Segmentation fault with floats

[lalvarez]

Hello to All

I had a problem with the floats in my run, it is a short time 40 days and a release floats 10, 15 and 20 days after the model starts, here is what I got. The total number of floats are 8400.

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 #0 0x7FE2D7CC7777
0x7FE2D7CC7777
#1 0x7FE2D7CC7D7E
#1 0x7FE2D7CC7D7E
#2 0x7FE2D6F8AD3F
#2 0x7FE2D6F8AD3F
#3 #3 0x51E10E in __interp_floats_mod_MOD_interp_floats
0x51E10E in __interp_floats_mod_MOD_interp_floats
#4 #4 0x4B4B2D in __step_floats_mod_MOD_step_floats_tile at 0x4B4B2D in __step_floats_mod_MOD_step_floats_tile at step_floats.f90:?
step_floats.f90:?
#5 #5 0x4B9164 in __step_floats_mod_MOD_step_floats0x4B9164 in __step_floats_mod_MOD_step_floats

#6 #6 0x44D832 in main3d_0x44D832 in main3d_

#7 0x7FE2D753F349
#8 0x7FE2D7321181
#7 0x407089 in __ocean_control_mod_MOD_roms_run
#9 0x7FE2D704E47C
#10 0xFFFFFFFFFFFFFFFF
Violación de segmento


Can anyone give a hand on this
Thanks in advance

[kate]

The first step is to recompile in USE_DEBUG mode. The resulting output should look similar but include line numbers for the .f90 files. You then need to look at the appropriate .f90 file and see what is at that line. Next would be to perhaps add print statements around there to see what is in the relevant variables.

[lalvarez]

Hello Kate

Thanks a lot for your ASAP answer, I'm working on it

Have a nice day

[wilkin]

When does the seg fault occur?

At model initialization, at first float release, or midway in the run?

[arango]

Segmentation violations (SIGSEGV) are usually a sign of memory problems on the computer and available stack for the run. Perhaps, your computer doesn't have too much memory for your application, and you are deploying too many floats for it. I will try to run without floats to see if it works. Then add floats but put a few.

[lalvarez]

Hello Hernan

Actually the model runs ok until the day it has to deploy the floats, I'll try to set fewer floats I guess that's my problem

Thank alot

[kate]

I suspect that the problem is not the number of floats but something else. The way to debug these things is to save a restart record at the time of the first float release, or just before. Then do your tests in a restart run so that it blows up right away and you're not running forever in debug mode.