[CI probe, do not merge] CCE OpenACC EOS helper diagnostics by sbryngelson · Pull Request #1812 · MFlowCode/MFC

sbryngelson · 2026-09-03T17:04:50Z

Temporary probe for the Frontier CCE OpenACC NaNs seen on #1811. Runs one lane on seven targeted tests and prints each EOS helper evaluated on the device next to its host value at startup. Will be closed once the failing construct is identified.

Any Mie-Gruneisen EOS p = p_ref + rho Gamma_G (e - e_ref) is MFC's rho e = Gamma p + Pi with Gamma = 1/Gamma_G and Pi = rho e_ref - p_ref/Gamma_G, so the existing operators need only the coefficients and dPi/drho. s_eos_coefficients is the single dispatch: the Mie-Gruneisen case supplies a linear-Hugoniot reference curve (u_s = c0 + s u_p, linear release) and one shared conversion produces Gamma, Pi and dPi/drho; a second family is one more case. Stiffened and ideal gas return the constants resolved at init, and no caller invokes the routine yet, so every existing answer is bit-identical - 27-case gate. Parameters mg_rho0, mg_c0, mg_s, mg_G0 follow the flat per-fluid style; the validator requires all four under mie_gruneisen, forbids them otherwise, and forbids qv there because e_ref carries the formation energy. Thirty pytest checks pin the maths against finite differences and a numerically integrated isentrope.

A Mie-Gruneisen fluid has neither gamma, pi_inf nor qv, and an ideal gas has no pi_inf: the validator now refuses them, and the suite's base fluid and every example drop the dead values. The ideal-gas half duplicates MFlowCode#1808 and drops out once it merges.

The mixture loop calls s_eos_coefficients per phase when any fluid's EOS is state dependent; the stiffened-gas arithmetic is untouched (27/27 bit-identical on the gate). The sound speed takes the partial densities as an optional argument and mixes frozen per-phase moduli, with the derivative term applied at that one site. Callers that cannot supply them are refused by the validator: 5-equation, HLL/HLLC/LF only, no bubbles, hypoelasticity, IBM, IGR, relativity, MHD, chemistry, acoustic source, probes, post-process c, characteristic BCs or Wood's law. Validation: symmetric-impact shock speed and plateau density match the Hugoniot to 1.4e-3, and a small pulse travels at the analytic sound speed to 1.9e-4 (the missing derivative term would give 24%).

gamma = gammas(1) and friends were assigned and never read; lit_gamma was assigned in both output routines and read in neither.

…eck passes from worktrees (cherry picked from commit 5ad5c85)

…ristic BCs, probes and post-process c

…isen-jwl

Isentropic release of a JWL fluid lands on the closed-form isentrope to 7.5e-5 at N = 200 and 9e-7 at N = 800; goldens for JWL beside an ideal gas, Mie-Gruneisen under Wood's law, and Mie-Gruneisen with characteristic walls. The acoustic pulse now comes from a rectangle patch and is tracked by its centroid: an analytic IC is compiled in and every distinct one costs the suite a full rebuild.

One reference-curve routine with a Gruneisen closure Gamma_G = Gamma_0 + a mu, one coefficient conversion, one sound-speed formula and one RK4 stepper for the phasic isentrope and the reference temperature. Adds the temperature T = T_ref + (e - e_ref)/c_v and its T_wrt output, the cubic Hugoniot, the Vinet cold curve, the 6-equation model on the phasic isentrope, and helpers that fetch their own coefficients so no feature reads gammas(i) directly: hypoelasticity, IBM, acoustic sources and the reactive burn open up. The validator refuses initial states outside a state-dependent EOS and keeps qv as the shared formation-energy zero.

Goldens for Vinet, the cubic Hugoniot, MG with hypoelasticity, with IBM, and a MG reactant burning to JWL products; the isentropic release now serves JWL and Vinet and the Hugoniot recovery a quadratic fit. The convergence harness runs a private copy of each case in a temporary directory without rebuilding, so specs that share a case file can run concurrently.

Copilot

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Pull request overview

Temporary CI probe to help isolate Frontier CCE OpenACC NaNs by (a) adding device/host EOS helper diagnostics at startup and (b) narrowing CI to a small set of targeted cases; it also extends the toolchain/params to recognize additional EOS selector values and related parameters.

