[gmx-developers] cray xt3 port, fio rename, misc fixes

David van der Spoel spoel at xray.bmc.uu.se
Wed Aug 30 11:52:30 CEST 2006

Axel Kohlmeyer wrote:
> hi everybody,
> please find attached three patches to the current gromacs
> development cvs tree.

I have implemented your patches in the cvs tree, and fixed all 
compilation errors and warnings.

Note that I did not test any of the code!

> gmx-fio-rename.diff:
>  renames all fio_XXX functions to gmx_fio_XXX as needed
>  for static linkage with PGI runtime libs. not all renames are
>  needed, but for the sake of consistency i changed all functions.
> gmx-misc-fixes.diff:
> contains a fix for a typo in a comment and fixes a segfault
> at the end of jobs that do _not_ use domain decomposition.
> gmx-cray-xt3-port.diff:
>  contains the port of gromacs to the cray xt3 architecture.
>  a few remarks about that. this machine uses two different
>  kernels on the head nodes and the compute nodes with otherwise
>  identical hardware. this means, that the serial compile is
>  like on a normal (opteron CPU), in fact you can copy and
>  use serial binaries from other machines, but for parallel
>  compilation, you have to 'cross-compile' using a different
>  libc adapted to the compute nodes which is linked statically
>  into the binary. the modified configure.ac script should take
>  care of that. the compute nodes have a very restricted
>  kernel (similar to BG/L) without support of sockets,
>  subprocesses, userids, shared libraries, no /proc etc.
>  the more generic features (no nice, no call system)
>  i wrapped into more generic defines GMX_NO_XXX.
>  then there is GMX_BROKEN_CALLOC which is a
>  special story: the default malloc in the special libc for
>  the compute nodes has a bad performance for many
>  small allocations, so there is also libgmalloc as a remedy,
>  but the one shipped with the current versions has a
>  broken callloc (it does not zero out all elements) which
>  explained the 'memory corruption' that shawn and i
>  reported earlier. luckily this can be easily fixed using
>  malloc and memset (i am still puzzled why cray is
>  not able to correct that). the rest are changes particular
>  for the cray xt3 architecture and thus wrapped into
>  GMX_CRAY_XT3 defines.
> please let me know ASAP if you're not happy with any
> of these changes, so i can fix them before i'll be working
> on implementing the changes for our project.
> enjoy,
>    axel.
> ------------------------------------------------------------------------
> _______________________________________________
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-developers-request at gromacs.org.

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