-
For the `repro-ci` process, we need to move our default build of access-om3 out of ik11 and into a vk83 folder so it can be accessed by the ci user.
I assume that builds going into `/g/data/vk83/a…
-
### Bug summary
When numpy is compiled with Intel classic compilers (v2021.5.0) on Rocky 8.8 Linux, matplotlib v3.8.4 (also compiled with Intel) is unable to generate a basic plot.
### Code for …
-
Currently spack pipelines take about 15 minutes each. If we were to set up a binary mirror using the GHCR, then we could speed up builds substantially.
In addition to this, it would be good to run …
-
The test_gpu.cpp example included in the testsuite builds with a non-gpu variant of hpx, but it fails with both rocm and cuda variants. There are cuda specific api calls in the code so I'm not surpris…
-
h5py via pip now installs its own hdf5 library, which is a different version than what spack installs. Do both installs via spack to keep them unified.
-
Tasks for the spack-stack-1.8.0 release:
- [x] Create release branches
- [x] Install release candidate on a few selected systems for testing. Finalize testing/report back/fix issues by September 6. …
-
**Describe the bug**
`cray-mpich` does not work well with the default enabled `cudaMallocAsync`.
When building Kokkos with Spack, there is no way to disable `cudaMallocAsync`.
see : spack/spack…
-
### What is wrong?
The vint executable reports a duplicate module name when building with g2/v3.4.9 and w3emc/v2.10.0:
```bash log
cd /scratch1/NCEPDEV/global/David.Huber/GW/gw_python_3117/sorc/g…
-
### Steps to reproduce
1. `spack -d install -v --keep-stage lammps@20240627%gcc +kokkos+rocm amdgpu_target=gfx90a`;
2. go to the stage folder of the build and read `spack-build-02-build-out.txt`;
…
-
**Describe the bug**
With certain allocation patterns and sizes QuickPool never coalesces a backing allocation to be large enough, leading to unnecessary reallocations.
**To Reproduce**
Compi…