[CPMD-list] COORSP working on IBM-SP, failing on other platforms

[Martin Konopka]

Dear CPMD developers,

It looks that there is a platform-dependent bug in CPMD 3.11
displaying itself when performing coordination constraint
COORSP runs. See my example in the attachment. The task worked
only on IBM-SP and it failed on other tested configurations:

LINUX_IA64_INTEL-MPI, LINUX-ALPHA-MPI, PGI-AMD64-MPI:
the run crashed when starting the MD stage (e.g. segmentation
fault on LINUX-ALPHA-MPI or reported signal 11 on PGI-AMD64-MPI);
the crash occured when writing to RESTART.1 file.

PC-PGI, PC-IFC:
the MD stage even did not normally start because the wavefunction
optimization could not converge.

I attach representative results from some of the mentioned platforms.

Thank you for your interest.

Greetings
Martin.


------------------------------------------------------------------------
Dr. Martin Konopka             http://www.ccms.elf.stuba.sk/konopka.html
Department of Physics, CCMS                      tel: +421 (0)2 60291714
Slovak University of Technology (FEI STU)        fax: +421 (0)2 65411483
Ilkovicova 3, 812 19 Bratislava, Slovakia
------------------------------------------------------------------------
-------------- next part --------------
A non-text attachment was scrubbed...
Name: coordin_cnstr_test.tar.bz2
Type: application/x-bzip2
Size: 32028 bytes
Desc: 
Url : http://cpmd.org/pipermail/cpmd-list/attachments/20060620/d96ca055/attachment.bz2