AMBER Archive (2006)

Subject: Re: AMBER: "NO CONVERGENCE IN SCF CALCULATION" in divcon in Amber9 package

From: Zhihong Yu (
Date: Thu Sep 14 2006 - 01:47:07 CDT

>On Thu, Sep 14, 2006, Zhihong Yu wrote:
>[divcon gives NaN energies and crashes when called from antechamber]:
>> operating system : RedHat Linux 9
>> compiler : Intel FC 9.0 (ifort 9.0.031)
>> CPU : Intel Xeon (3.06GHz) dual cpu
>> Generated config.h via command : ./configure -mpich -p4 -bintraj ifort_ia32
>> file of "libgm.a" exist in : /opt/xcat/gm/lib
>> MKL_HOME : /opt/intel/mkl70
>Thanks for the info. I have ifort version 9.0.026 installed on my machines,
>and don't see any problems with divcon. Maybe someone on the list can comment
>on problems (or lack of problems) with 9.0.031. [We also need a web page
>with "known good/bad compiler versions" --come on Amber developers, someone
>needs to step up to the plate here and pitch in!]
>I would like to confirm some other things that I thought I understood from
>your previous post:
>1. The qmmm2/crambin/Run.crambin_divcon test fails (presumably giving NaN's
> and no energies, like the example you posted earlier)?

yes, it's true that qmmm2/crambin/Run.crambin_divcon fails, my crambin_divcon.out file is attached

>2. All of the other tests in "make test.sander.DIVCON" pass(?)

yes, all of the other tests in "make test.sander.DIVCON" passed, divcon_test.log is attached
>3. The antechamber tests (which also call divcon) all pass (or show only
> minor diffs in charges).
No, not all antechamber tests had passed, such as ash, sustiva, guanine_amber. But they exactly showed minor diffs in charges.

ALL_TEST_FAILURES.diff and ALL_TEST.log are attached.

>Having this sort of information may help us narrow down the problem. For
>obvious reasons, if any of the test cases are failing, that is always the best
>thing to report, since we don't have to worry about tracking down input errors
>and the like for those.

thanks very much!

Zhihong Yu

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to