CHARMM Development Project Forums User Discussion & Questions Energy terms, Constraints, Restraints, and Solvation Segmentation fault error setting up a GSBP.

Dear Users,

I have installed charmm and have been trying to set up GSBP. As a start, I have taken the test files gsbptest1.inp from clongtest folder and the sccgsbp1_2cba.inp from the c32test folders (files attached).
I am using the latest charmm version downloaded from the website.

In both the runs, the simulations crashed with a segmentation fault error. Following instructions in previous posts, I have compiled a debug version of charmm and for the corresponding output. The error is as follows:

"
Backtrace for this error:
#0 0x7fee320e42da in ???
#1 0x7fee320e3503 in ???
#2 0x7fee31760f1f in ???
#3 0x55a7716449e6 in sphe_stpol
at /home/sridhar/charmm_debug/charmm/source/misc/pbeq.F90:16141
#4 0x55a7716afb93 in __pbeq_MOD_gsbp0
at /home/sridhar/charmm_debug/charmm/source/misc/pbeq.F90:954
#5 0x55a7716b8202 in __pbeq_MOD_pbeq0
at /home/sridhar/charmm_debug/charmm/source/misc/pbeq.F90:469
#6 0x55a770d58ced in maincomx_
at /home/sridhar/charmm_debug/charmm/source/charmm/charmm_main.F90:1151
#7 0x55a770d5b5eb in charmm
at /home/sridhar/charmm_debug/charmm/source/charmm/charmm_main.F90:396
#8 0x55a770d5b62c in main
at /home/sridhar/charmm_debug/charmm/source/charmm/charmm_main.F90:2
Segmentation fault (core dumped)

"

In both cases, the program is failing at the same lines in the code. I am wondering if it is the version of charmm or it has something to do with the input file.

Kindly suggest any changes in version/input script.

Thank you very much!!

Regards,
Sridhar

A seg fault is usually a coding error of some kind, either an outright error, or perhaps insufficient checks for bad data values, etc. The "invalid memory reference" message suggests it may be a legitimate bug.

It could also be an issue with how the program was built, i.e. compiler suite and version, parallel (mpi) library and version, and command line options used with either install.com or configure. It would be helpful to provide these names and versions. (Note that the GCC suite that comes with a Linux distribution may not be the best choice.)

Since these are test cases in principle they should work, but the ones in clongtest may not have been evaluated by the developers in quite a while, and the other one depends on SCCDFTB, which most developers are probably not using for routine testing. You should probably run the full test suite, as that may help pin down what the cause might be.

Thank you very much for your reply.

I have installed charmm versions using the command "./install.com gnu T" and "./install.com gnu T +DFTBMKL debug". with the default GCC (gcc (Ubuntu 7.4.0-1ubuntu1~18.04.1) 7.4.)

Both these versions did run for other calculations, including running dynamics with GSBP, where I was able to provide the input .mij and .pjix commands. Also, the SCCDFTB calculations were also running without any issues (in a case where set up of GSBP was not involved).

Would you suggest any other compilers?

Thank you again!!

Regards

I have also tried options of xlarge, just to make sure it was not an issue with charmm size. I still get the same error.

Also, I had tried the code on an em64t architecture with an ifort compiler using

"./install.com em64t large T +DFTBMKL". I have not used debug in this case. However the code dies with a segmentation fault error after the GSBP basis functions calculation.


Thank you.

I have a patch for this.

Apply the attached patch file to a fresh copy of the charmm source from the charmm source root directory with something like

$ patch -p1 -i <full path to patch file>

and then compile a new charmm executable as usual.

Thank you very much Josh Buckner.

It is working now.

Regards,
Sridhar