CGH0S7 c974a88d6d Pool fh work arrays in compute_rhs_bssn to eliminate allocation churn ...

Add _fh variants of fderivs, fdderivs, kodis, and lopsided that accept
a caller-provided fh work array instead of allocating one internally.
Declare two shared work arrays in compute_rhs_bssn (fh_work2 for
symmetry_bd(2) callers, fh_work3 for symmetry_bd(3) callers) and pass
them to all ~84 subroutine calls, eliminating ~77 redundant automatic
array allocations (~591 MB churn per RHS call, ~2.3 GB per timestep).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-02-07 21:49:12 +08:00

19 KiB

Raw Blame History

View Raw