Optimized code generation for heterogeneous computing environment using parallelizing compiler TINPAR

This paper presents a compiling technique to generate optimized codes for heterogeneous computing environment. This paper also proposes a new dynamic load redistribution mechanism which can adaptively and dynamically distribute tasks among computers according to their available computing power which may vary during the computation. As the results of the performance evaluation, we could confirm that the generated codes are effectively executed in a heterogeneous computing environment with dynamic load change