I've had no success in running through a minimization of a system of water
and nitric acid molecules. 2670 atoms in a 30x30x30 cubic periodic box.
I'm attaching a few files for the interested:
-topology
-input coords
-mdout of the run
The input (mdin) of the run is a rather simple one:
-----
General minimization script
&cntrl
IMIN = 1,
IWRAP = 1,
NTPR = 250,
NTMIN = 1,
MAXCYC = 100000, NCYC = 3500,
/
-----
the specific error I get (when running sander.MPI) is:
-----
Signal:11 info.si_errno:0(Success) si_code:1(SEGV_MAPERR)
Failing at addr:0x7dfabde0
[0] func:/usr/local/lib/libopal.so.0 [0x400f5b38]
[1] func:[0x40000440]
[2] func:sander.MPI(ewald_force_+0x1a63) [0x8131d83]
[3] func:sander.MPI(force_+0x1270) [0x8254ed8]
[4] func:sander.MPI(runmin_+0x598) [0x80fdc68]
[5] func:sander.MPI(sander_+0x5427) [0x80f17a3]
[6] func:sander.MPI(MAIN__+0x14ab) [0x80e8053]
[7] func:sander.MPI(main+0x42) [0x806a4da]
[8] func:/lib/tls/libc.so.6(__libc_start_main+0xd3) [0x8a5e23]
[9] func: sander.MPI [0x806a411]
*** End of error message ***
-----
When running a serial version (sander) the following error is produced:
-----
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
sander 08103B3C Unknown Unknown Unknown
sander 08102E10 Unknown Unknown Unknown
sander 08219F37 Unknown Unknown Unknown
sander 080D2CA7 Unknown Unknown Unknown
sander 080C6B43 Unknown Unknown Unknown
sander 080C36D1 Unknown Unknown Unknown
sander 0804B1BA Unknown Unknown Unknown
libc.so.6 008A5E23 Unknown Unknown Unknown
sander 0804B0F1 Unknown Unknown Unknown
-----
Changing the NCYC will allow the run to proceed until that number (i.e. if
NCYC = 500, the run crashes at 500, if NCYC = 2500, it crashes at 2500,
etc.) What I notice is that the energies listed in the mdout are rather
large, but I'm at a loss for how to reduce the initial values as I've got no
overlaps of atoms, and I've checked out the force field parameters and they
seem kosher. I've read through a number of the segfault topics on the
reflector that come close and it doesn't seem that anyone has come up with a
solution beyond blaming the compiler. I've been able to successfully
minimize a different system (similar, but all water molecules without any
nitric acid), so something tells me that this segmentation fault is indeed
system-specific: an artifact of the initial conditions and not the compiled
code.
The system used is a dual-processor athlon (32-bit), and the amber 9 source
was compiled with ifort 9.1.040. Openmpi 1.1.4 compiled with the same ifort.
Intel MKL 9.0 libraries were used.
Any help is greatly appreciated. I'm at a loss for whether the installation
was done correctly or not, and would like to pinpoint where the problem may
be. Please feel free to ask for further information on anything I've
forgotten to include.
--
~Eric Shamay
eric.shamay.gmail.com
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Mar 18 2007 - 06:07:15 PDT
