Builds with RRTMGP will fail when using the GNU and PGI compilers, on both summit and cori. The problem seems to be due to inconsistent treatment of the nswbands and nlwbands sizes. These are set in radconstants.F90 as integer parameters (because other parts of the code, I think places in chemistry, depended on them), but I previously tried to allow them to be set dynamically in radiation.F90 by reading the absorption coefficient data at initialization (in case we ever wanted to change the number of bands used). This causes some cryptic runtime errors with GNU and PGI though it seems, so we should revert to setting these in radconstants.F90 for now. PR with the fix coming shortly.
Builds with RRTMGP will fail when using the GNU and PGI compilers, on both summit and cori. The problem seems to be due to inconsistent treatment of the
nswbandsandnlwbandssizes. These are set inradconstants.F90as integer parameters (because other parts of the code, I think places in chemistry, depended on them), but I previously tried to allow them to be set dynamically inradiation.F90by reading the absorption coefficient data at initialization (in case we ever wanted to change the number of bands used). This causes some cryptic runtime errors with GNU and PGI though it seems, so we should revert to setting these inradconstants.F90for now. PR with the fix coming shortly.