software-layer
software-layer copied to clipboard
Software layer of the EESSI project
In https://github.com/EESSI/software-layer/pull/419 we encountered a failing test for `Highway 1.0.3` on `neoverse_v1`: ``` 211 - HwyBlockwiseShiftTestGroup/HwyBlockwiseShiftTest.TestAllShiftRightLanes/SVE_256 # GetParam() = 33554432 (Subprocess aborted) ``` While debugging things interactively, sometimes this one...
See e.g. https://github.com/EESSI/software-layer/pull/434#issuecomment-1866264093 and https://github.com/EESSI/software-layer/pull/496#issuecomment-1996876402 I.e. it says: ``` other under 2023.06/software/linux/aarch64/neoverse_v1 2023.06/scripts/gpu_support/nvidia/install_cuda_host_injections.sh 2023.06/scripts/gpu_support/nvidia/link_nvidia_host_libraries.sh 2023.06/scripts/utils.sh .lmod/cache/spiderT.lua .lmod/cache/spiderT.luac_5.1 .lmod/cache/timestamp .lmod/lmodrc.lua ``` Clearly, `2023.06/scripts/utils.sh` is not _under_ `2023.06/software/linux/aarch64/neoverse_v1`, as that would be...
We've collected duplicate code in various (shell) scripts (see for example #467), we should clean that up a bit by "centralizing" common parts, for example to functions in `scripts/utils.sh`
Release 0.2.3 of archspec causes a failure in our CI: ``` Traceback (most recent call last): File "./eessi_software_subdir.py", line 7, in from archspec.cpu.detect import compatible_microarchitectures, raw_info_dictionary ImportError: cannot import name...
After running `link_nvidia_host_libraries.sh` the file permissions are: ``` $ ls -ald /cvmfs/software.eessi.io/host_injections/2023.06/compat/linux/x86_64/lib/ drwxr-x--- 2 userxyz userxyz 4096 Dec 22 14:13 /cvmfs/software.eessi.io/host_injections/2023.06/compat/linux/x86_64/lib/ ``` This made it work for me (I happened...
This is primarily due to `check_missing_installations.sh` and all the usage of `--from-pr` in our easystack files
While running the OSU benchmarks on a single system (CentOS Linux release 7.9.2009 (Core)), OpenMPI is giving the following errors: ``` [EESSI pilot 2021.06] $ mpirun -n 2 osu_bw --------------------------------------------------------------------------...
We're seeing 8 test failures for PyTorch 2.1.2 on all aarch64 targets, while Easybuild by default only allows 3 failures. For more details about the failing tests, see https://github.com/EESSI/software-layer/pull/444#issuecomment-1890416171. For...
Currently we have very long paths to get to our **installations**. We could dramatically reduce the size of these paths by removing the need to make them human readable. We...