Changes:

Add state-dependent EOS selector values/params (Mie-Gruneisen/JWL/Vinet) to the Python toolchain and add selector/constraint tests.
Add GPU startup diagnostics that evaluate EOS helpers on device and host and print both for comparison.
Modify CI workflow/scripts to disable most lanes and run only a small targeted test subset for the probe.

Reviewed changes

Copilot reviewed 185 out of 188 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
toolchain/mfc/test_case_validator.py	Adds constraint tests for new EOS selectors/parameters.
toolchain/mfc/test/test_coverage_unit.py	Switches tests to import shared git-env scrub helper from coverage module.
toolchain/mfc/test/coverage.py	Scrubs inherited `GIT_*` env vars for subprocess git calls to avoid hook contamination.
toolchain/mfc/test/case.py	Removes ideal-gas `pi_inf` default from base case params.
toolchain/mfc/params_tests/test_eos_selector.py	Updates ideal-gas behavior to reject/ignore stiffness parameter.
toolchain/mfc/params/descriptions.py	Documents new `T_wrt` output option.
toolchain/mfc/params/definitions.py	Registers `T_wrt`, expands EOS selector mappings, and registers MG/JWL/Vinet parameters.
toolchain/mfc/eos.py	Adds Python reference implementations for MG/JWL/Vinet curves and related helpers (tests/validation).
tests/ED75D01D/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/A6846AD4/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/A421E318/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/7853BD45/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/6AE3FB4E/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/61AF4509/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/6191C8D0/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/5AC2F65D/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/590E4427/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/4A0CDF9C/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/471270BB/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
tests/120043F6/golden-metadata.txt	Adds/updates golden metadata for targeted probe runs.
src/simulation/m_time_steppers.fpp	Threads `alpha_rho` through cell-state and sound-speed computation (state-dependent EOS support).
src/simulation/m_start_up.fpp	Adds startup GPU/host EOS helper probe + updates internal-energy call signature.
src/simulation/m_sim_helpers.fpp	Extends `s_compute_cell_state` to return `alpha_rho` rather than a local scratch.
src/simulation/m_riemann_solver_lf.fpp	Passes `alpha_rho` into sound-speed routine.
src/simulation/m_riemann_solver_hypo_hlld.fpp	Uses per-phase bulk modulus helper for alt sound speed path.
src/simulation/m_riemann_solver_hllc.fpp	Passes `alpha_rho` into sound-speed routines and switches to phasic isentrope/internal-energy helpers.
src/simulation/m_riemann_solver_hll.fpp	Passes `alpha_rho` into sound-speed routine(s).
src/simulation/m_rhs.fpp	Switches bulk modulus computation to phasic bulk modulus helper.
src/simulation/m_reactive_burn.fpp	Uses `f_phase_temperature` instead of stiffened-gas-only thermal relation.
src/simulation/m_pressure_relaxation.fpp	Adds state-dependent EOS path via `s_phase_density_on_isentrope` and updates internal-energy call signature.
src/simulation/m_ibm.fpp	Updates internal-energy call signature to EOS-index-based API.
src/simulation/m_hypoelastic.fpp	Uses phasic bulk modulus helper for 2-material K computation.
src/simulation/m_global_parameters.fpp	Initializes newly-added EOS parameter fields in `fluid_pp`.
src/simulation/m_data_output.fpp	Threads `alpha_rho` through sound-speed computations and removes unused locals.
src/simulation/m_cbc.fpp	Passes `alpha_rho` into sound-speed and mixture-coefficient time-derivative routines.
src/simulation/m_acoustic_src.fpp	Inlines state-dependent per-phase modulus mixing to avoid OpenACC issue.
src/pre_process/m_global_parameters.fpp	Initializes newly-added EOS parameter fields in `fluid_pp` (pre-process).
src/pre_process/m_data_output.fpp	Removes unused `lit_gamma` plumbing.
src/post_process/m_start_up.fpp	Threads `alpha_rho` into sound-speed computation and adds `T_wrt` temperature output.
src/post_process/m_global_parameters.fpp	Initializes new EOS fields and defaults `T_wrt` to false.
src/post_process/m_derived_variables.fpp	Removes old `s_derive_sound_speed` implementation (sound speed handled elsewhere now).
src/post_process/m_data_output.fpp	Updates internal-energy and sound-speed calls to EOS-index-based API.
src/common/m_global_parameters_common.fpp	Adds device-resident arrays for EOS selector and reference-curve parameters.
src/common/m_derived_types.fpp	Extends `physical_parameters` with MG/JWL/Vinet parameter fields.
src/common/m_constants.fpp	Adds EOS selector constants and RK4/ODE constants for state-dependent EOS helpers.
examples/3D_turb_mixing/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_sphbubcollapse/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_rotating_sphere/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_recovering_sphere/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_rayleigh_taylor_muscl/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_rayleigh_taylor/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_performance_test/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_patch_spherical_harmonic/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_moving_lag_particles/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_mibm_sphere_head_on_collision/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_mibm_periodic_collision/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_lagrange_shbubcollapse/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_lagrange_bubblescreen/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_icpp_stl_cube/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_ibm_stl_test/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_ibm_stl_pyramid/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_ibm_stl_ellipsoid/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_ibm_bowshock/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_brio_wu/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_advection_convergence/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_TaylorGreenVortex_analytical/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_TaylorGreenVortex/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_IGR_jet_1fluid/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_IGR_jet/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_IGR_TaylorGreenVortex_nvidia/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_IGR_TaylorGreenVortex/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/3D_IGR_33jet/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_zero_circ_vortex_analytical/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_zero_circ_vortex/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_viscous_shock_tube/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_tumbling