I will take responsibility for this. I was testing the code for Carlos, but
I forgot that the test Makefile overwrites the TESTsander variable for the
parallel tests. So when I ran the parallel tests I wasn't testing what I set
TESTsander to, I was testing the default ($(AMBERHOME)/exe/sander.MPI). As a
bonus the error doesn't show up when the fdist subroutine isnt called (i.e
REM runs etc with 1 thread per group still passed the parallel tests). Sorry
everyone.
Can anyone suggest an easy way to modify the parallel tests so that you can
set a variable to point to the binary you want to test (like TESTsander in
the serial tests). I suppose you could have a SANDERMPI variable in the
makefile or something that TESTsander gets set to, like
export TESTsander=$(SANDERMPI)
and SANDERMPI can be set to $(AMBERHOME)/exe/sander.MPI if empty?
-Dan
On Fri, Mar 5, 2010 at 9:51 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
> Dear All,
>
> The AMBER cvs tree is currently completely hosed in parallel. All test
> cases
> segfault right now. Thus I have locked the CVS tree until I can work out
> why.
>
> The first culprit is the checkin around 4pm today:
>
> Log Message:
> qtw: Add PM3MM*ESCF for PM3/MM*.
>
> This contains code that cannot possibly work in parallel due to bcasts of
> unallocated arrays etc. I will be reversing out this commit.
>
> However, the tree prior to this commit also segfaults I suspect because of
> this commit:
>
> Modified Files:
> dynlib.f egb.f ene.f ene_decomp.h ew_directe.h ew_force.f
> force.f mdread.f parallel.f pimd_force.f pimd_init.f
> pimd_vars.f runmd.f sander.f short_ene.f
> Log Message:
> update from carlos for NEB changes
>
> I will likely reverse this out as well to get us back to a working tree.
>
> I'll send a message to the list once the tree is open for commits again.
>
> All the best
> Ross
>
>
> /\
> \/
> |\oss Walker
>
> | Assistant Research Professor |
> | San Diego Supercomputer Center |
> | Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
> | http://www.rosswalker.co.uk | http://www.wmd-lab.org/ |
>
> Note: Electronic Mail is not secure, has no guarantee of delivery, may not
> be read every day, and should not be used for urgent or sensitive issues.
>
>
>
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Mar 05 2010 - 21:00:03 PST
