Tests

ADAM uses a multi-level testing strategy to ensure correctness across backends and configurations.

Unit Tests

Finite difference operator tests with trigonometric analytical solutions:

# Build a specific test
fobis build --mode test-fdv-gradient-trigonometric-gnu
fobis build --mode test-fdv-divergence-trigonometric-gnu
fobis build --mode test-fdv-curl-trigonometric-gnu
fobis build --mode test-fdv-laplacian-trigonometric-gnu
fobis build --mode test-fdv-operators-trigonometric-gnu
fobis build --mode test-fdv-operators-step-gnu

# Run with MPI
mpirun -np <N> exe/<test_executable>

Regression Tests

The PRISM regression suite is the structural-change baseline for the Maxwell/EM solver: CI-tuned two-rank cases run on both the CPU and FNL (GPU) backends, compared tolerance-aware against committed field-digest and residuals goldens. It covers the forest multi-realm seam machinery (α/β cadence, inter-realm 1:1 mirror seams), the intra-realm 2:1 AMR seams (with check.sh-driven div(B) oracles), reflux, and the fWLayer. See the Forest guide for the seam machinery itself.

To stand the suite up on a new machine, compiler, or GPU architecture, follow the regression bring-up tutorial.

Current status (cf16e20d)

The suite runs again after the [fWLayer] C → width input migration, but is not fully green: 6 PASS, 1 FAIL, 3 SKIP. The rmf-amr failure is an unadjudicated FV-path behaviour change, and rmf-fwl is still un-migrated. See the suite page for details.

Research and development cases live separately under src/tests/prism/<backend>/{fd,fv}/<family>/ — long integration times, full AMR, sized for physics validation. They are not regression anchors and run.sh never descends into them.

Integration Tests

NASTO integration tests with known CFD benchmarks:

Test	Description
Sod-X	1D Riemann Problem of Sod along X axis
Sod-Y	1D Riemann Problem of Sod along Y axis
Sod-Z	1D Riemann Problem of Sod along Z axis
Shock-Sphere	Shock-sphere interaction simulation

Compiler Testing Matrix

All changes affecting core numerics or parallel logic must pass:

• GNU gfortran — baseline CPU, strict bounds checking
• NVIDIA nvfortran — GPU targets (CUDA Fortran, OpenACC)
• Intel ifort/ifx — aggressive optimization, vectorization reports

Use -fbounds-check -fcheck=all (GNU) or -check all -traceback (Intel) during development.