[AMBER] Inconsistencies between reported lines and executed lines in source code

From: Albert Lee <hojoon.lee.yahoo.com>
Date: Thu, 31 Mar 2011 11:12:23 -0700 (PDT)

Hi, everyone.

I'm a research assistant using a modified form of AMBER 10, and I'm trying to perform some optimization on this modified form. Naturally, I've been working with a debugger for this purpose (GDB, mostly), and I've come across some strange behavior in _short_ene.f. Once the program hits the short_ene_ subroutine, it jumps about 300 lines and ends up in the wrong place. That is, GDB will report stepping from the start of the subroutine, at line 521, to line 862, when in reality, the program is actually executing line 951. In other words, there is an offset between the reported line by GDB and the executed line by sander.

I've noticed this happening in other areas, but it is most striking in short_ene_, as far as I can tell. Line deviations range from about 500 lines to 450 lines in the eedmeth "if" block. My question, then, is: why does this behavior occur?

The modified form of AMBER isn't the issue, as after I noticed this, the first thing I did was to check the basic AMBER 10 install, which produces the same results. All of the compilations of AMBER I've done have been with -g and -O0 options when applicable; I've compiled with gfortran and g95, and tried out dbx, to no avail.

I'm running AMBER 10 on CentOS Linux, i386, as a virtual machine on Oracle VirtualBox.

AMBER mailing list
Received on Thu Mar 31 2011 - 11:30:03 PDT
Custom Search