"Segmentation fault: address not mapped to object" running DMP solver in mfix-22.2.2

MFiX
#1

Hello everyone:
I used the case (fluid_bed_dem_2d) implemented in the mfix tutorials to test the dmp in the mfix-22.2.2. the operating system is centos.
the dmp solver can be build successful, however, when run the solver, the following error occurred:

image
image1056×530 46 KB

What should I do? Please help!!!
the case is uploaded.
fluid_bed_dem_2d-5.zip (44.9 MB)

#2

Note, I changed the topic subject line to be more specific.

What is your CentOS version? You made a .zip file of the project directory instead of using “Submit bug report” from the main menu, so the system information is not generated, and a lot of unnecessary files (the build subdir) are included. In the future, please use “Submit bug report”.

What DMP domain decomposition are you running with? The .mfx file you attached has nodesi = nodesj = nodesk = 1

It’s possible that the problem is due to stack size limitations as described in this post:

Do you get a core dump? What is the output of ulimit -a ?

Error encountered when submitting tasks to the cluster
#3

Hi, cgw, thank you for your answer.
I used the command line in a HPC, there isn’t a GUI, so which command should I use to get the “sbumit bug report” —.zip file?
image

#4

Hi, cgw, can you help me,please?
I really need your help!!!

#5

“Submit bug report” is not available from the command line - sorry.
Please see the Stack Overvlow link I posted, and send the output of ulimit -a as I asked.

#6

Hi cgw, I checked the HPC rencently, but I still don’t know the reason about the “segmentation fault”.
the output of ulimit -a is as follows:

fbaf59ceb42fce9ce4c7aa55c50549d
fbaf59ceb42fce9ce4c7aa55c50549d1756×788 68.2 KB

and to build the mfix-solver, I used the command:
(mfix-22.2.2) [songchao@cu01 fluid_bed_dem_2d-5]$ build_mfixsolver --batch --dmp -j 10
to run the case, I used the command:
(mfix-22.2.2) [songchao@cu01 fluid_bed_dem_2d-5]$ mpirun -np 2 ./mfixsolver -f fluid_bed_dem_2d.mfx

#7

First, we need to rule out any setting error. Please provide your DMP domain decomposition.

What happens if you do

mpirun -np 2 ./mfixsolver -f fluid_bed_dem_2d.mfx NODESJ=2
#8

You may be running against stack size limitations

#9

Hi, jeff
I used the commmand you provided, and the result is below:

image
image1057×130 7.91 KB

(mfix-22.2.2) [songchao@cu01 fluid_bed_dem_2d-5]$ mpirun -np 2 ./mfixsolver -f fluid_bed_dem_2d.mfx NODESJ=2
[cu01:40745:0:40745] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x3ff00000001)
[cu01:40746:0:40746] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x3ff00000001)

prterun noticed that process rank 0 with PID 40745 on node cu01 exited on
signal 11 (Segmentation fault).

#10

Can you increase the 8k stack size limit (as shown by ulimit -a)? You may need to get your local system admins to help with this.

#11

hi cgw, thanks for your suggestion.
How big should I set this value?

#12

That’s an experimental question. Try 64k.