From: "Thomas A. Spraggins" <tas@holmes.acc.Virginia.EDU>
Newsgroups: comp.parallel.mpi
Subject: Problems testing MPICH 1.1.0 on DEC Alpha
Date: Tue, 25 Aug 1998 08:08:59 -0400
Organization: University of Virginia
Message-Id: <Pine.A32.3.95.980825080228.129566B-100000@holmes.acc.Virginia.EDU>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


I installed MPICH on a DEC Alpha running OSF 4.0 revision 878 and Fortran
5.1, specifying -device=ch_p4 and -arch=alpha.  The simple fpi and cpi
test programs seem to run fine.

I suspected a problem, however, since my attempts to port "Amber" (a
computational chemistry package) failed, while someone at DEC was able to
port it with no problems.  The only difference we can discern is that he's
using an older MPICH.

When I try to run the more thorough MPICH tests, and look at the "diff's",
I see things like:

------------------------------------------------------------------

**** Testing MPI Point-to-point routines ****
alpha10.bioch.virginia.edu : Tue Aug 25 07:43:24 EDT 1998
**** Testing MPI_Send and MPI_Recv ****
Differences in sendrecv.out
2,9c2,275
< mpirun: M-^A^C^A^KM-@: not found
< mpirun: ^W8M-TM-c^P^C^HM-Bb^HM-^@M-^P^A^BM-^@^D^A@
< ^A@M-^M^A@M-^QM-^\#M-
@0^A@0^A^AM-^]M-cM-?h/^HM-.^UM-b(^Q^HM-^P^R^L@KM-^R^P:
not found
< mpirun: M-B^CM- dM-b^CM- hM-$^E^AM-^R^DM-^@^Q/^HM-.^UM-bHM-^P^P: not
found
< mpirun: M-^AM-GM-`^HM-^AM-h^PM-^@^FM-^B^P: not found
< mpirun: M-^@M- a^K^RM-^@K^AM-h^CM- dM-B#M- M-^PM-^P^P: not found
< mpirun: M-^@^D^AM-^AM-CM-`^H^AM-^P^P: not found
< mpirun: syntax error at line 5: `&' unexpected
---
> [sendrecv]: Test 'Send-Receive Test, Type 0 (MPI_CHAR)' Passed.
> [sendrecv]: Test 'Send-Receive Test, Type 1 (MPI_SHORT)' Passed.
> [sendrecv]: Test 'Send-Receive Test, Type 2 (MPI_INT)' Passed.
> [sendrecv]: Test 'Send-Receive Test, Type 3 (MPI_LONG)' Passed.
> [sendrecv]: Test 'Send-Receive Test, Type 4 (MPI_UNSIGNED_CHAR)' Passed.
> [sendrecv]: Test 'Send-Receive Test, Type 5 (MPI_UNSIGNED_SHORT)'
Passed.
> [sendrecv]: Test 'Send-Receive Test, Type 6 (MPI_UNSIGNED)' Passed.

<snip>

---------------------------------------------------

Has anyone else seen anything like this?  Is there a workaround?

Thanks in advance for any help.

Tom Spraggins
tas@virginia.edu