arm64, ppc64le), but official support is for x86_64.
HW support is highly dependent on LLVM version included in Fedora, so if you have newer HW, you might need to wait for newer LLVM versions to be introduced to Fedora (usually in the spring and the autumn).
Installation
Setup
To enable access to GPU resources for non-root users, the users must be added to the video
or render
group. The AMD ROCm Installation Guide (v5.2) and FAQs recommend the video
group for all ROCm-supported operating systems.
To add the current user to the video
group:
sudo usermod -a -G video $LOGNAME
To check for ROCm support of the running HW, you can install and run rocminfo
:
sudo dnf install rocminfo rocminfo
OpenCL
ROCm OpenCL can be installed with:
sudo dnf install rocm-opencl
As well, rocm-clinfo
or clinfo
can be installed to verify it is working, e.g.:
sudo dnf install rocm-clinfo rocm-clinfo
HIP
HIP is included in Fedora 39 and is pending for Fedora 38. For Fedora 38, you will need to add --enable-repo=updates-testing to dnf if it's not available yet.
Install with:
sudo dnf install rocm-hip
For use with Blender, you might need to also install "rocm-hip-devel" if it does not detect HIP.
Intel
TODO: incomplete, take me!
Packages
Packages that use HC
Blender
Build with:
rpmbuild --with rocm