Project

General

Profile

Bug #364

NAN in g_anaeig -proj

Added by Alexander empty almost 10 years ago. Updated almost 10 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Erik Lindahl
Category:
analysis tools
Target version:
Affected version - extra info:
Affected version:
Difficulty:
uncategorized
Close

Description

Hi!
I run into a problem trying to project a trajectory on the eigenvector(s) (with g_anaeig -proj) from covariace matrix. All projections I have calculated are nan. However, the eigenvalues are OK and the g_anaeig -comp -v2 -eig2 -over options works OK so eigenvectors seems are OK (but g_anaeig -2d fails to nan too). I am using gromacs-4.0.5 on x86-64 CentOS5 (compiled with gcc-34).
Regards,
Alex.
P.S. To reproduce bug unzip attachment and enter g_anaeig_d -v sss_1000_eigenvec.trr -f sss_after_mdsi_1000.gro -s sss_mdsi_1000.tpr -first 1 -last 1 -n sss.ndx -proj. Then check proj.xvg for nan.

problem.tar.gz (2.15 MB) problem.tar.gz files for bug reproducing Alexander empty, 11/09/2009 07:29 PM

History

#1 Updated by Berk Hess almost 10 years ago

Could you add the attachment?

Berk

#2 Updated by Alexander empty almost 10 years ago

Created an attachment (id=401)
files for bug reproducing

#3 Updated by Berk Hess almost 10 years ago

The was a bug with mass-weighted covariance analysis when the fitting group
did not consist of the first N atoms in the system.
I fixed it for 4.0.6 and git master.

Berk

Also available in: Atom PDF