From: schuelaw@carrot.whitman.edu (Albert Schueller) Newsgroups: comp.parallel.mpi Subject: A real puzzler? Date: Thu, 15 Jul 1999 21:03:52 GMT Organization: Posted via RemarQ, http://www.remarQ.com - The Internet's Discussion Network Message-Id: Reply-To: schuelaw@whitman.edu User-Agent: slrn/0.9.5.4 (UNIX) Xref: ukc comp.parallel.mpi:5331 Hi all, I have a simple fortran program that uses mpi. I start the communicator, call a subroutine that has some parallelization in it, shut down the communicator and stop the program. As I was modifying the program, it started throwing cores at the mpi_finalize statment, which immediately follows the one subroutine call. I put a print*, myid after the subroutine, but before the mpi_finalize and only proc 0 prints out its id (while ALL the procs SHOULD print this), which suggests only proc 0 is getting returned from the subroutine. However, if I also put a print*, myid immediately before the RETURN statement in the aforementioned subroutine, all of the procs sucessfully print their ids. This indicates that all of the procs are getting to the return statement in the subroutine, but only proc 0 is actually getting back to the main program. Why are the other procs not getting back to the calling program? I should also point out that the program is running properly, except for the core dump at mpi_finalize. Clues? Thanks, Albert -- Albert Schueller Department of Mathematics Office Phone: 509-522-4432 Whitman College Public Key: http://carrot.whitman.edu/pgp.key Walla Walla, WA USA 99362