_rectangle/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_triple_point/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_synthetic_turbulence/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_shockdroplet_muscl/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_shockdroplet/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_shockbubble/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_shock_cloud_rmhd/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_riemann_test_muscl/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_riemann_test/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_richtmyer_meshkov/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_rayleigh_taylor/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_poiseuille_thickening_nn/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_poiseuille_nn/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_patch_modal_shape_exp/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_patch_modal_shape/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_orszag_tang_hyper_cleaning/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_orszag_tang/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_moving_lag_bubs/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_mibm_shock_cylinder/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_mibm_particle_cloud/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_mibm_cylinder_in_cross_flow/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_mhd_rotor/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_mhd_magnetic_vortex/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_lid_driven_cavity_nn/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_lid_driven_cavity/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_laplace_pressure_jump/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_lagrange_rising_bubble/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_lagrange_in_crossflow/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_lagrange_bubblescreen/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_kelvin_helmholtz/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_jet/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_isentropicvortex_analytical/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_isentropicvortex/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_icpp_stl_circle/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm_viscous_drag_over_cylinder/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm_stl_wedge/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm_stl_test/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm_stl_MFCCharacter/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm_poiseuille_nn/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm_multiphase/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm_ellipse/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm_cfl_dt/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm_airfoil/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_ibm/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_hypo_shear_contact/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_herschel_bulkley_poiseuille_nn/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_hardcoded_ic/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_forward_facing_step/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_bingham_poiseuille_nn/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_backward_facing_step/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_axisym_shockbubble/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_advection_muscl/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_advection_convergence/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_advection/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_acoustic_pulse_analytical/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_acoustic_pulse/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_acoustic_broadband/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_TaylorGreenVortex/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_IGR_triple_point/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_IGR_2fluid/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/2D_GreshoVortex/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_vacuum_restart/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_vacuum/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_titarevtorro_analytical/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_titarevtorro/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_sodshocktube_muscl/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_sodshocktube/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_sod_convergence/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_shuosher_wenoz5/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_shuosher_wenom5/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_shuosher_wenojs5/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_shuosher_teno7/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_shuosher_teno5/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_shuosher_old/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_shuosher_analytical/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_mhd_smooth_alfven_wave/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_laxshocktube/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_kapilashocktube/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_hypo_2materials/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_euler_convergence/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_dai_woodward_hlld/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_dai_woodward/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_convergence/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_brio_wu_rmhd/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_brio_wu_hlld/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_brio_wu/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_advection_convergence/case.py	Removes ideal-gas `pi_inf` from example inputs.
examples/1D_mg_impact/case.py	Adds a new MG impact validation case generator.
examples/1D_mg_acoustic/case.py	Adds a new MG acoustic-pulse validation case generator.
examples/1D_isentropic_release/case.py	Adds a new JWL/Vinet isentropic-release validation case generator.
docs/documentation/contributing.md	Updates contributor docs for EOS helper inventory.
docs/documentation/case.md	Documents `T_wrt` output option in case configuration docs.
.github/workflows/test.yml	Disables most CI jobs and narrows the self-hosted matrix for the probe.
.github/workflows/common/test.sh	Runs only a fixed small subset of tests on pull requests.
.github/scripts/check_coverage_map_health.py	Switches git calls to shared `_git` helper.
.claude/rules/common-pitfalls.md	Adds a note about a CCE OpenACC GPU_ROUTINE argument pitfall (context for #1811).

