Abstract:
The strategies and tactics for the parallelization of various stages of large-scale
computational experiments are considered: geometric and functional modeling,
grid generation, approximation of the original problem, solving algebraic systems,
postprocessing and visualization of results. The main tools are the domain decomposition
and the mapping of algorithms onto the multiprocessor computer architecture.
The principles of parallelization are described in the framework of a Basis System
of Modeling intended to solve large-scale problems on high-performance supercomputers
of new generations.
Keywords:parallelization of algorithms; mathematical modeling; technological stages; domain decomposition; multiprocessor and multicore supercomputers; large-scale problems.