#!/usr/bin/env python3 import os petsc_hash_pkgs=os.path.join(os.getenv('HOME'),'petsc-hash-pkgs') # This test is done on grind.mcs.anl.gov. It uses ILP64 MKL/BLAS packaged # with MATLAB. # Note: regular BLAS [with 32-bit integers] conflict with # MATLAB BLAS - hence requiring -known-64-bit-blas-indices=1 # Note: MATLAB build requires PETSc shared libraries # Some versions of MATLAB [R2013a] conflicted with -lgfortan - so the following workaround worked. # export LD_PRELOAD=/usr/lib/gcc/x86_64-linux-gnu/4.6/libgfortran.so # find MATLAB location import os from shutil import which matlab_dir=os.path.dirname(os.path.dirname(which('matlab'))) if __name__ == '__main__': import sys import os sys.path.insert(0, os.path.abspath('config')) import configure configure_options = [ '--package-prefix-hash='+petsc_hash_pkgs, '--download-mpich=1', # /usr/bin/mpicc does not resolve '__gcov_merge_add'? and gcc-4.4 gives gcov errors '--with-display=140.221.10.20:0.0', # for MATLAB example with graphics '--with-blaslapack-dir='+matlab_dir, '--with-matlab=1', # matlab-engine is deprecated, no longer needed but still allowed '--with-matlab-engine=1', '--with-shared-libraries=1', '-known-64-bit-blas-indices=1', '--with-ssl=0', '--with-coverage=1', '--with-strict-petscerrorcode', ] configure.petsc_configure(configure_options)