Suppressed comments (3)

toolchain/mfc/test_case_validator.py:1

The new test classes are defined after the unittest.main() call. If this module is executed directly (e.g., python test_case_validator.py), unittest.main() runs before these classes exist, so they won't be discovered/executed. Move the if __name__ == \"__main__\": unittest.main() block to the end of the file (after all test class definitions), or remove the direct-execution entrypoint if it's not intended to be used.
src/simulation/m_start_up.fpp:1
The probe allocates device memory, launches a GPU kernel, performs a device→host update, and prints at startup unconditionally. Even if intended as temporary, this is user-visible output and adds runtime overhead to all runs (including non-CI). Consider guarding this behind a dedicated runtime option (e.g., eos_probe = T), a compile-time probe flag, or an environment variable check so the diagnostics can be enabled only when needed.
toolchain/mfc/test/coverage.py:1
Scrubbing all GIT_* environment variables globally for every git subprocess is broader than the documented motivating issue (GIT_DIR, GIT_INDEX_FILE). If any callers rely on other GIT_* variables (e.g., GIT_SSH_COMMAND), this could cause surprising behavior. A narrower scrub (explicitly dropping only GIT_DIR/GIT_INDEX_FILE/GIT_WORK_TREE) or an opt-in parameter on _git (e.g., scrub_git_env=True) would preserve the hook fix while reducing unintended side effects.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

 if [ "${GITHUB_EVENT_NAME:-}" = "pull_request" ]; then
-    select_opts="--select-enforce"
+    select_opts="--only 9DAC4DDC 2F35A1FE A421E318 5AC2F65D 29C5D458 34F3999B 3A8359F6"
 fi


-      needs.lint-gate.result == 'success' &&
-      needs.file-changes.result == 'success' &&
-      needs.file-changes.outputs.checkall == 'true'
+    if: false


  case-optimization:
    name: "Case Opt | ${{ matrix.cluster_name }} (${{ matrix.device }}-${{ matrix.interface }})"
-    if: github.repository == 'MFlowCode/MFC' && needs.file-changes.outputs.checkall == 'true' && github.event.pull_request.draft != true
+    if: false


github-actions · 2026-09-03T17:52:52Z

Claude Code Review

Head SHA: 6269824

Files changed:

164
.claude/rules/common-pitfalls.md
.github/scripts/check_coverage_map_health.py
.github/workflows/common/test.sh
.github/workflows/test.yml
docs/documentation/case.md
docs/documentation/contributing.md
src/common/m_variables_conversion.fpp
src/simulation/m_pressure_relaxation.fpp
src/simulation/m_start_up.fpp
toolchain/mfc/case_validator.py

Findings:

