Next:
Contents
Contents
CCSM Coupler Architecture
Version 6, the ``Next Generation'' Coupler
CCSM Coupler 6 Development Team
DRAFT June 18, 2001, DRAFT
Contents
1 Synopsis
2 Requirements
2.1 Scientific Requirements
2.1.1 Component Model Control
2.1.2 Flexibility
2.1.3 Extensibility
2.1.4 Time Stepping Method
2.1.5 Computations
2.1.6 Grids
2.1.7 Fault Detection
2.2 Computational Requirements
2.2.1 Target Platforms
2.2.2 Language
2.2.3 Parallel Implementation
2.2.4 Parallel Reproducibility
2.2.5 Sequencing
2.2.6 Performance
2.2.7 Single Vs. Multiple Executables
2.2.8 Processor Allocation
2.2.9 Communications
2.2.10 Mapping
2.2.11 I/O
2.2.12 Fault Detection
2.3 Imposed Functionality Requirements
2.3.1 Single vs. Multiple Executable
2.3.2 Fault Detection
2.3.3 Handshaking
3 System overview
4 Design considerations
4.1 The coupler, MCT, and MPH
4.2 Processor layout
4.3 Single vs. multiple executables
4.4 Control and sequencing
4.5 F90 source code
4.6 Staged delivery development
4.7 Future Considerations
4.7.1 SPMD Configuration
4.7.2 Component References
4.7.3 Wrapper Code Functionality
5 System architecture
5.1 Layering strategy
5.2 Layer 1a: Command and control
5.2.1 Main event loop and sequencing
5.2.2 Control module
5.2.3 Data declarations
5.3 Layer 1b: major subsystems and modules
5.3.1 Mapping module
5.3.2 Message passing module
5.3.3 Merge module
5.3.4 Flux module
5.3.5 Restart module
5.3.6 History module
5.3.7 Diagnostic module
5.4 Layer 1c: Shared data types and utilities
5.4.1 Data type definitions
5.4.2 Calendar module
5.5 Layers 2-5: MCT and other external libraries
5.6 MCT wrapper code provided to component models
5.7 Distributed and threaded multi-processing
6 Review Status
7 Glossary
About this document ...
csm@ucar.edu
