In this section there are some traces that show in which MPI state threads are. For example: running, waiting, sending, etc. We just analyzed one coupling iteration due to the fact that this kind of trace is very huge and, as a consequence, difficult to handle it. Note that the original trace contains more information, like IPC, useful duration, cycles, etc. The trace is available on /gpfs/projects/bsc32/share/traces/EC-Earth_3.2beta/complete_1cplng_iter_with_northfold_opt_ifs-128_nemo-128 (MareNostrum).

Configuration:

• Coupled EC-Earth 3.2beta: IFS, NEMO 3.6, LIM3, XIOS, Runoff-mapper, OASIS3-MCT
• 128 NEMO tasks
• 128 IFS tasks
• 1 Runoff-Mapper task
• 1 xios task
• 1 day of simulation.
• MareNostrum III, 16 tasks/node
• T255L91-ORCA1L46