[Bioclusters] mpi library problem

[andy law (RI)]

All,

The answer is probably out there, but pointers to the answer will save me time.

We are setting up a test cluster (of 2 machines) as a pilot to more extensive work on our 30 CPU cluster. The software to be compiled/tested requires mpi and pgapack. We are running on Red Hat 7.x.

One of my troops compiled and installed mpich 1.2.5 (he is off on leave for a couple of weeks and I don't have immediate access to his notes). I know (because we spoke about it efore he went off) that we had problems with the test suite initially, but once we worked out that we could circumvent the race condition inherent in the test suite by installing on an NFS mount then everything went fine. Thus, we have mpich in /local/Mpich/current (softlink to /local/Mpich/mpich-1.2.5) on the two nodes and 'make test' is as happy as it could be.

I installed pgapack. I gave it the path to libmpich.a that it was looking for, and the path to the includes directory. It was happy with 'make' and 'make test'. However, the end bit of 'make' suggested running a script called 'instverf' (installation verification) and although that seemed to run OK, it did mention that there were errors in each of the tests (although it reported them in a 'by the way there were some errors' rather than a 'STOP!!!!! THRERE ARE ERRORS' kind of way).

Now we are trying to compile the *real* code, we are getting a failure at the link stage, with reference to a whole load of mpi_* routines (amongst others). We have the necessary -L/local/Mpich/current/lib and -lmpich flags set, but nothing seems to work. objdump on the library suggests that at least some of them are in there (although I am no expert at decoding the output of objdump - it could be that there is a problem in there).

Are there any obvious things that we might have done wrong, or ways that we can track it down further?

Thanks in advance,

Later,

Andy

-------------
Dr. Andy Law
--------------------
Head of Bioinformatics - Roslin Institute

Unfortunately, legal niceties require me to add the following to this message...

The information contained in this e-mail (including any attachments) is confidential and is intended for the use of the addressee only.   The opinions expressed within this e-mail (including any attachments) are the opinions of the sender and do not necessarily constitute those of Roslin Institute (Edinburgh) ("the Institute") unless specifically stated by a sender who is duly authorised to do so on behalf of the Institute.