Skip to content

[GPU] evaluateDerivatives (SPH and SolidSPH)#456

Merged
ldowen merged 69 commits into
developfrom
feature/gpu/eval-deriv
Apr 24, 2026
Merged

[GPU] evaluateDerivatives (SPH and SolidSPH)#456
ldowen merged 69 commits into
developfrom
feature/gpu/eval-deriv

Conversation

@ldowen
Copy link
Copy Markdown
Collaborator

@ldowen ldowen commented Jan 15, 2026

Summary

  • This PR is a feature.
  • It does the following:
    • Ports the loop over node pairs in evaluateDerivatives to run on device for some of the SPH methods.

ToDo :

  • Annotate RELEASE_NOTES.md with notable changes.
  • Create LLNLSpheral PR pointing at this branch. (PR#204)
  • LLNLSpheral PR has passed all tests.

…n, added openmp as one of the cpp test execution options, fixed the cpp tests so gpu counters are only filled if the test is a proper gpu test
…xed type discrepancy in artificial viscosity classes, added basic exec policy to config that is subject to improvement, started porting the eval deriv in SPH.cc
@ldowen ldowen linked an issue Jan 15, 2026 that may be closed by this pull request
ldowen added 22 commits January 28, 2026 09:26
… compiler flags to prevent issues with clang 20
…ion off workarounds, added function to increase GPU stack to resolve device side virtual function errors
ldowen added 24 commits April 8, 2026 12:44
…raja_test ATS value so RAJA tests are skipped when the build has HIP enabled but the ats run does not, removed the cap on ranks for cray machines which will cause issues with historical benchmarks
…pped because test name was the same causing the ATS to continue from the last run, removed unused capture of test-logs since it retained logs from any previous failed pipeline
… and Q interpolators, moved touch, move, and free wrappers into GPUUtils to minimize ifdef statements, added element-by-element comparison operator for PairwiseField
… so that it is properly set when running with unified memory
…me in NodePairListView, other small cleanups
… run_perf and into the ci script, no longer names directories as latest
…les, added chmod and chgrp calls for performance test, redid hist plots so each plot is for a single spec and test base with lines for the test variations
Comment thread docs/developer/dev/gpu_dev.rst
@ldowen ldowen linked an issue Apr 24, 2026 that may be closed by this pull request
@ldowen ldowen merged commit 548becc into develop Apr 24, 2026
2 checks passed
@ldowen ldowen deleted the feature/gpu/eval-deriv branch April 24, 2026 17:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[GPU] evalDerivatives Replace use of std::iterator

2 participants