AMBER: running sander_classic on multiple processors

From: Richard Smith <dicksmit.umich.edu>
Date: Tue, 14 Oct 2003 09:48:23 -0400

Hi,

        I am trying to get sander_classic of AMBER 6 to run on a dual
processor machine (trying to get AMBER to utilize both processors) with
MPICH version 1.2.5.2, running on Linux (Red Hat version 7.1).
Everything appeared to compile properly using the -DMPI flag, and the
CPP and LOADLIB variables pointing to the appropriate include and
library files for MPICH in the MACHINE file. However, when I try to run
sander_classic with the command "mpirun -np 2
$AMBERHOME/exe/sander_classic -i md.in etc.", sander_classic starts up
on multiple processors, and starts writing to the output file, but after
a short period of time (~20 seconds) it only runs on one processor, and
does not write to any of the output files, so I can't tell if the
calculation is proceeding. The same calculation using the command
"mpirun -np 1 $AMBERHOME/exe/sander_classic -i md.in etc." runs just
fine, and finishes before the two processor job. Below is the MACHINE
file and input file for the mdrun.
        Unfortunately I don't get any error messages from either
sander_classic or MPICH.

Here are the settings of the MACHINE file:

#
setenv MACHINE "linux/FreeBSD/Windows PC"
setenv MACH Linux
setenv MACHINEFLAGS "-DREGNML -DMPI"

# CPP is the cpp for this machine
setenv CPP "/lib/cpp -traditional -I/users/MPICH/mpich-1.2.5.2/include"
setenv CC gcc

# SYSDIR is the name of the system-specific source directory relative to
src/*/
setenv SYSDIR Machines/g77

# COMPILER ALIASES:

# LOADER/LINKER:
setenv LOAD "g77"
setenv LOADLIB "-lm -L/users/MPICH/mpich-1.2.5.2/lib -lmpich"
setenv G77_COMPAT " "
#
# following seems to actually slow down code, though it works well on
CHARMM:
#setenv G77_OPT "-O6 -m486 -malign-double -ffast-math
-fomit-frame-pointer -funroll-loops -funroll-all-loops -mcpu=pentiumpro
-march=pentiumpro -ffloat-store -fforce-mem -frerun-cse-after-loop -fex
pensive-optimizations -fugly-complex"

# following appears to be the best we have found so far:
setenv G77_OPT " "

# little or no optimization:
setenv L0 "g77 -c -g"

# modest optimization (local scalar):
setenv L1 "g77 -c $G77_OPT $G77_COMPAT"

# high scalar optimization (but not vectorization):
setenv L2 "g77 -c $G77_OPT $G77_COMPAT"

# high optimization (may be vectorization, not parallelization):
setenv L3 "g77 -c $G77_OPT $G77_COMPAT"

# ranlib, if it exists
setenv RANLIB ranlib

Here is a copy of the input file:

temperature ramping input file
 &cntrl
  NTX = 1,
  IREST = 0,
  INIT = 3,
  SCEE = 1.2,
  NTB = 1,
  NTC = 2,
  NTF = 2,
  CUT = 12.0,
  NSNB = 30,
  NSTLIM = 10000,
  DT = 0.001,
  NDFMIN = 6,
  TEMPI = 10,
  TEMP0 = 70,
  HEAT = 2.6457513,
  NTT = 1,
  TAUP = 0.2,
  NTP = 0,
  NTWR = 500,
  NTWE = 500,
  NTWX = 500,
  NTWXM = 0,
  NTWEM = 0,
  NTWVM = 0,
  JFASTW = 0,
  NTCM = 1,
  IBEllY = 0,
 &end
        
Thank you for any assistance you may provide.

Richard Smith
Graduate Student
University of Michigan
Department of Biophysics



-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Tue Oct 14 2003 - 14:53:01 PDT
Custom Search