Is it possible to parallelize running gulp on a cluster using a scheduler? Currently gulp seems to only use 1 CPU per task.
If you select the MPI option during compilation then it will run on as many cores as you have (subject to scaling tests to check efficiency)
The compilation with mpi seems to be working fine for the first steps with mpif90 until it reaches the linker step with the .o files. When it reaches there it keeps saying that there’s an undefined reference to many functions like “dcopy_” and “lsame_” etc… I’m not sure what this means as it seems to be finding “libscalapack.a” correctly.
You are correct - the issue is that scalapack, blas etc are not being linked on your machine. The default is to compile and link with gfortran and scalapack installed from Macports on a Mac. If you specify the Intel compiler then it should also be OK (depending on version). For other cases you will probably have to modify mkgulp where SLIB/MLIB is specified to the location of scalapack, blacs, pblas and blas on your computer.