Single processor running error：Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

[Kobe]

hellow，every professors，
When I run ./wrf.exe with a single processor, a segmentation fault occurs.
I set three layers of nesting, and the simulation time was 48 hours. Every time the simulation reached the first 12 hours, the wrf program terminated and reported an error.
I thought it was due to fnl weather data, so I replaced the July data with the December data, and the error still occurs.
So I tried to run the July and December data with 8 processors, but not only did the error not get resolved, a new error appeared:
-------------- FATAL CALLED ---------------
FATAL CALLED FROM FILE: <stdin> LINE: 774
ZDC + Z0C + 2m is larger than the 1st WRF level - Stop in subroutine urban - change ZDC and Z0C
-------------------------------------------
Abort(1) on node 2 (rank 2 in comm 0): application called MPI_Abort(MPI_COMM_WORLD, 1) - process 2

I then tried to change the e_vert parameter from 35 to 40, but that did not resolve the issue.
I modified the parameters FRC_URB, ROOF_WIDTH, ROAD_WIDTH, Roof level in URBPARM.TBL. I don't know if this is the cause of the error.
The following rsl.zip file is my error log and namelist parameter settings.(The naming format of the error log folder is “simulation time_number of processors”)
This has confused me for a long time, and I look forward to your responses.

 

[kwerner]

Hi,
For the size of your domains, you will likely need to use more than a single processor. Try using several more (as many as you can, based on the size of the smallest domain) to see if that makes a difference. See Choosing an Appropriate Number of Processors for reference.

 

[Kobe]

Hi,
For the size of your domains, you will likely need to use more than a single processor. Try using several more (as many as you can, based on the size of the smallest domain) to see if that makes a difference. See Choosing an Appropriate Number of Processors for reference.

hi，kwerner，
I tried to miprun -np 8 ./ wrf with 8 processors. It ran into an error halfway through.
I captured the cfl in the rsl file and it can still be captured:
d03 2020-07-19_12:41:20 6 points exceeded cfl=2 in domain d03 at time 2020-07-19_12:41:20 hours
d03 2020-07-19_12:41:20 MAX AT i,j,k: 116 70 3 vert_cfl,w,d(eta)= 3.70504260 -94.8293915 9.93639231E-03
-------------- FATAL CALLED ---------------
FATAL CALLED FROM FILE: <stdin> LINE: 774
ZDC + Z0C + 2m is larger than the 1st WRF level - Stop in subroutine urban - change ZDC and Z0C
-------------------------------------------
Abort(1) on node 2 (rank 2 in comm 0): application called MPI_Abort(MPI_COMM_WORLD, 1) - process 2

I tried setting the epssm parameters to 0.5-0.8 and set the time_step to 30, but the problem is still not solved.
My computer has 8*2 processors. Will there be any problems if I use 64 processors?
I don't know if you can reproduce my error. If not, I suspect that there is something wrong with my software.
Looking forward to your reply.

 

[Kobe]

hellow，every professors，

I would like to add that I modified the parameters FRC_URB, ROOF_WIDTH, ROAD_WIDTH, Roof level in URBPARM.TBL. I don't know if this is the cause of the error.This is my URBPARM.TBL.Any help is appreciated! Thanks

 

[kwerner]

Hi,
It shouldn't be a problem to try using 64 processors.

Regarding the CFL errors, try increasing epssm to 0.9 to see if that makes a difference.

As a test, you can try to run this using clean code (i.e., no modifications to the urban parameters you mention above) to see if you are able to get past FATAL error. If so, then you'll know that your modifications are causing the issues and you can play around with those to try to get them right.