Add OpenMP hybrid parallelism to RRTMGP shortwave block loop#86

Merged

tclune merged 12 commits intofeature/sdrabenh/gcm_v12from

feature/tclune/solar_openmp

May 7, 2026

Commits on May 4, 2026

Refactor(Step 1): uniform per-block alloc/dealloc in RRTMGP block loop
tclune
committed
Refactor(Step 2a): extract compute_gas_optics from RRTMGP block loop
tclune
committed
Refactor(Step 2b): extract compute_aer_optics from RRTMGP block loop
tclune
committed
Step 2c: extract compute_cloud_optics_mcica
tclune
committed
Step 2d: extract compute_sprlyr_diags_predelta
tclune
committed
Step 2e: extract compute_delta_scale
tclune
committed
Step 2f: extract compute_sprlyr_diags_postdelta from SORADCORE block loop
tclune
committed
Step 2g: extract compute_rte_sw from SORADCORE block loop
tclune
committed
Step 3: assemble PROCESS_RRTMGP_BLOCK wrapper with all per-block state as locals
tclune
committed
Step 4: add ! PARALLEL DO over block loop; comment out non-thread-safe MAPL timers
tclune
committed
Fix: move aer_props increment to call site to avoid passing unallocated polymorphic dummy
tclune
committed
bugfix3: guard compute_aer_optics call when aer_props unallocated
tclune
committed