[RE]AMBER: RE: I got the error meassage in testcase of QMMM

From: Lee Kyung-koo <daum203.hanmail.net>
Date: Wed, 08 Nov 2006 01:46:25 +0900 (KST)

Dear Walker,

 

I tried to test the testcase of some cases.

 

In the testcase in amber9, there are not difference with itanium patched and non-patched version.

 

In single mode,  sander displayed error message in previous e-mail.

 

In parallel version with one process (-np 1), there are some message that is similiar with single mode but fewer than single mode.

 

When I executed 'make test.sander.QMMM.MPI' with 'DO_PARALLEL=mpirun -np 2', there are the following errer in crambin.out.

 

   4.  RESULTS
--------------------------------------------------------------------------------

 

   NSTEP       ENERGY          RMS            GMAX         NAME    NUMBER
      1     NaN            NaN              4.2278E+03     SG        373

 BOND    =       59.6787  ANGLE   =       74.0627  DIHED      =  NaN
 VDWAALS =     1350.9930  EEL     =    -2611.8056  HBOND      =        0.0000
 1-4 VDW =      183.5164  1-4 EEL =     1450.1218  RESTRAINT  =        0.0000
 PM3ESCF =      -40.0704

QMMM: WARNING!
QMMM: Unable to achieve self consistency to the tolerances specified
QMMM: No convergence in SCF after   1000 steps.
QMMM: Job will continue with unconverged SCF. Warning energies
QMMM: and forces for this step will not be accurate.
QMMM: E =  NaN         DeltaE =  NaN         DeltaP =   0.0000E+00
QMMM: Smallest DeltaE =   0.1000E+31 DeltaP =   0.1000E+31 Step =      0

 

   NSTEP       ENERGY          RMS            GMAX         NAME    NUMBER
      2     NaN            NaN              0.0000E+00     N           1

 BOND    =  NaN           ANGLE   =  NaN           DIHED      =  NaN
 VDWAALS =        0.0000  EEL     =        0.0000  HBOND      =        0.0000
 1-4 VDW =  NaN           1-4 EEL =  NaN           RESTRAINT  =        0.0000
 PM3ESCF = NaN

QMMM: WARNING!
QMMM: Unable to achieve self consistency to the tolerances specified
QMMM: No convergence in SCF after   1000 steps.
QMMM: Job will continue with unconverged SCF. Warning energies
QMMM: and forces for this step will not be accurate.
QMMM: E =  NaN         DeltaE =  NaN         DeltaP =   0.0000E+00
QMMM: Smallest DeltaE =   0.1000E+31 DeltaP =   0.1000E+31 Step =      0

 

 

 

Run.crambin ,Run.crambin_md and Run.crambin_md_calc_mulliken were runned with 2 process.

 

Run.crambin_md_qmgb2 was runned with 1 process and still run.

 

 

 

 

and

 

cat /proc/cpuinfo

 

processor  : 0
vendor     : GenuineIntel
arch       : IA-64
family     : Itanium 2
model      : 1
revision   : 5
archrev    : 0
features   : branchlong
cpu number : 0
cpu regs   : 4
cpu MHz    : 1595.699000
itc MHz    : 1595.699000
BogoMIPS   : 2390.75

processor  : 1
vendor     : GenuineIntel
arch       : IA-64
family     : Itanium 2
model      : 1
revision   : 5
archrev    : 0
features   : branchlong
cpu number : 0
cpu regs   : 4
cpu MHz    : 1595.699000
itc MHz    : 1595.699000
BogoMIPS   : 2390.75

 

[sunkist.alpha2 test]$ ifort -V
Intel(R) Fortran Itanium(R) Compiler for Itanium(R)-based applications
Version 9.0    Build 20051201 Package ID: l_fc_c_9.0.031
Copyright (C) 1985-2005 Intel Corporation.  All rights reserved.
FOR NON-COMMERCIAL USE ONLY

 

 

 

If you need some files, I will post that quickly.

 

Thank you.

 

Take care


 

---------[ ¹ÞÀº ¸ÞÀÏ ³»¿ë ]----------
Á¦¸ñ : AMBER: RE: I got the error meassage in testcase of QMMM
³¯Â¥ : Sat, 4 Nov 2006 10:09:13 -0800
º¸³½ÀÌ : "Ross Walker" <ross.rosswalker.co.uk>
¹Þ´ÂÀÌ : <amber.scripps.edu>

Dear Lee,

Firstly did you apply the patch I sent several days ago to address hanging
when running in parallel? Did it work?

>I installed amber9 single version in intel itanium server called white box.
>After install, I runned testcase with make test.sander.QMMM.
>In crambin_md.out, I got strange messages as follow.

This is in serial yes?

>QMMM: WARNING!
>QMMM: Unable to achieve self consistency to the tolerances specified
>QMMM: No convergence in SCF after 1000 steps.
>QMMM: Job will continue with unconverged SCF. Warning energies
>QMMM: and forces for this step will not be accurate.
>QMMM: E = -0.2405E+06 DeltaE = 0.3389E-01 DeltaP = 0.2716E-01
>QMMM: Smallest DeltaE = 0.3947E-02 DeltaP = 0.5110E-01 Step = 74
>
>vlimit exceeded for step 12; vmax = 97.6111
>
>How can I overcom this warning?

Since this test case passes on other system I have tried I once again
suspect that this is a case of the compiler not doing what it is supposed
to. Can you please give me detailed specs of the machine so that I can try
to track this down.

For this I will need:

Operating system and version
Details of "cat /proc/cpuinfo"
Compiler version (use ifort -V if the intel compiler - I need the full
version number. E.g. 9.1.033)
Options passed to ./configure
A complete copy of crambin_md.out from the above test case.
The results of "echo $TESTsander" and "echo $DO_PARALLEL" in the same shell
in which you are running make test.sander.QMMM

Is this the only test case that fails? Can you do exactly the following...

cd $AMBERHOME/test/
make clean
make

Let all tests run to completion and then send me the file
$AMBERHOME/test/TEST_FAILURES.diff

Once I have these details I can try and reproduce the problem myself and see
if I can come up with yet another workaround...

All the best
Ross

/\
\/
|\oss Walker

| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
| http://www.rosswalker.co.uk | PGP Key available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.


-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu



"¿ì¸® ÀÎÅͳÝ, Daum" http://www.daum.net ¡ºÆò»ý¾²´Â ¹«·á ÇѸÞÀϳݡ»
----------------------------------------------------------------------- The AMBER Mail Reflector To post, send mail to amber.scripps.edu To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu Received on Wed Nov 08 2006 - 06:07:35 PST
Custom Search