-
Notifications
You must be signed in to change notification settings - Fork 32
Spack 1.1 & TPL v2025.12.0 updates #1769
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
| hip_link_flags = "" | ||
|
|
||
| rocm_root = os.path.dirname(spec["llvm-amdgpu"].prefix) | ||
| rocm_root = spec["llvm-amdgpu"].prefix |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Here and in toss_4_x86_64_ib_cray/spack.yaml, I incorporated changes from @adrienbernede 's PR #1739 to get things working with Spack 1.1 .
| +// Include OpenMP header for Windows | ||
| +#if defined(_OPENMP) | ||
| +#include <omp.h> | ||
| +#endif |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This patch file is part of my attempt to get raja v25.12.0 working with Windows.
Following error without this patch:
D:\a\axom\axom\uberenv_libs\vcpkg\buildtrees\raja\src\v2025.12.0-beae201c76.clean\include\RAJA/policy/openmp/thread.hpp(35): error C3861: 'omp_get_max_threads': identifier not found
D:\a\axom\axom\uberenv_libs\vcpkg\buildtrees\raja\src\v2025.12.0-beae201c76.clean\include\RAJA/policy/openmp/thread.hpp(41): error C3861: 'omp_get_thread_num': identifier not found
b1ced68 to
2b540e7
Compare
|
Additional context from testing: On Windows, I've tried building RAJA EDIT: Between v2025.09.0 and v2025.09.1, this seems to be the most relevant PR to openmp changes? llnl/RAJA#1884 |
2e8ff18 to
6c57b7f
Compare
|
@bmhan12 to be clear: it works with RAJA 2025.09.0 and breaks with RAJA 2025.09.1 and newer? |
Yes. |
Thank you. I've got John looking into it. His RAJA PR seems to be the cause of the breakage. |
9f2f2dc to
ea0f6f7
Compare
6358442 to
a2c6564
Compare
| "conduit", | ||
| "lua", | ||
| "mfem", | ||
| "openmp", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
After talking with @kennyweiss, I decided we can disable OpenMP on Windows for now, and can enable it again at a later date.
It was also mentioned that Umpire is actually disabled in our Windows testing, I left a comment in this file to track these changes.
Tag: @johnbowen42
| umpire: | ||
| require: | ||
| - spec: "@2025.09.0" | ||
| - spec: "@git.0372fbd6e1f17d7e6dd72693f8b857f3ec7559e9=develop" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why use the commit here compared to the version tag ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
2025.12.0 isn't in upstream spack yet, so I went with the commit hash.
| buildable: false | ||
| externals: | ||
| - spec: mvapich2@2.3.7.clang_19~alloca~cuda~debug~hwloc_graphics~hwlocv2+regcache+wrapperrpath build_system=autotools ch3_rank_bits=32 fabrics=mrail file_systems=auto process_managers=slurm threads=multiple | ||
| - spec: mvapich2@2.3.7.clang_19~alloca~cuda~debug~hwloc_graphics~hwlocv2+regcache+wrapperrpath ch3_rank_bits=32 fabrics=mrail file_systems=auto process_managers=slurm threads=multiple |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm curious how you verify those variants for external installs. Are they provided by the spack external detection tool ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, that is where I originally got all these variants, using spack to detect them.
Turns out this is probably not the best idea, since the packages and detection can change between Spack versions.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You actually don't need all the variants anymore. It was needed when you couldn't specify which mvapich you needed per compiler before toolchains. Basically we were forcing the concretizer to select the right one. You can do this now:
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@white238 so is just the version sufficient now for externals?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes, as long as you specify it in the toolchain
rhornung67
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thank you for working through this stuff @bmhan12 LGTM!
Co-authored-by: Adrien M. Bernede <bernede1@llnl.gov>
c85da13 to
9b928ef
Compare
This PR:
TODO:
Latest attempt: https://github.com/llnl/axom/actions/runs/20835993048