Single-column, SHOC-only ("in-and-out") DYCOMS RF01, dt = 60 s, 360 substeps (6 h), exp2 (one shoc_main call per substep, nadv=1). Both models read the same high-precision ShocInOut_IC_*.txt files. E3SM: branch maint-3.0-scm-mchinita-shoc_in_and_out_cpp (thv-staleness fix), Intel, l_turb_standalone=.true. EAMxx C++: standalone driver shoc_in_and_out (same branch), GNU, double precision, thv fix ported into shoc_main.
Top rows: profiles at t = 1/61/181/360 min, E3SM Fortran (solid) vs C++ (black dashed). Bottom rows: column-max relative difference vs time (bounded at ~1e-6..1e-5; attributable to Intel-vs-GNU transcendental rounding — with the same compiler the two codes agree to 1e-14..1e-11).
Runs: /global/cfs/projectdirs/m4359/mchinita/in_and_out_sims_thvfix (Fortran, E3SM case dirs without build/), /global/cfs/projectdirs/m4359/mchinita/in_and_out_cpp_runs (C++ driver outputs + IC files + plotting script). 2026-07-03, Maria Chinita + Claude Fable 5.
Related: exp1 vs exp2 equivalence page exp1 vs exp2 equivalence page.
#124; EAM vs EAMxx interface audit.
