Code:
$ ./describe_version
ccs_config at tag ccs_config_cesm0.0.109
M HEAD detached at 797acd7
Changes not staged for commit:
(use "git add <file>..." to update what will be committed)
(use "git restore <file>..." to discard changes in working directory)
modified: machines/config_batch.xml
modified: machines/config_machines.xml
no changes added to commit (use "git add" and/or "git commit -a")
share at tag share1.0.19
cime at tag cime6.0.246
mct at tag MCT_2.11.0
mpi-serial at tag MPIserial_2.5.0
cam at tag cam6_3_162
M HEAD detached at ab476f9b
Changes not staged for commit:
(use "git add/rm <file>..." to update what will be committed)
(use "git restore <file>..." to discard changes in working directory)
deleted: cime_config/testdefs/testmods_dirs/cam/outfrq9s_waccm_ma_mam4/shell_commands
deleted: cime_config/testdefs/testmods_dirs/cam/outfrq9s_waccm_ma_mam4/user_nl_cam
deleted: cime_config/testdefs/testmods_dirs/cam/outfrq9s_waccm_ma_mam4/user_nl_clm
no changes added to commit (use "git add" and/or "git commit -a")
ww3 at tag ww3i_0.0.2
rtm at tag rtm1_0_79
pysect at tag 3.2.2
mosart at tag mosart1_0_49
mizuroute at tag cesm-coupling.n02_v2.1.2
fms at tag fi_240516
parallelio at tag pio2_6_2
cdeps at tag cdeps1.0.37
cmeps at tag cmeps0.14.63
cice at tag cesm_cice6_5_0_9
cism at tag cismwrap_2_2_001
clm at tag ctsm5.2.007
mom at tag mi_240522
testfails = 0, local mods = 2, needs updates 0
The submodules labeled with 'M' above are not in a clean state.
The following are options for how to proceed:
(1) Go into each submodule which is not in a clean state and issue a 'git status'
Either revert or commit your changes so that the submodule is in a clean state.
(2) use the --force option to git-fleximod
(3) you can name the particular submodules to update using the git-fleximod command line
(4) As a last resort you can remove the submodule (via 'rm -fr [directory]')
then rerun git-fleximod update.
./create_newcase --case /glade/derecho/scratch/jbuzan/cases/Test_24_nodes --compiler intel --compset F2000climo --res ne30pg3_ne30pg3_mg17 --driver nuopc --mpilib mpich --run-unsupported --mach derecho
Changes to env_mach_pes (timing files below).
Describe your problem or question:
I am attempting to increase the number of cores for the F2000 case for scaling tests on Eiger and Derecho.
Eiger and Derecho are all but the same machine, and their performance is almost the same between the same processor and case setup.
However, when I increase the number of cores from 12 nodes to 24 nodes (with ATM = CPL), the ATM Run Time increases, but the TOT Run Time does barely changes. My understanding is that the model should strongly scale up to the "element" number of the grid.
Example:
24 nodes
TOT = 18.37 myears/wday ATM = 33.73 myears/wday
12 nodes
TOT = 14.79 myears/wday ATM = 19.18 myears/wday
6 nodes
TOT = 9.33 myears/wday ATM = 10.61 myears/wday
Is this the expected behavior for CESM? This is showing a strong disconnect between the ATM performance and the TOT model performance. Detailed timing files below.
Thank you for your help!
Cheers,
-Jonathan
Example on
Derecho with 24 nodes to demonstrate the same performance as Eiger.
Eiger with 24 nodes
Eiger with 12 nodes
Eiger with 6 nodes
