[gmx-developers] some notes on compiling the current gmx cvs sources on cray xt3

David van der Spoel spoel at xray.bmc.uu.se
Wed May 17 10:18:42 CEST 2006


Axel Kohlmeyer wrote:
> hi everybody,
> 
> im currently trying to get the current gmx development cvs sources
> compiled on a cray xt3 (basically an opteron cluster with an embedded
> kernel and a special low-lentency network).
> the catch is, that one has to basically cross-compile, which only
> seems to work, when i edit the configure file to default to 
> cross-compilation.
> otherwise ./configure will not finish and create the makefiles etc...
> 
> while trying several options, i noticed that a) you cannot currently 
> compile
> serially and b) there are a number of c++ style comments in the c-code
> that have PGI's pgcc choke heavily (gcc is more forgiving). a patch that
> attempts to fix both problems is attached.

I have fixed the // problems. I don't understand the remark about 
compiling serially.

> 
> as already reported by shawn brown elsewhere, the gcc compiled code
> tends to crash
> in different places (e.g. segfault in add_gbond(), mpi error in 
> splitter.c).

This is weird, the add_gbond error might depend on the system studied 
however.

> 
> the PGI compiled binary seems to work, however, and for the DPPC
> benchmark input it has a speed increase even for large numbers of
> nodes (just with standard settings and no i/o buffer hacks etc, that
> should help the code to scale even more on this machine). the same
> code checkout with the same optimization compiled on a dual-opteron
> myrinet cluster running the identical input scales out at 8 processors...
> 
Your numbers look good for the ring parallellization scheme that we have 
used until now. It will quite soon be possible to obtain even better 
scaling using domain decomposition, Berk has been working very hard to 
implement it. The DPPC benchmark scales to 32 Opteron cores on my Gbit 
network already, so it will be interesting to see whether it will be 
even better on the Cray.

-- 
David.
________________________________________________________________________
David van der Spoel, PhD, Assoc. Prof., Molecular Biophysics group,
Dept. of Cell and Molecular Biology, Uppsala University.
Husargatan 3, Box 596,  	75124 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://folding.bmc.uu.se
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++



More information about the gromacs.org_gmx-developers mailing list