In the past, I have been running all my MPI-test runs on a alpha debian woody. About a week ago I started getting these error messages:
--snip--
Fatal error; unknown error handler
May be MPI call before MPI_INIT. Error message is MPI_BARRIER and code is 197
Fatal error; unknown error handler
May be MPI call before MPI_INIT. Error message is MPI_BARRIER and code is 197
p0_6838: p4_error: interrupt SIGSEGV: 11
Killed by signal 2.
Killed by signal 2.
Killed by signal 2.
bm_list_6839: p4_error: interrupt SIGINT: 2
--snip--
The code is exactly the same as I have used before (I simply changed a the initial values of some variables) and the error changes to the following if I remove the final call to MPI_FINALIZE(). (Oh, the code is fortran.)
-snip--
Fatal error; unknown error handler
May be MPI call before MPI_INIT. Error message is MPI_BARRIER and code is 197
--snip--
That is, nothing gets sigint'ed or sigsegv'ed any more! What worries me
most is that there are no calls to MPI_BARRIER() in the code! Of course,
MPI_FINALIZE() may call it itself. The system is running woody (last
updated at exactly the same time this problem appeared) with mpich
1.2.1-16 and g77-3.0 3.0.4-7, which I use to compile the code. Changing to 2.95-4 does not help.
Anyone any insights?
--
-----------------------------------------------
| Juha Jäykkä, juolja@utu.fi |
| home: http://www.utu.fi/~juolja/ |
-----------------------------------------------
Attachment:
pgp5qncg3Ek4y.pgp
Description: PGP signature