- Oct 14, 2024
-
-
RODIER Quentin authored
merge the _OPT_LINEARIZED_LOOP version for PHYEX shuman_phy (bit-repr); TO BE TESTED in AROME for PERF on CPU
-
- Oct 10, 2024
-
-
RODIER Quentin authored
-
RODIER Quentin authored
-
- Oct 09, 2024
-
-
RODIER Quentin authored
Merge branch 'MNH-56X-dev-OPENACC-FFT-PHYEXturb-ice4' of mesonh2.aero.obs-mip.fr:/gitrepos/MNH-git_open_source-lfs into MNH-56X-dev-OPENACC-FFT-PHYEXturb-ice4
-
RODIER Quentin authored
-
RODIER Quentin authored
-
RODIER Quentin authored
-
- Oct 08, 2024
-
-
RODIER Quentin authored
condensation is strongly modified with respect to the AROME version: a duplication is made while new pft fonctions will be written and tested on the future HPC see Issue #29 of pyft repo
-
RODIER Quentin authored
-
- Oct 04, 2024
-
-
RODIER Quentin authored
Quentin 04/10/2024: end OpenACC ice_adjust: must be used with pyft commit 59b7877 (convertTypesInCompute deals with if-then-stmt + handle manualy the transfo of ICED%XRTMIN within a do loop)
-
- Oct 02, 2024
-
-
RODIER Quentin authored
-
RODIER Quentin authored
- remove IC=0 (useless in the IF condition), bit-repro CPU/GPU - remove EXIT of the loop (impossible to use with OpenACC) - add kernels to internal loop with atomic capture of IC (with change of variable IDX)
-
RODIER Quentin authored
Quentin 02/10/2024: OpenACC opti, reverse loops on JIJ/JK to avoid a compiler problem on reduction of ZMAX_TSTEP; parallelized only on NIJT
-
- Sep 30, 2024
-
-
RODIER Quentin authored
-
- Sep 27, 2024
-
-
RODIER Quentin authored
-
RODIER Quentin authored
-
- Sep 26, 2024
-
-
RODIER Quentin authored
Quentin 26/09/2024: add kernels ice4_rainfr_vert, pick the splitted version with the vertical loop on top
-
RODIER Quentin authored
-
RODIER Quentin authored
-
RODIER Quentin authored
-
RODIER Quentin authored
-
RODIER Quentin authored
-
RODIER Quentin authored
Quentin 26/09/2024: bugfix PRCACCR(:) = PHLC_HRC(JL)... for CPU/GPU (already removed in 5.7). This bf changes results
-
RODIER Quentin authored
-
RODIER Quentin authored
-
RODIER Quentin authored
-
- Sep 25, 2024
-
-
RODIER Quentin authored
-
RODIER Quentin authored
-
- Sep 24, 2024
-
-
RODIER Quentin authored
Quentin 24/09/2024: remove useless D1D_TO_3D array-dimensions conversion in GX_M_U (lower perf on GPU)
-
RODIER Quentin authored
Quentin 24/09/2024: add expansion of All Arrays within PHYEX for adding DO CONCURRENT + collapse independent at every mnh_expand
-
RODIER Quentin authored
Quentin 24/09/2024: add back ifdef around mnh_undef(LOOP) and mnh_undef(OPENACC) because the macro is handled by CPP and not pyft
-
- Sep 18, 2024
-
-
RODIER Quentin authored
-
- Sep 13, 2024
-
-
RODIER Quentin authored
Quentin 13/09/2024: Cray re-merge mnh_undef macro with pyft --craybyPassDOCONCURRENT fonction. Use with pyft_tool commit of 13/09
-
- Sep 02, 2024
-
-
Juan Escobar authored
Juan 02/09/2024:mode_turb_hor_dyn_corr.f90, comment completely "!$acc ! wait" not accepted by Nivida compiler
-
- Aug 26, 2024
-
-
ESCOBAR MUNOZ Juan authored
-
- Aug 19, 2024
-
-
ESCOBAR MUNOZ Juan authored
-
ESCOBAR MUNOZ Juan authored
Juan 19/08/2024:MESONH_HOST2005/run_mesonh_step1_xyz_mppdb, Update for use with rocm/Cray-MPI/SPAWN environnement
-
ESCOBAR MUNOZ Juan authored
Juan 19/08/2024:mode_turb_hor_thermo_flux.f90, For Cray compiler, WERY LONH =1hour compilation time, try to split the code without succes
-
ESCOBAR MUNOZ Juan authored
Juan 19/08/2024:mode_turb_hor_dyn_corr.f90, For Cray compiler, deactived BUGGED asynchronus kernel, due to SHUMAN automatic transformation by pyft_tool '
-
ESCOBAR MUNOZ Juan authored
Juan 19/08/2024:mode_tridiag_tke.f90/turb.f90 For Cray compiler & managed memory add present_crm/update_crm , to bypass PB "data not present"
-