Re: AMBER: Amber8 parallel compilation problems on an OSCAR Cluster

From: Aaron J. Greenwood <agreenwo.uci.edu>
Date: Fri, 03 Feb 2006 07:57:30 -0800

Thank you for taking the time to reply to my message.

There is a new twist you might find interesting. Last night I logged
onto my system from home and started over. Whatever the problem was no
longer existed. Try as I may I could not recreate the errors I had at
work. My guess is that the environment variables may have been setup up
incorrectly.

Best regards


David A. Case wrote:
> On Thu, Feb 02, 2006, Aaron J. Greenwood wrote:
>
>
>> On our system the serial version of amber8 compiles without error but
>> the parallel version does not compile. I've looked but cannot find the
>> missing libraries. I get many messages of the form:
>>
>> egb.o(.text+0x1eeb): In function `genborn_mp_egb_':
>> : undefined reference to `mpi_allreduce_'
>>
>
> We need more information about the statement "I've looked but cannot find
> the missing libraries". What is the LOADLIB variable you have in your
> config.h file? If you go to that location, do you find anything that
> looks like mpi libraries?
>
> Since you did *not* get an error message saying that the libraries were
> missing, (but rather that references inside them were missing) my best guess
> is that you may have installed LAM with a different compiler than you used
> to install Amber. (Is that possible?)
>
> You can use the "nm" program to find out what entries are in a library;
> try something like this:
>
> cd $LAM_HOME/lib
> nm liblamf77mpi.a | grep mpi_allreduce
>
> this will report what "mpi_allreduce" commands are found. (You may have to try
> different library names, depending on which version you have.) If there
> is some problem (e.g. MPI_ALLREDUCE is in caps, or the trailing underscore
> is missing) this points to an installation incompatibility.
>
> If you can't find mpi_allreduce in any LAM library, you will also have to
> look at your configuration.
>
> Many other scenarios are possible, but this is what you must do to debug
> the situation. It is nearly impossible for people on the mailing list
> to figure out all the possible things that might be wrong.
>
> ...good luck...dac
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sat Feb 04 2006 - 06:10:06 PST
Custom Search