Segmentation fault with GADGET 3 modification

From: Joan Bachs-Esteban <joan.bachs_at_tecnico.ulisboa.pt>
Date: Tue, 21 Dec 2021 17:17:30 +0000

Hello everyone,

I am working with the developer version of GADGET3. I am working with
beyond-LCDM modifications that account for Modified Gravity, in
particular, the modgrav package. I have introduced a further
modification that makes the code consider radiation and other species
instead of the usual one in LCDM when computing cosmological values such
as the Hubble parameter.

It is not a deep modification so I was not expecting it to affect the
performance of the code, but I have encountered a segmentation fault. I
have found that the code stops running when it hits the following line
(folder: modgrav, file: modgrav_fieldsolve.c, line: 1242):

MPI_Allreduce(&phi_max_local, &phi_max_global, 1, MPI_DOUBLE, MPI_MAX,
MPI_COMM_WORLD);

where phi_max_local = -1e30 and phi_max_global are declared just above
(line 1227) at the beginning of a function. Apparently, it is a simple
calculation that does not consider any parameters of my modification, so
I do not understand what is going on. I guess that perhaps I am using
MPI incorrectly, but I execute the same command without my modification
off and the code runs.

Has anyone encountered a similar problem or have any idea where I should
look or how my modification might be messing with the program?

Thanks in advance,

Joan Bachs-Esteban
Instituto Superior Técnico - University of Lisbon
Received on 2021-12-21 18:17:42