[gmx-users] Ewald summation: segmentation fault!

[Vincent Ballenegger]

David wrote:

>It's weird anyway. We appreciate people doing this kind of tests, since
>we can't do everything...
>
>The fact that your md.log file ends in the middle of printing is not
>indicative of anything (check the filesize, it is probably a multiple of
>4096, i.e. file buffering in action).
>
>You may want to rerun in the following way:
>setenv LOG_BUFS 0
>mdrun -debug
>
>Please report back what happens.
>
When I run mdrun in this way, the last 7 lines in the .log file become:
.................................................................
Initial temperature: nan K

Testing x86 processor CPUID...

Testing x86 SSE capabilities...
CPU and OS support SSE.
Using Gromacs SSE single precision assembly innerloops.
.................................................................
The crash seems therefore to occur before MD step 0.

Since the temperature is not a number (?), I removed the temperature 
coupling in the grompp.mdp file (tcoupl=no). This does not solve the 
problem: the simulation runs with PME but crashes with segmentation 
fault with Ewald. (Adding a solvent to have a well defined temperature 
doesn't solve the problem either).

Vincent