Abstract:
In order to continue to achieve the computational throughput of present-day simulations, the next version of the Community Climate System Model (CCSM) will need to be modified to use existing computing systems more efficiently and optimized for next-generation massively parallel systems. This will require improving scalability to thousands of processors and carrying out more comprehensive performance analysis in order to attain that goal. We plan to (1) improve scalability of the atmosphere and ocean component models with respect to processor count; (2) investigate and optimize performance at high spatial resolution; (3) improve scalability of the full coupled model through improved load balance; to (4) port to and optimize on next generation high performance computing systems; and (5) improve I/O performance of the component models through parallel I/O. |
