Dear Easybuilders,
We're configuring an AMD EPYC 7313 server with two AMD MI210 GPUs. AMD
provides instructions for installing Release-specific AMDGPU and ROCm
Repositories on Linux Distributions in the page
https://rocm.docs.amd.com/en/latest/deploy/linux/os-native/install.html
The AMD instructions allow us to install multiple versions of ROCm,
however, I'm missing the ability to create software modules which
conveniently set PATH, LD_LIBRARY_PATH etc.
I looked at the EB software list
https://docs.easybuild.io/version-specific/supported-software/#rocm but
only an ancient version 4.5.0 is offered. The currently latest version is
5.7.2. We'd like to have at least version 5.3.3 which is installed on the
LUMI supercomputer so that we would be compatible (to some extent) with LUMI.
Question: Has anyone created EB module files to build ROCm manually, or as
a wrapper about AMD's ROCm RPM packages?
Thanks a lot,
Ole
--
Ole Holm Nielsen
PhD, Senior HPC Officer
Department of Physics, Technical University of Denmark