Skip to content

Refactor tests around shared MPI and backend helpers - #346

Open
ia267 wants to merge 11 commits into
xcompact3d:mainfrom
ia267:293_modular_test_cases
Open

Refactor tests around shared MPI and backend helpers#346
ia267 wants to merge 11 commits into
xcompact3d:mainfrom
ia267:293_modular_test_cases

Conversation

@ia267

@ia267 ia267 commented Aug 20, 2026 •
edited
Loading

Copy link
Copy Markdown
Collaborator

Closes #293

  • Centralise MPI initialisation, result reporting, and finalisation in shared test utilities
  • Add shared runtime setup for backend, allocator, and CUDA device selection while retaining OMP, CUDA, and OMP-target support
  • Use backend-neutral field access and solver operations to remove repeated preprocessor and select type blocks
  • Isolate backend-specific Fortran modules in the test build.

Testing

Built and ran the affected GNU/OpenMP and NVHPC/CUDA tests locally. OMP-target paths are covered by CI.

@ia267 ia267 added this to the Version 0.1.0 milestone Aug 20, 2026
@ia267 ia267 self-assigned this Aug 20, 2026
@ia267 ia267 added the enhancement New feature or request label Aug 20, 2026
@ia267
ia267 requested review from CFD-Xing and kaanolgu August 20, 2026 13:32
@ia267
ia267 marked this pull request as ready for review August 20, 2026 13:56
@ia267 ia267 mentioned this pull request Aug 21, 2026
Open
ia267 added 9 commits August 27, 2026 12:48
@ia267
add initialise_mpi common routine for tests
bcea74f
@ia267
add finalise test into common test utils
41e9736
@ia267
add backend_runtime abstraction and per-target module dirs
6b777f4
@ia267
simplify test_reordering with backend_runtime and field data helpers
b88c2e6
@ia267
add omp_tgt support to backend_runtime
b629a02
@ia267
Refactor tests to use shared backend runtime helpers
0ca18e7
@ia267
Refactor Thomas tests to use the backend runtime
5ae7709 
Expose Thomas solves, synchronisation, and device bandwidth information through the backend interface. Migrate the Thomas verification and performance tests away from backend-specific setup and kernel calls.
@ia267
refactor(tests): standardise MPI lifecycle and reporting
a26850b 
- adopt shared MPI initialisation and test finalisation helpers
- aggregate rank-local test results where required
- make statistics and scalar-output failures update allpass
- remove duplicate success reporting
@ia267
refactor(tests): centralise CUDA runtime handling
732dcd3 
- share rank-aware device selection across tests and benchmarks
- query bandwidth information from the active CUDA device
- standardise MPI lifecycle in remaining CUDA tests
- propagate benchmark runtime failures across MPI ranks
@ia267
ia267 force-pushed the 293_modular_test_cases branch from e61128a to 732dcd3 Compare August 27, 2026 11:53
kaanolgu
kaanolgu requested changes Sep 4, 2026
View reviewed changes
Comment thread tests/common/test_utils.f90 Outdated
Comment thread tests/CMakeLists.txt Outdated
Comment thread src/backend/backend_runtime.f90 Outdated
Comment thread tests/performance/perf_thom.f90 Outdated
@ia267
Address review comments
7a2e340 
- Aggregate allpass across ranks to synchronise failure state across ranks.
- Moved backend_runtime.f90 to common directory
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@kaanolgu kaanolgu kaanolgu requested changes
@CFD-Xing CFD-Xing Awaiting requested review from CFD-Xing

Assignees

@ia267 ia267

Labels

enhancement New feature or request

Projects

None yet

Milestone

Version 0.1.0

Development

Successfully merging this pull request may close these issues.

Simplifying the test cases

2 participants

@ia267 @kaanolgu