[lammps-users] segmentation fault with pppm, using example - peptide

LAMMPS LAMMPS Mailing List Mirror
1

Hello,
I got a segmentation fault in the example simulation using kspace_style pppm but not in a simulation using ewald. I have tried to use pair_modify table 0 and tried playing with the pppm precision setting. Either didn't work. LAMMPS is crashing after setuping pppm. ===========
LAMMPS (12 Feb 2007)
Scanning data file ...
  3 = max bonds/atom
  6 = max angles/atom
  14 = max dihedrals/atom
  1 = max impropers/atom
Reading data file ...
  1 by 1 by 1 processor grid
  2004 atoms
  2004 velocities
  1365 bonds
  786 angles
  207 dihedrals
  12 impropers
Finding 1-2 1-3 1-4 neighbors ...
  4 = max # of 1-2 neighbors
  7 = max # of 1-3 neighbors
  14 = max # of 1-4 neighbors
  18 = max # of special neighbors
Finding SHAKE clusters ...
  19 = # of size 2 clusters
  6 = # of size 3 clusters
  3 = # of size 4 clusters
  640 = # of frozen angles
84 atoms in group peptide
PPPM initialization ...
  G vector = 0.268725
  grid = 15 15 15
  RMS precision = 4.85953e-05
  brick FFT buffer size/proc = 10648 3375 5808
Setting up run ...
Segmentation fault

2

If you're running examples/peptide/in.peptide, then it works
for us obviously, on several boxes. So can you try to
figure out where it is crashing, e.g. run it under valgrind if
you are compiling with g++ on a Linux box.

Also can you tell us what compiler, system you are running on?

Paul, do you have any ideas on this?

Steve

3

Hi all,

For what its worth, I had some trouble with that example file on my linux cluster too with an older version of Lammps (gcc 4.1.0, mpich 1.2.6, fftw-mpi 2.1.3). If I remember correctly, the problem was with the pppm because using a straight ewald summation enabled it to run.

Cheers,
Andrew

4

Seems I need to install valgrind to run valgrind.

This is the system info: (uname=) Linux lsa.umich.edu 2.6.9-22.ELsmp #1 SMP Sat Oct 8 19:11:43 CDT 2005 i686 athlon i386 GNU/Linux

This is the makefile I used and no compiling error when it was compiled. Thanks. JW

# System-specific settings

CC = g++
CCFLAGS = -O -I../STUBS \
                -I/home/ubecker/JIANWEI/apps/fftw-2.1.5/include -DFFT_FFTW
DEPFLAGS = -M
LINK = g++
LINKFLAGS = -O -L../STUBS \
                -L/home/ubecker/JIANWEI/apps/LAMMPS/lammps2fft/lib/meam \
                -L/opt/pgi-5.1/linux86/5.1/lib \
                -L/home/ubecker/JIANWEI/apps/fftw-2.1.5/lib
USRLIB = -lmpi -lfftw -lmeam -lpgf90 -lpgf90_rpm1 -lpgf902 -lpgf90rtl -lpgftnrtl -lm -lpgc -lgcc
SYSLIB =
ARCHIVE = ar
ARFLAGS = -rc
SIZE = size

# Link target

Quoting Steve Plimpton <[email protected]>:

5

Seems I need to install valgrind to run valgrind.

This is the system info: (uname=) Linux lsa.umich.edu 2.6.9-22.ELsmp #1
SMP Sat Oct 8 19:11:43 CDT 2005 i686 athlon i386 GNU/Linux

This is the makefile I used and no compiling error when it was
compiled. Thanks. JW

please try adding -fno-strict-aliasing to CCFLAGS do
a make clean and recompile. seems like newer g++ versions
are more picky about what they compile correctly or not.
i've just been down that road. :wink:

cheers,
   axel.

6

The example problem runs fine on my box as is. Do you have the latest version of LAMMPS? Are you using an unmodified version of examples/peptide/in.peptide? Looks like you have compiler optimizations off, and tabling off. Checking for memory leaks using Valgrind would indeed be helpful to us. You could just avoid the issue by using Ewald instead of PPPM. Does the problem occur when running other input scripts that call for PPPM?

Paul