[CPMD-list] Wrong kinetic energy when running with LAM (but not for serial exec.)

[Henryk Wicke]

Dear all,

I've recently compiled CPMD 3.11.1 for a small cluster operating with 
Intel Core 2 Quad CPUs.
I've used ifort 9.1.043 with the Intel MKL 9.1.023 for the serial 
version which seems to be running fine so far. (Specific optimizations 
used: -O2 -tpp7 -xT)

For the parallel version I've also used LAM 7.1.4 which I had previously 
built with icc/icpc 9.1.051 and ifort 9.1.043.
The compilation of the parallel version went fine and first tests from 
the test suite were promising. Now I went on to check some calculations 
from the CPMD tutorial at A. Kohlmeyer's website and ran into trouble 
with the calculation 2-h2o-pbc-geo-linsc in section 6.2.2. Especially 
the kinetic energy in the "FINAL RESULTS" section is conspicuous...

The serial calculation seems to yield good results. The results differ 
for the LAM calculations though:

Serial binary:
KINETIC ENERGY =           12.82302639 A.U.

LAM with ONE CPU core:
KINETIC ENERGY =           12.82302824 A.U.

Completely different for LAM with 2 cores:
KINETIC ENERGY =            6.36582557 A.U.

It might be interesting that the kinetic energy listed shortly before 
the line "END OF GEOMETRY OPTIMIZATION" is still reasonable, i.e. 
12.82302... A.U., even in the last case.

I found a similar case in the mailing list archives where Axel Kohlmeyer 
recommended to recompile wrener.F without optimizations. I recompiled 
everything with -O0 just to see if it makes a big difference, but it 
doesn't:

LAM with ONE CPU core:
KINETIC ENERGY =           12.82302922 A.U.

LAM with 2 cores:
KINETIC ENERGY =            6.36582683 A.U.

Any help and/or insight is very much appreciated, especially because I'm 
new at both CPMD as well as operating my own cluster and compiling such 
relatively complex programs.

Regards,
Henryk Wicke

-- 
Dipl.-Phys. Henryk Wicke

Leibniz Universitaet Hannover
Faculty of Mathematics and Physics
Center for Radiation Protection and Radioecology (ZSR)

Herrenhaeuser Str. 2
D-30419 Hannover
Germany

Phone +49(0)511 762 3069
Fax   +49(0)511 762 3008
Mail  wicke at zsr.uni-hannover.de
Web   www.zsr.uni-hannover.de