Fill ptil so probe output stops writing undefined ptilde and ptot

[sbryngelson]

Description

ptil is allocated in m_global_parameters.fpp:929 and pushed to the device, but nothing in the codebase ever assigns it. The probe path reads it anyway:

• m_data_output.fpp:1300 ptilde = ptil(j - 2, k, l)
• m_data_output.fpp:1301 ptot = pres - ptilde
• m_data_output.fpp:1471 both are passed through s_mpi_allreduce_sum
• m_data_output.fpp:1509 both are written to the probe file

So 1D Euler-Euler bubble runs with probe_wrt = T and num_fluids = 3 have been writing undefined values in the ptilde and ptot columns. Because the values are MPI-reduced first, they are not reproducible across runs or rank counts. Fixes #1741.

Changes

s_compute_ptilde in m_bubbles_EE.fpp evaluates the correction at cell centers, mirroring what the HLLC solver applies to reconstructed face states in m_riemann_solver_hllc.fpp:736-746:

ptil = alpha (p_l - <R^3 p_bw>/<R^3> - rho <R^3 Rdot^2>/<R^3>)

which is Eq. (2.5) of Ando (Caltech thesis, 2010). The small_alf degenerate branch and the QBMM moment path (mom_sp 1, 3, 4) are both mirrored so the diagnostic cannot drift from the solver.

It is called from s_compute_rhs rather than from s_compute_bubble_EE_source. The latter is guarded by (.not. adap_dt) .and. (.not. qbmm) at m_rhs.fpp:839, so hanging the fix off it would have left ptil unfilled in exactly the QBMM and adaptive-timestep cases.

m_start_up.fpp also gains a GPU_UPDATE(host='[ptil]') next to the existing q_cons sync in the probe_wrt block. Without it the array stayed device-resident and the host-side probe read would still have been invalid on GPU builds.

Type of change

• Bug fix (non-breaking change which fixes an issue)

Scope

• This PR comprises a set of related changes with a common goal

Testing

Built simulation locally and ran examples/1D_bubblescreen (bubbles_euler, probe_wrt) with temporary instrumentation on ptil. Values are finite and evolve smoothly. The magnitudes check out against the case setup: the background patch has alpha = 1e-12, below small_alf = 1e-11, so it takes the degenerate branch and gives ptil = alpha * p of about 1e-12, which matches the observed maximum. The bubble screen at vf0 = 4e-5 contributes near zero because the bubbles start in equilibrium, so p_l - p_bw and Rdot both vanish. The instrumentation was removed before commit.

./mfc.sh precheck passes.

ptil is diagnostic only and never feeds back into the solution, and the new computation is gated on probe_wrt, so no golden values move.

Note on coverage

No example or test case combines 1D, bubbles_euler, probe_wrt, and num_fluids = 3, which is why this went unnoticed. Adding one would be a reasonable follow-up.

[Copilot]

Pull request overview

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.

Fixes undefined ptilde/ptot probe outputs in Euler–Euler bubble runs by computing and syncing the ptil diagnostic field before probe writes.

Changes:

• Compute ptil in the RHS path (gated on bubbles_euler + probe_wrt) via new s_compute_ptilde.
• Add GPU host sync for ptil in the probe-related startup sync block.
• Implement cell-centered ptil evaluation mirroring the HLLC solver’s pressure-correction logic (including QBMM and degenerate branches).

Reviewed changes

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

File	Description
src/simulation/m_start_up.fpp	Adds host synchronization of ptil for GPU builds when probes are enabled.
src/simulation/m_rhs.fpp	Calls s_compute_ptilde during RHS evaluation when probe output is enabled.
src/simulation/m_bubbles_EE.fpp	Introduces s_compute_ptilde to fill the probe diagnostic ptil consistently with solver logic.

---

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

[github-actions]

Claude Code Review

Head SHA: 962130b

Files changed:

• 3
• src/simulation/m_bubbles_EE.fpp
• src/simulation/m_rhs.fpp
• src/simulation/m_start_up.fpp

Findings:

• ptil is only ever populated by s_compute_ptilde, which m_rhs.fpp gates on bubbles_euler .and. probe_wrt (src/simulation/m_rhs.fpp:848), but the corresponding GPU_UPDATE(host='[ptil]') added in m_start_up.fpp is gated on bubbles_euler alone, with no probe_wrt check (src/simulation/m_start_up.fpp:624). This host sync runs during startup, before the time-stepping loop (and hence before s_compute_ptilde has ever executed), so if bubbles_euler is on the device-side ptil buffer — which Fortran does not zero-initialize on allocation — gets pulled to host and is available to be written into the initial (t_step 0) probe output even when it was never computed, or before it was computed. The two guards should match (bubbles_euler .and. probe_wrt) to avoid surfacing uninitialized data as a diagnostic value.

[codecov]

Codecov Report

❌ Patch coverage is 74.07407% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 61.61%. Comparing base (d4ffa29) to head (962130b).

Files with missing lines	Patch %	Lines
src/simulation/m_bubbles_EE.fpp	72.00%	4 Missing and 3 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1742      +/-   ##
==========================================
+ Coverage   61.59%   61.61%   +0.01%     
==========================================
  Files          84       84              
  Lines       21493    21520      +27     
  Branches     3176     3180       +4     
==========================================
+ Hits        13239    13259      +20     
- Misses       6078     6082       +4     
- Partials     2176     2179       +3     

☔ 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.