' CCSM/CAM Unification Plan - need for single executable system - only single executable? - concern that single executable in robust on all platforms - multiple machine runs may prefer single executable - performance multiple executable vs single executable - acceptance criteria, ease of use, performance, run-time environment - unified model has same climate as CAM stand-alone in some mode, need to determine method for moving CCSM and CAM forward - new components, merged components - NO components will contain other components (for atm, ocn, ice, lnd) runoff is acceptable as currently implemented? - we need a way to instantiate components and add them easily to the system - hierarchy of components - thermo-only ice model - full ice model - data ice model - SOM - full ocean model - mixed layer ocean model - upper ocean model - data ocean model (SST) - active land (runoff, glaciers) - data land model - aqua planet mode - data assimilation models - transport only CAM - CAM adjoint model - regional modeling - ensemble components CAM internal components (some on different grids) - chemistry, mozart - single column model - WACCM - super parameterizations - regional modeling - data aerosol/prognostic aerosol models CPL component - how many? Components - science - instantiate components for testing - code looks better - is component shared program, is component visible only at main.F, is component something that meets structure at top and can be any hierarchy - need to start drawing diagrams - building blocks of components, flexibility is critical - should we hardwire number of top components like CCSM now (probably not) - is it something that talks through a coupler - coupling strategies - ESMF - flux averaging - coupling every timestep, coupling frequency flexibility? - call and return for coupling? - CAM bc/ac ordering, extra 1/2 timestep at end? - update bcs every timestep in CAM - flux computation flexibility, where do they live in components - model timestep and coupling frequency relationship - serial vs concurrent execution - is serial required for 1 pe mode, does it have to work without MPI available? - do we need several different CCSM drivers for different applications of the unified model? - ESMF and mph work together on virtual machine, hierarchy, system needs, mediation of system resources between components and at different hierarchys - lightweight components for trivial coupling - resolution support - all current CAM resolutions (T5-T340, FV 2,1,.5,.25, etc) - some ocean/ice/land configurations will need to run on the CAM grid (will need identity mapping in cpl) - should CLM be able to run at a different resolution? - should POP and CSIM be able to run at different resolutions? - land fractions? specified land fraction vs complement of ocean, are both required? - common land fraction mechanism - land fractions are a critical issue - limited grids, regional grids - mesh refinement inside components - flexible grid at boot-up - hardware support - CCSM platforms - Lahey strict compliance - laptop - scientific requirements for overall system - WACCM impact on system? - BGC impact on system? - Chemistry impact on system? - support for varied coupling fields - SE requirements/desires - bfb capability with different pes only for certain configurations? as an optional capability? - flexibility wrt coupling frequency, coupling concurency, grids, number of components - single pe - error growth, validation - (run time) dynamics load balance, dynamics decompositions = structures - flexible statically allocated load balance at initialization, automatically - dynamic grid refinement (probably not) - dynamic load balance needs to be enabled in order to test - science requirement for load balance, dynamic vegetation for instance - SE infrastructure, scripts, etc. - common scripts in unified model, tools for building and generating namelist, etc. - common test framework, complete test framework, flexible test framework - flint, lahey strict, build checks - scripting practices