pyMPI originally was built to work with MPICH 1. It was never designed to work with MPICH 2, but in most cases appears to run ok. Last conversation on this topic from Pat Miller (creator of pyMPI) was in late 2006:
"I haven't build pyMPI with MPICH2 though I had expected that it would work. I have a parallel Fedora machine at home so I can try this over the weekend and let you know if I can repeat your results.
I have not talked to Pat Miller for some time and he has not made changes to pyMPI since he left LLNL. Your first error could probably be found by using strace (assuming you are on Linux, you didn't say). The 14hr error looks like it might be due to a memory leak - which could easily be in MPICH itself. The only way to prove that is to run a similar long-run C based MPICH2 process. You can actually attach to a running python process and debug it - you just need the right development tool.
----- Original Message ----
From: Ricardo Mata <ricardo.mata@...>
Sent: Friday, October 3, 2008 6:30:40 AM
Subject: [Pympi-users] make check fails
I had two questions, although the answer to the first one could help
on the second. I know the project hasn't been very active, but I would
really appreciate any help I could get...
I have recently downloaded pyMPI and installed it on an Intel Quad
Core, with mpich2-1.0.7 on the machine. After installation, I tried to
run "make check", but it just hangs on the "pyMPI_request_001" test,
consuming 100% on one of the processors. So, the first question would
be: where could this be coming from?
Now the second question. Desperate as I was to get my project going, I
used pyMPI on my python program, which worked out fine. Its the call
for an external program, read information, do something, rinse and
repeat kind of thing. So I only used gather and scatter commands for
the communication. However, it always gets stuck after 14 hours or so.
Up till then it does its job (correctly) for about 300 times. Is there
any way to debug this? Or any clues on what could be causing this?
(note: it hangs with four pyMPI processes running close to 100% on
each processor, but with nothing happening. It has hung twice, each
time at a different location).
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
Pympi-users mailing list