.github/workflows/test.yml disables the primary CI matrix: the github job (Ubuntu/macOS, gfortran/Intel) is set to if: false, the case-optimization job is set to if: false, and the self-hosted cluster matrix is gutted (Phoenix and most Frontier CCE/AMD shards removed, leaving one Frontier ACC entry, plus runs-on: group: phoenix and the NODE_OPTIONS env are dropped). If merged, this PR turns off the gfortran/Intel compiler gate and case-optimization coverage entirely, directly contradicting the "must compile with gfortran, nvfortran, Cray ftn, and Intel ifx" requirement — this looks like a debugging leftover, not an intended change.
.github/workflows/common/test.sh replaces --select-enforce (coverage-based PR test selection) with a hardcoded --only list of 7 test UUIDs for every pull_request event. Once merged, all future PRs on this branch would run only those 7 tests in CI instead of the normal coverage-selected set, silently gutting regression coverage repo-wide.
src/simulation/m_start_up.fpp (s_initialize_gpu_vars, ~line 1103-1141): a new unconditional "CI probe" block allocates probe, runs a GPU kernel and host-side calls to f_bulk_modulus/s_phase_coefficients/etc., and unconditionally prints eos probe device/eos probe host to stdout on every simulation run (rank 0). This has no debug/feature guard and runs in production for every case, adding stdout noise and extra device work — looks like leftover development scaffolding that should be a unit test, not part of s_initialize_gpu_vars.
src/simulation/m_pressure_relaxation.fpp (~line 3107, s_correct_internal_energies init loop): the pressure floor clamp on pres_K_init(i) is now gated by if (.not. any_state_dependent_eos), the case-global flag, whereas the equivalent clamp on pres_relax a few lines later (~line 3120) is correctly gated per-fluid by if (.not. f_is_state_dependent(i)). In a mixed-EOS case (e.g. one stiffened-gas fluid plus one JWL/Mie-Gruneisen/Vinet fluid), any_state_dependent_eos is true globally, so the stiffened-gas fluid's pres_K_init clamp is skipped even though that fluid still uses the stiffened-gas closed form (isentrope_B/isentrope_n) downstream in f_phase_pressure_on_isentrope. An unclamped, near-vacuum initial pressure for the SG phase can then drive (pres + isentrope_B(i))**isentrope_n(i) to a negative base with non-integer exponent inside the isentrope evaluation, producing NaNs for the SG fluid in a mixed-EOS case.

…isen-true

…NaN-safe harness, coverage for the temperature, IBM and cubic paths; EOS helpers as subroutines

…a device-array element passed by reference

github-actions · 2026-09-04T07:31:57Z

Lines of Code

File	Lines	Diff
`src/common/m_variables_conversion.fpp`	1411	+283
`src/post_process/m_derived_variables.fpp`	355	-27
`src/post_process/m_global_parameters.fpp`	380	+22
`src/pre_process/m_global_parameters.fpp`	450	+21
`src/simulation/m_global_parameters.fpp`	769	+21
`src/common/m_derived_types.fpp`	460	+20
`src/post_process/m_start_up.fpp`	786	+20
`src/simulation/m_pressure_relaxation.fpp`	170	+18
`src/simulation/m_acoustic_src.fpp`	515	+15
`src/simulation/m_riemann_solver_hllc.fpp`	1278	+10
`src/common/m_global_parameters_common.fpp`	247	+8
`src/simulation/m_hypoelastic.fpp`	597	+6
`src/simulation/m_riemann_solver_hypo_hlld.fpp`	782	+6
`src/common/m_constants.fpp`	92	+5
`src/pre_process/m_data_output.fpp`	624	-5
`src/simulation/m_rhs.fpp`	1956	+5
`src/simulation/m_data_output.fpp`	1308	-2
`src/simulation/m_ibm.fpp`	1253	+2
`src/simulation/m_start_up.fpp`	1247	+2
`src/simulation/m_time_steppers.fpp`	829	+2
`src/post_process/m_data_output.fpp`	1183	+1
`src/simulation/m_cbc.fpp`	1115	+1

Directory	Lines	Diff
common	10262	+316
pre_process	4513	+16
simulation	27754	+86
post_process	3343	+16
total	45872	+434

…binary

codecov · 2026-09-04T16:25:34Z

Codecov Report

❌ Patch coverage is 88.94737% with 42 lines in your changes missing coverage. Please review.
✅ Project coverage is 62.70%. Comparing base (8010057) to head (20ef9a3).
⚠️ Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
src/common/m_variables_conversion.fpp	94.27%	3 Missing and 8 partials ⚠️
src/post_process/m_start_up.fpp	25.00%	8 Missing and 1 partial ⚠️
src/simulation/m_acoustic_src.fpp	10.00%	8 Missing and 1 partial ⚠️
src/simulation/m_pressure_relaxation.fpp	75.00%	1 Missing and 5 partials ⚠️
src/post_process/m_data_output.fpp	0.00%	3 Missing ⚠️
src/simulation/m_data_output.fpp	81.81%	2 Missing ⚠️
src/simulation/m_riemann_solver_hllc.fpp	85.71%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1812      +/-   ##
==========================================
+ Coverage   62.26%   62.70%   +0.44%     
==========================================
  Files          84       84              
  Lines       21558    21847     +289     
  Branches     3188     3199      +11     
==========================================
+ Hits        13423    13700     +277     
- Misses       5937     5938       +1     
- Partials     2198     2209      +11

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

sbryngelson · 2026-09-04T22:16:39Z

Closing: this probe has served its purpose. It answered two questions that the full suite could not, because both needed one Frontier lane running one test repeatedly.

