GM_BOGUS_SEND_INTERRUPT
"Davis, Harvey" <[email protected]>
| Newsgroups | gmane.network.myrinet.general |
|---|---|
| Message-ID | <DB0C881A51F5D311999F0090274DD43702E56FB9@salex003.dsto.defence.gov.au> |
Hello I was wondering if anyone can tell me why I am getting a GM_BOGUS_SEND_INTERRUPT. I have code that runs fine when spawned across a single machine (irregardless of the number of processess), but that fails when spawned across multiple machines. The error reported is send failed to complete, host unreachable. The hosts are reachable and the required processess are running on them. Sometimes error messages are written to /var/log/messages complaing about bogus pages. I found the error message generated in the $(GM_HOME)/drivers/gm.c file when the send receives a bogus interrupt. Could threading be doing this?? All my threads are spawned before a call to MPI init and only one thread access the mpi routines. Does anyone know what else might cause this error. Help please!!!!!!!!!!!