Molcas Forum

mcurie747

Member

Registered: 2019-12-12

Posts: 21

SIGSEGV: Segmentation fault

Hello everyone,
Hope you guys are doing well.
I recently got OpenMolcas installed but I'm having issues running calculation in parallel. Regarding the requirements for openmolcas I used the following versions:
- OpenMPI v4.1.4
- OpenBLAS v0.3.17
- GlobalArrays v5.8.1
- Cmake v3.24.0-rc1

Everything was smoothly. I didn't have warning or any issues compiling or installing the program. When I tried to run a calculation with just 1 processor it worked well. However when I specified more than one processor, I got this problem "Program received signal SIGSEGV: Segmentation fault - invalid memory reference."

This is my input (really simple, I was just testing):

&GATEWAY
Basis set
H.3-21G.....
H1 0.0  0.0 0.0
End of basis

&SEWARD
Grid input
 RQuad= Log3; nGrid= 50000; GGL; lMax= 26; Global
End of Grid Input

&SCF; Occupations=1; KSDFT=LDA5; Iterations= 1 1

This is the job file:

#!/bin/bash
  
#SBATCH -p intel
#SBATCH --job-name=h2o
#SBATCH -N 1
#SBATCH --tasks-per-node=2
#SBATCH --mem-per-cpu=16000
#SBATCH -o job-%j.stdout
#SBATCH -e job-%j.stderr

#module purge
#module load openmolcas/pymolcas

# OpenMP settings
export PATH="$PATH:/home/mariab/opt/openmpi/bin"
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/home/mariab/opt/openmpi/lib/"
export OMP_NUM_THREADS=1

# Molcas environment
export MOLCAS_WORKDIR=/home/mariab/scratch/$SLURM_JOB_ID
export MOLCAS_MEM=4000
export MOLCAS_NNODES=1
export MOLCAS_NPROCS=2

/home/mariab/build/pymolcas h2o.inp -oe h2o.log -b 1

This is the output

--- Start Module: gateway at Tue Jun 28 14:08:29 2022 ---

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
#0  0x7f73ce15b2ed in ???
#1  0x7f73ce15a503 in ???
#2  0x7f73cb4ddf0f in ???
#3  0x7f73cb901ab3 in ???
#0  0x7fe0bd1f82ed in ???
#1  0x7fe0bd1f7503 in ???
#2  0x7fe0ba57af0f in ???
#3  0x7fe0ba99eab3 in ???
#4  0x55b314f41e38 in comex_init
        at src-mpi/comex.c:1376
#5  0x55b314f41fef in comex_init_args
        at src-mpi/comex.c:1429
#4  0x55f8b49b3e38 in comex_init
        at src-mpi/comex.c:1376
#5  0x55f8b49b3fef in comex_init_args
        at src-mpi/comex.c:1429
#6  0x55b314f3c148 in PARMCI_Init_args
        at src-armci/armci.c:416
#7  0x55b314f14cda in ???
#8  0x55b314e48362 in ???
#9  0x55b314d7f456 in ???
#10  0x55b314d62e92 in ???
#11  0x55b314cf8170 in ???
#12  0x7fe0ba55dc86 in ???
#13  0x55b314cf81b9 in ???
#14  0xffffffffffffffff in ???
#6  0x55f8b49ae148 in PARMCI_Init_args
        at src-armci/armci.c:416
#7  0x55f8b4986cda in ???
#8  0x55f8b48ba362 in ???
#9  0x55f8b47f1456 in ???
#10  0x55f8b47d4e92 in ???
#11  0x55f8b476a170 in ???
#12  0x7f73cb4c0c86 in ???
#13  0x55f8b476a1b9 in ???
#14  0xffffffffffffffff in ???
--------------------------------------------------------------------------
Primary job  terminated normally, but 1 process returned
a non-zero exit code. Per user-direction, the job has been aborted.
--------------------------------------------------------------------------

Any help with this issue will be appreciated.
Thanks

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,011

Re: SIGSEGV: Segmentation fault

Did you use an externally compiled GlobalArrays? Those problems are typically caused by mismatch between the configurations of different libraries. Note that for OpenMolcas you need an 8-byte integer lapack, but a 4-byte integer MPI. You could try using the "built-in" Global Arrays (GA_BUILD=ON)

To rule out possible bugs in seldom-used features, try also without the "Grid input" block.

mcurie747

Member

Registered: 2019-12-12

Posts: 21

Re: SIGSEGV: Segmentation fault

Hi Ignacio, thanks for your response. Yes, I used an externally compiled GA. I just tried with "GA_BUILD=ON" but I got the same issue. I don't know what the problem can be.

mcurie747

Member

Registered: 2019-12-12

Posts: 21

Re: SIGSEGV: Segmentation fault

Hi Ignacio,
The problem in fact was the GA compilation. I downloaded the source code version and I recompiled it. That solved the problem!
Thank you very much