amber-developers: AMBER-DEV: partition error in shake (fwd)

From: Scott Brozell <sbrozell.scripps.edu>
Date: Fri, 23 Apr 2004 12:58:48 -0700

Hi Guanglei,

This indicates that a constrained pair of atoms are on different
processors.
I have slightly cleaned up the amber9/src/sander/shake.f and
added a more detailed error message. If this is really a bug
then you will probably have to trace back to the creation of the
ib and jb arguments.

The ./Run.ubiquitin test fails in amber9 using an identical config.h
to that in amber8 where the test passes.
This occurs on several linux machines using ifc 7.1.

Tail of amber9 mdout:
| TOTAL SIZE OF NONBOND LIST = 3971390
 vlimit exceeded for step 0 ; vmax = 807276600744896.

 NSTEP = 1 TIME(PS) = 1000.001 TEMP(K) =????????? PRESS
=********
 Etot = ?????????????? EKtot = ?????????????? EPtot =
**************
 BOND = 220.7117 ANGLE = 637.0885 DIHED =
738.4775
 1-4 NB = 275.3945 1-4 EEL = 3165.3009 VDWAALS =
**************
 EELEC = -15813.8927 EHBOND = 0.0000 RESTRAINT =
0.0000
 EKCMT = 4180.1500 VIRIAL = ************** VOLUME =
167676.3480
 EPOLZ = -391.0542 E3BODY = 0.0000

Diff with amber8:
...
< Box X = 65.656 Box Y = 51.770 Box Z = 49.331

---
>      Box X =   63.571   Box Y =   50.126   Box Z =   47.764
135c135
<      NFFT1 =   72       NFFT2 =   54       NFFT3 =   50
---
>      NFFT1 =   64       NFFT2 =   50       NFFT3 =   48
181,183c181,182
< | Local SIZE OF NONBOND LIST =    3971390
< | TOTAL SIZE OF NONBOND LIST =    3971390
<  vlimit exceeded for step            0 ; vmax =    807276600744896.
---
> | Local SIZE OF NONBOND LIST =    4374419
> | TOTAL SIZE OF NONBOND LIST =    4374419
185,186c184,185
<  NSTEP =        1   TIME(PS) =    1000.001  TEMP(K) =?????????  PRESS
=********
<  Etot   = ??????????????  EKtot   = ??????????????  EPtot      =
**************
---
>  NSTEP =        1   TIME(PS) =    1000.001  TEMP(K) =   298.34  PRESS =
-15.2
>  Etot   =    -38796.4732  EKtot   =      9246.2358  EPtot      =
-48042.7090
...
Scott
---------- Forwarded message ----------
Date: Fri, 23 Apr 2004 11:18:19 -0400
From: Guanglei Cui <cuigl.csb.sunysb.edu>
Reply-To: amber.scripps.edu
To: amber.scripps.edu
Subject: AMBER: partition error in shake
Dear all,
After I managed to get SCALI to work with AMBER8, I found I can't use
more than 4 CPUs for one of my systems (80,000 atoms), but it worked
fine with a regular DNA test case (10 base pairs). It failed because of
"partition error in shake". Before I dive into shake.f, I wonder if
anyone has a quick guess what might be the problem. Thanks in advance.

Received on Wed Apr 05 2006 - 23:50:03 PDT