Skip to content

copilot-instructions.md

Latest commit

 

History

History
52 lines (38 loc) · 3.58 KB

copilot-instructions.md

File metadata and controls

52 lines (38 loc) · 3.58 KB

AI Code Review Instructions for MFC

These instructions guide AI code reviewers (GitHub Copilot, CodeRabbit, Qodo, Cubic, CodeAnt, etc.) when evaluating pull requests in this repository.

Coding standards, common pitfalls, and contribution guidelines: docs/documentation/contributing.md GPU macro API reference: docs/documentation/gpuParallelization.md

Formatting and linting are enforced by pre-commit hooks. Focus review effort on correctness, not style.

1 Project Context

  • Project: MFC (exascale many-physics solver) written in modern Fortran 2008+, preprocessed with Fypp.
  • Directory layout:
    • Sources in src/, tests in tests/, examples in examples/, Python toolchain in toolchain/.
    • Most source files are .fpp (Fypp templates); CMake transpiles them to .f90.
  • Fypp macros are in src/<subprogram>/include/, where <subprogram> is simulation, common, pre_process, or post_process.
  • Only simulation (plus its common dependencies) is GPU-accelerated via OpenACC or OpenMP target offload (--gpu acc or --gpu mp). GPU code uses backend-agnostic GPU_* Fypp macros (in src/common/include/parallel_macros.fpp) that dispatch to the correct backend at compile time.
  • Code must compile with GNU gfortran, NVIDIA nvfortran, Cray ftn, and Intel ifx.
  • Precision modes: double (default), single, and mixed (wp = working precision, stp = storage precision).
  • Python toolchain requires Python 3.10+ — do not suggest from __future__ imports or other backwards-compatibility shims.

2 What to Review

See the Common Pitfalls section of docs/documentation/contributing.md for the full reference. Key review priorities:

  1. Correctness over style — logic bugs, numerical issues, array bounds (non-unity lower bounds with ghost cells)
  2. Precision mixing — stp vs wp conversions, no double-precision intrinsics (dsqrt, dble, real(8), etc.)
  3. Memory — @:ALLOCATE/@:DEALLOCATE pairing, GPU pointer setup (@:ACC_SETUP_VFs/@:ACC_SETUP_SFs)
  4. MPI — halo exchange pack/unpack offsets, GPU_UPDATE around MPI calls, buffer sizing
  5. GPU — no raw OpenACC/OpenMP pragmas (use Fypp GPU macros), private(...) on loop-local variables, no stop/error stop in device code
  6. Physics — pressure formula must match model_eqns, case_validator.py constraints for new parameters
  7. Compiler portability — all four compilers, Fypp macros for GPU and CPU builds
  8. src/common/ blast radius — changes affect all three executables

3 PR-Pattern Triggers

Flag these patterns when reviewing a pull request:

  • PR adds a parameter in toolchain/mfc/params/definitions.py but does not update the namelist /user_inputs/ in src/*/m_start_up.fpp or declare it in src/*/m_global_parameters.fpp
  • PR adds a parameter with cross-parameter constraints but does not add validation in toolchain/mfc/case_validator.py
  • PR modifies files in src/common/ but does not mention testing all three targets (pre_process, simulation, post_process)
  • PR adds @:ALLOCATE calls without matching @:DEALLOCATE in the corresponding finalization subroutine
  • PR renames or moves files referenced in docs/documentation/contributing.md or docs/documentation/gpuParallelization.md without updating those docs
  • PR adds a new .fpp module missing implicit none or missing private/public declarations
  • PR modifies MPI pack/unpack logic in one sweep direction without updating all directions
  • PR adds a new physics feature or model without a corresponding regression test