Opened 4 years ago

Closed 3 years ago

#116 closed Defect (fixed)

FLEXPART-WRF v3.2 runtime segmentation fault

Reported by: tauhonson Owned by: jbrioude
Priority: major Milestone: FLEXPART_WRF_3.2_FPbase_9
Component: FP other Version: FLEXPART-WRF
Keywords: Segmentation fault Cc:

Description

Hi,everyone:
I could successfully compile FLEXPART-WRF v3.2 with intel,
the exe of flexwrf32_intel_serial works just fine to finish simulation;
however, when I type mpiexec -n 6 ./flexwrf32_intel_mpi flex..., it stops soon with error message:


Allocating fields for nested and global output (x,y): 100 100

====================================================================

BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES

EXIT CODE: 11

CLEANING UP REMAINING PROCESSES

YOU CAN IGNORE THE BELOW CLEANUP MESSAGES

====================================================================
YOUR APPLICATION TERMINATED WITH THE EXIT STRING: Segmentation fault (signal 11)

I previously used FLEXPART-WRF v3.2 with pgi13.7 smoothly in another cluster. I have to build FLEXPART-WRF v3.2 in this new cluster which is not installed with pgi. I would appreciate it if someone do me a favor.

Change History (4)

comment:1 follow-up: Changed 4 years ago by jbrioude

  • Owner changed from somebody to jbrioude
  • Status changed from new to accepted

Hi,
It would be good to test it in serial mode first (in debug mode would be even better).
Can you also check the size of the 3D fields in the WRF output you are trying to read, and compare it to the max size value you have in par_mod.f90?

comment:2 in reply to: ↑ 1 Changed 4 years ago by tauhonson

Replying to jbrioude:

Hi,
It would be good to test it in serial mode first (in debug mode would be even better).
Can you also check the size of the 3D fields in the WRF output you are trying to read, and compare it to the max size value you have in par_mod.f90?

Hi
The serial mode compiled by Intel works smoothly,and the wrfout file is 107 G.

comment:3 Changed 4 years ago by jbrioude

Okay.
Make sure you properly setup the number of OMP cores per node. Ideally, you should use one MPI instance per node, and as many OMP threads as cores available per node.
If you have set it up correctly, then it is most likely a memory problem. 107Gb per WRF output is huge.
You should compile flexpart in debug mode and see if you get any meaningful error message.

comment:4 Changed 3 years ago by jbrioude

  • Resolution set to fixed
  • Status changed from accepted to closed
Note: See TracTickets for help on using tickets.
hosted by ZAMG