Add ROCm-6.4.1 easyconfigs#25576
Conversation
Updated software
|
|
@zerefwayne Can you share a brief overview in the PR description how this differs from @Thyre's open PRs?
Should we first merge those PRs, or just go ahead with this one only? cc @casparvl |
|
Test report by @casparvl |
I'm guessing they don't: source is @Thyre 's repo where he had some customized easyconfigs for these from before there was a They should be the same I guess, but @zerefwayne please do a diff on each to ensure that's the case, then deduplicate by taking the ones out from your PR that are already provided by #25138 or #25148 The upside: this PR allowed me to do a test build with all This shows that easybuilders/easybuild-framework#5099 and easybuilders/easybuild-easyblocks#3861 are both working really well. |
|
@casparvl, @zerefwayne, @boegel Worth noting that I'm totally fine with closing my PRs and focusing on this one. |
| # Author: Jan André Reuter <j.reuter@fz-juelich.de> | ||
|
|
There was a problem hiding this comment.
I think you got the wrong file here by mistake 😄
There was a problem hiding this comment.
Thanks, fixed it :D
| # Author: Jan André Reuter <j.reuter@fz-juelich.de> | ||
|
|
|
If I'm seeing correctly, Feel free to pull them over from my PRs, then I'd go ahead and close mine. |
|
Test report by @casparvl |
| easyblock = 'CMakeMake' | ||
|
|
||
| name = 'rocm_bandwidth_test' | ||
| version = '1.4.0' |
There was a problem hiding this comment.
TODO: Update to 2.6.0
| } | ||
|
|
||
| # Sanity check commands fail on machines without AMD GPU (device not found) | ||
| # TODO: Reinstate them later to enable crosscompilation |
5705aa0 to
6f0a88a
Compare
...uild/easyconfigs/r/ROCmValidationSuite/ROCmValidationSuite-1.1.0_do-not-hardcode-paths.patch
Show resolved
Hide resolved
easybuild/easyconfigs/r/rocPRIM/rocPRIM-3.4.0-rocm-compilers-19.0.0-ROCm-6.4.1.eb
Outdated
Show resolved
Hide resolved
easybuild/easyconfigs/r/rocRAND/rocRAND-3.3.0-rocm-compilers-19.0.0-ROCm-6.4.1.eb
Outdated
Show resolved
Hide resolved
easybuild/easyconfigs/r/rocSOLVER/rocSOLVER-3.28.0-rocm-compilers-19.0.0-ROCm-6.4.1.eb
Outdated
Show resolved
Hide resolved
easybuild/easyconfigs/r/rocSOLVER/rocSOLVER-3.28.0-rocm-compilers-19.0.0-ROCm-6.4.1.eb
Outdated
Show resolved
Hide resolved
easybuild/easyconfigs/r/rocSOLVER/rocSOLVER-3.28.0_use_easybuild_paths.patch
Show resolved
Hide resolved
| builddependencies = [ | ||
| ('binutils', '2.42'), | ||
| ('CMake', '3.31.3'), | ||
| ('ROCm-LLVM', '19.0.0', f'-ROCm-{_rocm_version}', ('GCCcore', local_gcc_version)), |
There was a problem hiding this comment.
No longer needed?
| ('ROCm-LLVM', '19.0.0', f'-ROCm-{_rocm_version}', ('GCCcore', local_gcc_version)), |
easybuild/easyconfigs/r/rocSPARSE/rocSPARSE-3.4.0_fix-hardcoded-paths.patch
Show resolved
Hide resolved
| ('CMake', '3.31.3'), | ||
| ] | ||
| dependencies = [ | ||
| ('ROCm-LLVM', '19.0.0', f'-ROCm-{_rocm_version}', ('GCCcore', local_gcc_version)), |
There was a problem hiding this comment.
No longer needed?
| ('ROCm-LLVM', '19.0.0', f'-ROCm-{_rocm_version}', ('GCCcore', local_gcc_version)), |
| ('HIP', _rocm_version), | ||
| ('rocPRIM', '3.4.0'), | ||
| ('rocRAND', '3.3.0'), | ||
| # Tests will require SQLite, and a patch to remove the forced download of it |
There was a problem hiding this comment.
So, what does this mean? Test step isn't run now, but if we ever want to run it, we need to add SQlite as dep + a patch to not download SQlite?
| ('HIP', _rocm_version), | ||
| ('rocBLAS', '4.4.0'), | ||
| ('googletest', '1.17.0'), | ||
| ('rocm-smi', '7.6.0', '-ROCm-6.4.1', ('GCCcore', '14.2.0')), |
There was a problem hiding this comment.
Shouldn't be needed, it's a subtoolchain
| ('rocm-smi', '7.6.0', '-ROCm-6.4.1', ('GCCcore', '14.2.0')), | |
| ('rocm-smi', '7.6.0', '-ROCm-6.4.1'), |
| ('PyYAML', '6.0.2'), | ||
| # Tensile uses both msgpack Python bindings and C++ header files when actually using Tensile | ||
| # Therefore, add msgpack-cxx as a dependency and msgpack as a Python package, even though | ||
| # its messy. Also bumps this EasyConfig to GCC from GCCcore, due to msgpack-cxx relying on Boost... |
There was a problem hiding this comment.
"bumps to GCC" yet msgpack-cxx is now at rocm-compilers level.
There was a problem hiding this comment.
Since msgpack-cxx has been moved to rocm-compilers, this comment seems to be redundant. Removing it.
easybuild/easyconfigs/h/hipRAND/hipRAND-2.13.0-rocm-compilers-19.0.0-ROCm-6.4.1.eb
Outdated
Show resolved
Hide resolved
…pilers` as toolchain. Co-authored-by: Caspar van Leeuwen <33718780+casparvl@users.noreply.github.com>
|
@boegelbot please test @ jsc-zen3 |
|
@casparvl: Request for testing this PR well received on jsczen3l1.int.jsc-zen3.fz-juelich.de PR test command '
Test results coming soon (I hope)... Details- notification for comment with ID 4179851072 processed Message to humans: this is just bookkeeping information for me, |
Something went wrong somewhere :D |
easybuild/easyconfigs/r/rocm-core/rocm-core-6.4.0-GCCcore-14.2.0-ROCm-6.4.1.eb
Outdated
Show resolved
Hide resolved
Co-authored-by: Caspar van Leeuwen <33718780+casparvl@users.noreply.github.com>
|
@boegelbot please test @ jsc-zen3 |
|
@casparvl: Request for testing this PR well received on jsczen3l1.int.jsc-zen3.fz-juelich.de PR test command '
Test results coming soon (I hope)... Details- notification for comment with ID 4180234527 processed Message to humans: this is just bookkeeping information for me, |
|
Test report by @boegelbot |
Needs rocm-compilers toolchain from PR: easybuilders/easybuild-framework#5099Needs ROCmComponent EasyBlock from PR: easybuilders/easybuild-easyblocks#3861
The easyconfigs are adapted from @Thyre's work in #25138 and #25148 and modified to use the new rocm-compilers toolchain (instead of GCCcore) and ROCmComponent easyblock. (EB_ROCmComponent). Some more easyconfigs are ported from https://github.com/Thyre/easybuild-custom/tree/support-passing-amdgcn