Segmentation fault by ifort
raysou opened this issue · 1 comments
raysou commented
Sometimes. Not always.
raysou commented
- Use
DEBUG=set
while running make with ifort:make FC=ifort DEBUG=set
- nasa doc
- ifort docs
In our tests the issue seems to be with the linked list : changing rcut
in the code resolves the issue. At the "wrong" rcut
s we have x(l,i), y(l,i) = NaN. We don't yet understand why that is the case. The right rcut
s give physically valid checkpoints as we checked.
We hope cleaning up the code with the modified solely beadbased 1D linked list shall fix this.
Also, as per the intel doc above, Large Array Temporaries Caused by Passing Noncontiguous Array Sections, may cause segfault. In our current code every array reference is non-contiguous as the beads are stored in rows. Changing that might fix things too.