[CPMD-list] crash of parallel CP slab calculations with Vanderbilt PPs
Martin Konopka
konopka at elf.stuba.sk
Thu Aug 22 20:18:26 CEST 2002
Hello all,
Does anybody know how to fix the folowing problem?
CPMD 3.5.2 crashes after I start a large parallel Car-Parinello MD
job on IBM SP with a slab of copper atoms (and apparently many other
systems, too). Features:
* No error message in the standart output file
* The folowing messages in the standart error file:
ERROR: 0031-250 task 0: Terminated
ERROR: 0031-250 task 1: Segmentation fault
ERROR: 0031-250 task 2: Segmentation fault
.........
ERROR: 0031-250 task 15: Segmentation fault
(I.e., it is a 16-processor run.)
* one core file created by each process (may be not by the 0th one,
I forgot)
* The crash occurs at the beginning of the run. The last part of the stdout
file looks like
47 Cu 14.531590 9.687727 8.621875
48 Cu 14.531590 14.531590 8.621875
****************************************************************
DEGREES OF FREEDOM FOR SYSTEM: 144
EWALD SUM IN REAL SPACE OVER 5* 5* 5 CELLS
================================================================================
Job sp5cws.rz.uni-karlsruhe.de.156990.0 stopped at 22.08.2002/16:51:11 (RC=139)
* I am using a Vanderbilt pseudopotential.
* No crash occurs when I do just wavefunction or geometry optimization.
* No crash occurs even when I do Car-Parinello MD with Cu2 molecule.
(tested on 4 and 16 processors)
* Crash occurs for both tested Cu slabs: with 48 and 96 atoms.
* Crash occurs for both OLDCODE and NEWCODE.
I am using a version of CPMD 3.5.2 that has at least two bugs fixed
(one concerns MD and was fixed by some patch and the other was causing
crashes when RESTART files where saved after the 400th MD step.)
I have not tested this behaviour on other platforms.
Below I add parts of my input file.
Thank you for your interest.
Martin Konopka.
&CPMD
RESTART WAVEFUNCTIONS COORDINATES LATEST
RESTFILE
2
STORE
150
MOLECULAR DYNAMICS CP
TEMPERATURE
1000.0D0
MAXSTEP
1000
NOSE IONS MASSIVE
1000.0D0 1000.0D0
EMASS
400.0D0
TIMESTEP
5.0D0
ALEXANDER MIXING
0.3D0
SPLINE POINTS QFUNCTIONS
2001
CENTER MOLECULE OFF
MEMORY BIG
&END
&DFT
OLDCODE
GRADIENT CORRECTION PBEX PBEC
GC-CUTOFF
5.0D-5
&END
&SYSTEM
CELL VECTORS
19.37545337626741472795D0 0.0D0 0.0D0 0.0D0 19.37545337626741472795D0
0.0D0 0.0D0 0.0D0 22.0D0
TESR
4
CUTOFF
25.0D0
CHARGE
0
&END
&ATOMS
*Cu.uspp BINARY NEWF TPSEU
LMAX=D
48
0.0000000000 0.0000000000 15.4721327209
..........................
14.5315900317 14.5315900317 8.6218754857
&END
-----------------------------------------
Martin Konopka
Department of Physics, FEI STU
Ilkovicova 3, 842 19 Bratislava, Slovakia
tel: +421 2 60291 714
fax: +421 2 6541 1483
http://www.ccms.elf.stuba.sk/konopka.html
-----------------------------------------
More information about the CPMD-list
mailing list