Supporting incremental parallelization

OpenMP has become the de facto mainstream standard of parallel programming on shared memory processors. To enable parallel programming with OpenMP standard on multicore DSP

a compiler is designed for OpenMP programs running on multicore DSP. Our main work consists of a translator and runtime. The translator part converts the OpenMP directives into the function interface in runtime where the concrete implementation is provided. How to design the tactic to distribute computing task among the cores is the main concern of parallel compilation

which corresponds to the concept of parallel region in OpenMP. By means of the translation of parallel directives and support provided by runtime

parallel executing of master and slave cores is achieved

which offers instructive meaning to the design of parallel compilation.