Cray OpenACC. Three cycles narrowed the NaNs in the EOS helpers to an array element passed by reference into a device routine containing a loop. Recorded in #1815 and reproduced minimally in sbryngelson/compiler-bugs.

Cray OpenMP. Repointed at the reacting Roe HLLC shocktube, three passes of one binary gave three different answers:

0.07593517165078
0.07593517172084
0.07593517170526

That is a data race, not compiler noise. The cause was c_sum_Yi_Phi, written inside the HLLC kernel and missing from the pure-fluid private list while present in the hypoelastic one. It predates #1811 and is fixed on master in #1823.

No code from this branch is intended for merge.

sbryngelson added 10 commits September 2, 2026 16:19

Delete the dead fluid-1 EOS shortcuts in the output routines

2bf3f0d

gamma = gammas(1) and friends were assigned and never read; lit_gamma was assigned in both output routines and read in neither.

toolchain: keep coverage git calls out of the hook's GIT_DIR so prech…

f11722b

…eck passes from worktrees (cherry picked from commit 5ad5c85)

Add JWL and evaluate state-dependent EOS through Wood's law, characte…

635b00e

…ristic BCs, probes and post-process c

Merge remote-tracking branch 'upstream/master' into feature/mie-grune…

636b5c7

…isen-jwl

Copilot AI lite review requested due to automatic review settings September 3, 2026 17:04

Copilot AI reviewed Sep 3, 2026

View reviewed changes

Copilot started reviewing on behalf of sbryngelson September 3, 2026 17:40 View session

sbryngelson added 4 commits September 3, 2026 12:56

Let the convergence harness build the per-case configurations it runs

3fcd291

Read device-resident arrays directly inside the EOS device routines

4fa3d78

Merge remote-tracking branch 'upstream/master' into feature/mie-grune…

e644cf7

…isen-true

Address review: star-state coefficients, per-fluid relaxation floor, …

c167a0a

…NaN-safe harness, coverage for the temperature, IBM and cubic paths; EOS helpers as subroutines

sbryngelson force-pushed the probe/cce-acc branch from 5c0069f to 2829c41 Compare September 3, 2026 22:59

sbryngelson added 3 commits September 3, 2026 21:42

Pass scalars into the EOS device routines: Cray OpenACC misaddresses …

1ee119b

…a device-array element passed by reference

ci: keep build/venv when the build retry nukes the build directory

c35488b

CI probe: one Frontier CCE OpenACC lane, eight targeted tests

d7c2127

sbryngelson force-pushed the probe/cce-acc branch from 1c7b22f to d7c2127 Compare September 4, 2026 05:19

CI probe: CCE OpenMP lane, the reacting shocktube three times on one …

20ef9a3

…binary

sbryngelson closed this Sep 4, 2026

Navigation Menu

Uh oh!

[CI probe, do not merge] CCE OpenACC EOS helper diagnostics - #1812

[CI probe, do not merge] CCE OpenACC EOS helper diagnostics#1812
sbryngelson wants to merge 18 commits into
MFlowCode:masterfrom
sbryngelson:probe/cce-acc

sbryngelson commented Sep 3, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

github-actions Bot commented Sep 3, 2026

Uh oh!

github-actions Bot commented Sep 4, 2026

Uh oh!

codecov Bot commented Sep 4, 2026

Uh oh!

sbryngelson commented Sep 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

sbryngelson commented Sep 3, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

github-actions Bot commented Sep 3, 2026

Uh oh!

github-actions Bot commented Sep 4, 2026

Lines of Code

Uh oh!

codecov Bot commented Sep 4, 2026

Codecov Report

Uh oh!

sbryngelson commented Sep 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants