Coarse Grain Task Parallel Processing on a Shared Memory Multiprocessor System

摘要

This paper proposes an implementation method named "one-time single level tread gen- eration" for a coarse grain task parallel processing scheme on an off the shelf shared memory multiprocessor system. The coarse grain task parallel processing is important to improve the effective performance of wide range of multiprocessor systems from a single chip multiproces- sor to a high performance computer beyond the limit of the loop parallelism.