Friday - March 21, 2008
12:00 noon Doug
Sondak and Kadin Tseng Office
of Information Technology
to Scientific Computing"
scientific computing often requires large amounts
of CPU time and/or memory, hindering the ability
of researchers to examine systems of the desired
size or resolution. We will give a broad-brush
view of scientific computing, discussing some
of the associated issues and ways in which they
are tackled. This will include some basics of
parallelization, hardware-related performance
issues, and a description of the Scientific
Computing Facility (SCF) at Boston University.
The use of MPI to parallelize a Fortran code
will be discussed, showing the resulting performance
improvement. A brief introduction to high-performance
computing with MATLAB will also be covered.