AMBER: sander ntc=2 ntf=1

From: Nicolas Lux Fawzi <fawzin.berkeley.edu>
Date: Mon, 18 Dec 2006 19:10:42 -0800

If I'm running an amber9 sander ff99 protein + TIP3P (or TIP4P) water
REMD simulation ntc=2 ntf=2, I get a problem on AIX and opteron
platforms.

My question is, can I run with ntc=2, ntf=1, which seems to work? Will
this give the "correct" and expected output. As for the errors with
ntc=ntf=2, I just get "terminated" mpi signal... I suspect if might be
some type of MPI timeout. On an opteron with infiniband (i don't know
which MPICH but it's NERSC's jacquard which Bob Duke is familiar with),
I was getting VAPI timeout errors, which may be fixed with changing the
timeouts and retries... see the output below

Please let me know if you have any insight on if ntc=2, ntf=1 is
appropriate.
-Nick


bash-2.05b$ cat mdinremd.1
   sander input for TIP3P plus peptide MD
  &cntrl
         nstlim = 1000,
         ntx = 5,
         irest=1, iwrap = 1,
         ntt = 2,
    temp0 = 270
         ntc = 2,
         ntb = 1,
         ntf = 1,
         cut = 9.0,
         ntwx = 500,
         numexchg = 4000,

  /

  &ewald
         ew_type = 0,
         skinnb = 2.0,
  /

END
END
_________________________________________

from the standard output of an AIX loadleveler job:

         Group = 14
         WorldRank = 14
         NodeID = 0

         Group = 15
         WorldRank = 15
         NodeID = 0


  NSTEP = 400 TIME(PS) = 19465.400 TEMP(K) = 412.19 PRESS =
     0.0
  Etot = -11501.2579 EKtot = 4008.6951 EPtot =
-15509.9530
  BOND = 25.7644 ANGLE = 79.2867 DIHED =
79.8073
  1-4 NB = 29.9351 1-4 EEL = 629.7989 VDWAALS =
2454.6631
  EELEC = -18809.2085 EHBOND = 0.0000 RESTRAINT =
  0.0000
  Ewald error estimate: 0.8865E-05
   
------------------------------------------------------------------------
------

ERROR: 0031-250 task 6: Terminated
ERROR: 0031-250 task 31: Terminated
ERROR: 0031-250 task 46: Terminated
........................


-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Dec 20 2006 - 06:07:36 PST
Custom Search