AMBER Archive (2001)Subject: sander on SP3
From: Sanjeev B.S. (sanjeev_at_mbu.iisc.ernet.in) 
Date: Tue Apr 24 2001 - 16:59:16 CDT
 
 
 
 
Hello,
 
        I run sander(AMBER6) on IBM SP3 and get a variety of problems. I
 
cannot comprehed the possible reasons. The important problems are:
 
 1) Running the benchmark of AMBER6
 
   -------------------------------
 
    This runs for 2-9 processors (nodes). Gives problem for single
 
processor or more than ten. In fact, even for 9 processors it gave
 
problems and used to hang(?) after a few steps. I had to run it repeatedly
 
so that it completed 100 steps. This doesn't work with processors greater
 
than nine.
 
 2) If I run the same benchmark (or another input of my own), it works fine
 
with the thin node (ie 4-node SP3 processor). With the main one (16 node)
 
it hangs after sometime. (This is what is written above).
 
 3) I could even run a 16 node job on 4 node successfully, though it took
 
much longer time to complete of course).  But when I run on the main
 
(thick) node using all 16 (or anything greater than 9), the following
 
error crops up (without shake). [Also for higher processes, it comes out
 
saying that it has reached maximum (3000) steps for SHAKE, IF I use
 
shake.]
 
 --------------------------------------------------------------------------
 
  NSTEP =     0  TIME(PS) =    0.000  TEMP(K) =     0.00  PRESS =   9811.84
 
 Etot   =  -58205.0586  EKtot   =       0.0000  EPtot      =  -58205.0586
 
 BOND   =    8976.8757  ANGLE   =     312.4856  DIHED      =     759.3461
 
 1-4 NB =     426.9862  1-4 EEL =    3832.4532  VDWAALS    =   24893.4084
 
 EELEC  =  -97406.6138  EHBOND  =       0.0000  CONSTRAINT =       0.0000
 
 EKCMT  =       0.0000  VIRIAL  =  -50908.9538  VOLUME     =  240306.6153
 
                                                Density    =       0.7817
 
 Ewald error estimate:   0.3273E-03
 
 --------------------------------------------------------------------------
 
  EWALD BOMB in subroutine ew_recip/grid_weights
 
 nmine exceeds num_ks 
 
 System must be very inhomogeneous. Cannot continue.
 
ERROR: 0031-250  task 1: Terminated
 
ERROR: 0031-250  task 2: Terminated
 
ERROR: 0031-250  task 3: Terminated
 
ERROR: 0031-250  task 0: Terminated
 
 --------------------------------------------------------------------------
 
 I am totally confused with this this enigmatic behavior of this machine.
 
 Can anyone tell me where I faltered!? I am attaching the MACHINE file that
 
I have used for compiling. (Also I recompiled on thick node and I am also 
 
using unlimit command).
 
 Thanking in advance,
 
-Sanjeev
 
 Attachment Converted: "c:\eudora\attach\MACHINE1"
 
 
  
 |