iCESM1_2_0_1_geotrace_n02c Segmentation fault during the run

[liuxiaoqing]

I'm changing PE layout to optimize the the cost and throughput on Purdue's machine bell. an error "Segmentation fault: address not mapped to object at address 0xfffffffc042748d0" showed up when NTASKS of all components were increased to 256. When NTASKS of all components were 128, the simulation was run successfully. May I ask for your suggestion to solve this issue? The attached zip file includes log files and env_mach_pes.xml file.

 

[jedwards]

You have MAX_TASKS_PER_NODE=128 so I would expect 256 tasks to run on two nodes, but
the cesm log suggests that you are using 3 nodes. From that log I see tasks 0-85 on bell-a121
tasks 86-170 on bell-a163 and tasks 171-255 on bell-a206 - 86 tasks per node is very strange.
You may want to check with your system administrators to figure out why this is happening.