Using more nodes than I expect

[ep2764@columbia_edu]

Hi all,I am using a BWCN compset with total pes of 352 on yellowstone. I have set pes per node to 16, thus I expect to be using 22 nodes. However, I keep getting 44 nodes. I am not using multi-threading. Also, the variable COST_PES is set to 704 automatically, which seems to be double of what I would have expected. I am pasting my env_mach_pes.xml file below. Thanks for any help/advice!
-Ethan 

 

[jedwards]

Everything looks okay in this file except for the ROOTPE_LND which shoud be 0 to avoid overlap with the ice model.If you send me a pointer to your case directory on yellowstone I can take a closer look.

 

[ep2764@columbia_edu]

The case directory is: /glade/u/home/epeck/case/CO2x1SvarEmin_yBWCNI have added read permissions so you can see it. Thanks for your assistance!-Ethan

 

[jedwards]

In your run script the line: #BSUB -R "span[ptile=8]"controls the number of mpi tasks per node.   If you change that to 16, you will run on the correct number of nodes.  
We have found that using 15 tasks per node on yellowstone is generally faster than 16 tasks per node for CESM and I think that in implementing the 15 tpn setup we introduced a bug for 16 tpn.