I had trouble building paramesh3.3beta with real8. Apparently the
macro expansion #define MPI_REAL MPI_DOUBLE_PRECISION in
paramesh_preprocessor.fh corrupts a later included mpif.h, because
all source files include paramesh_preprocessor.fh before mpif.h.
Type definitions from mpif.h get macro-expanded, i.e. MPI_REAL to
MPI_DOUBLE_PRECISION. This of course, cannot work. I fixed it
temporarily with commenting out all MPI_REAL in mpif.h. This is fine
for me, because I don't use the single precision for now. I am using
lam-7.1.1 with g95 on OSX.
Log in to post a comment.