lammps: GPU/Kokkos package updates#35885
Merged
alalazo merged 3 commits intospack:developfrom Mar 8, 2023
Merged
Conversation
This is for non-Kokkos GPU builds only.
alalazo
approved these changes
Mar 8, 2023
Member
alalazo
left a comment
alalazo
left a comment
There was a problem hiding this comment.
Checked package syntax and idioms and LGTM
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
LAMMPS has multiple GPU backends. The most actively developed one is for Kokkos, but the older GPU package also supports various backends, including HIP.
This adds the missing
rocmvariant to lammps. By default, if kokkos is not enabled, it will build with the GPU package, as has been the case with the existing+opencland+cudavariants.The patch is necessary because of the hipcub dependency. It gets installed into a non-standard location when used via Spack. The current LAMMPS cmake was making some assumptions which break in that case. Since I have to choose some sort of cutoff for maintaining this, the patch is only applied beginning from the last stable release.
This and another issue in upstream
developshould be resolved soon with lammps/lammps#3674I've also added a new dependent variant
gpu_precision, which was/is actually the most significant benefit of the GPU package over Kokkos, as it allows you to choose between single, mixed, or double precision when doing the offloading.For
+kokkos, I've also added the propagation of thecuda_archandamdgpu_targetproperties.