Open SeanBryan51 opened 4 months ago
gadi.nci.org.au
pearcey.hpsc.csiro.au
MatthiasCuntz@INRAE
intel-compiler-llvm/2023.0.0
netcdf/4.9.2
-fpp -O3 -nofixed -assume byterecl -fp-model precise -ip -diag-disable=10382,15009 -march=broadwell -axSKYLAKE-AVX512,CASCADELAKE,SAPPHIRERAPIDS -D__INTEL__ -D__INTEL_COMPILER__ -D__NETCDF3__
host_mcin
) differences:
-fpp -O3 -nofixed -assume byterecl -fp-model precise -ip -diag-disable=10382 -xHost -D__INTEL__ -D__INTEL_COMPILER__ -D__CRU2017__ -D__NETCDF3__ -D__C13DEBUG__
-fpp -O0 -debug extended -traceback -g -check all,noarg_temp_created -warn all -fp-stack-check -nofixed -assume byterecl -fp-model precise -diag-disable=10382 -fpe0" # -fpe-all=0 -no-ftz -ftrapuv -init=arrays,snan -D__INTEL__ -D__INTEL_COMPILER__ -D__CRU2017__ -D__NETCDF3__ -D__C13DEBUG__
-cpp -O3 -Wno-aggressive-loop-optimizations -ffree-form -ffixed-line-length-132 -march=native -D__GFORTRAN__ -D__gFortran__ -D__CRU2017__ -D__NETCDF3__ -D__C13DEBUG__
-cpp -O -g -pedantic-errors -Wall -W -Wno-maybe-uninitialized -ffree-form -ffixed-line-length-132 -fbacktrace -ffpe-trap=zero,overflow -finit-real=nan" # -ffpe-trap=zero,overflow,underflow -D__GFORTRAN__ -D__gFortran__ -D__CRU2017__ -D__NETCDF3__ -D__C13DEBUG__
-O4 -fpp -colour -unsharedf95 -ideclient -ieee=full -free -not_openmp -mismatch -mdir . -march=native -D__NAG__ -ideclient -unsharedrts -D__CRU2017__ -D__NETCDF3__ -D__C13DEBUG__
-C=alias -C=array -C=bits -C=dangling -C=do -C=intovf -C=present -C=pointer -C=recursion -g -nan -O0 -strict95 -gline -fpp -colour -unsharedf95 -ideclient -ieee=full -free -not_openmp -mismatch -mdir . -march=native -D__NAG__ -ideclient -unsharedrts -D__CRU2017__ -D__NETCDF3__ -D__C13DEBUG__
host_mc16
) differences:
host_mcmi
) differences:
host_vm_o
) differences:
host_logi
) differences:
host_auro
) differences:
host_zlle
) differences:
ld
flags: -Wl,-Bsymbolic-functions -flto=auto -ffat-lto-objects -Wl,-z,relro -Wl,-z,now
intel-mpi/2021.5.1
mpiifort
netcdf/4.8.0
-fpp -O3 -nofixed -assume byterecl -fp-model precise -ip -diag-disable=10382 -xCASCADELAKE -xCORE-AVX2 -axSKYLAKE-AVX512,CASCADELAKE -xBROADWELL -xSKYLAKE -D__INTEL__ -D__INTEL_COMPILER__ -D__MPI__ -D__CRU2020__ -D__NETCDF3__
-fpp -O0 -debug extended -traceback -g -check all,noarg_temp_created -warn all -fp-stack-check -nofixed -assume byterecl -fp-model precise -diag-disable=10382 -fpe0 -fpe-all=0 -no-ftz -ftrapuv -D__INTEL__ -D__INTEL_COMPILER__ -D__MPI__ -D__CRU2020__ -D__NETCDF3__
I haven't used the new build system yet. With "CABLE4" to date I have been modifying the CABLE3-like build/Makefile (albeit the one from the CABLE-POP-TRENDY side [so it is very similar to the CABLE3 build3.sh (and Makefile) version - just older]). My estimation was that upon completion the Makefile etc would look just like the CABLE3 ones you have just done. BUT the CABLE4 I "was" doing has no overlap with offline applications so will need some extra bits to build with POP offline.
WRT to all the user/machine profiles. In the beginning we tried to cater for everyone. We then went to a self writing build script to include new profiles. It just got really messy. In the end we landed on:
"We support the use of CABLE (corresponding builds etc) at NCI. We are happy to help with alternatives, but really this is on you. "
@JhanSrbinovsky in the case for compiling on Gadi, the Makefile build in CABLE-POP_TRENDY branch differs in compiler flags for debug/release configurations as well as module versions for dependencies. We should determine the differences that should be incorporated/discarded when we do the merge of CABLE-POP_TRENDY into main.
Context:
This issue thread documents the differences in build configurations between CABLE-POP_TRENDY and main.
Useful links