I would like to use MPI for CAMB with intel compiler mpifort since I have to launch the executable "camb" multiple times in a loop.
I saw in the Makefile this part where I modified F90C and FFLAGS (keeping -omp to mention multi-processor) :
Code: Select all
#Digital/Compaq fortran, -omp toggles multi-processor #F90C = f90 F90C = mpifort #FFLAGS = -omp -O4 -arch host -math_library fast -tune host -fpe1 FFLAGS = -omp
Unfortunately, at the execution by doing :
mpirun -np 64 ./camb params.ini
CAMB seems to not be distributed on multi-processors, runtime is even higher than for a classical single processor execution.
Does anyone know how to make it run with MPI ? (especially the different flags to set in Makefile) ?