The performance of the GCM ECHAM5/MESSy

The runtime overhead due to the MESSy infrastructure is in the order of 3% to 5%, provided that the model setups (ECHAM5 versus ECHAM5/MESSy) are honestly harmonised:

• the comparison is made with ECHAM5 version 5.3.01, which is the basemodel of the current ECHAM5/MESSy version,
• all MESSy submodels (except for RAD4ALL, CLOUD, CONVECT) are switched OFF
• the radiation calculation is triggered with the same time interval: trigrad4all = 2,'hours','first',0 in namelist RADCTL (ECHAM5) and in namelist file rad4all.nml (ECHAM5/MESSy),
• the sub-submodel TRACER_PDEF (which does not exist in ECHAM5) is switched OFF: L_PDEF = F in tracer.nml,
• the output in the ECHAM5/MESSy setup is reduced, or switched off completely in both model setups for runtime performance tests. (Note: ECHAM5/MESSy outputs in the default setup about 3 times the amount of data as ECHAM5.)