Segmentation fault (core dumped) for gmx covar

User discussions
1

I try to run the PCA analysis with ‘gmx covar’, but it show me this:
Calculating the average structure …
Reading frame 0 time 0.000
WARNING: number of atoms in structure file (18148) and trajectory (18148) do not match
Last frame 100 time 1000.000
Constructing covariance matrix (54444x54444) …
Last frame 100 time 1000.000
Read 101 frames

Trace of the covariance matrix: 269.083 (nm^2)
100%
Diagonalizing …
Segmentation fault (core dumped)

How can I deal this problem?Thanks

2

With a 54444^2 size matrix your machine will likely run out of memory. But something seems very wrong here. The warning already doesn’t make sense, as the atom counts actually match. But the dimensions of the covariance matrix are much larger than the atom count. I don’t understand what is happening here.

Which GROMACS version are you using?
How large is the group that you selected for the analysis?

3

My version is 2024.4
And the Group 1 “Protein” (18195 atoms)

4

:)

5

That explains the confusing error message.

I now also see that 54444 = 18148 * 3, so this is actually fully correct. But 54444^2 is larger than max int, which is not supported. We should add a check for this.

You should choose fewer atoms for your analysis. I guess excluding hydrogens would make it run, but it might still take extremely long. C-alpha only runs fast.

6

Ok,I will choose group with less atoms. Thanks.