Parallel execution scheme of conditional branches with graph unfolding for the parallel processing system - Harray

The purpose of this paper is to propose and evaluate a new scheme, called the Preceding Activation Scheme with Graph Unfolding, which translates a FORTRAN program into a dataflow graph and executes it efficiency. The problems in restructuring a FORTRAN program into a dataflow graph is that a FORTRAN program has an explicit control flow, which results in little parallelism because many gate-operations, such as T/F gates, are introduced in the dataflow graph to synchronize the data mevement. Thus, discarding these gate-operations is the key to expose parallelism from a FORTRAN program, which is the main purpose of the proposed scheme. In the software simulation, it is shown that the execution speed with the proposed scheme for flow graphs without backward branches is about 1.5 times as fast as that of the pure dataflow computer. Moreover, the execution speed is 2.7 times as fast as that of the pure dataflow computer if a flow graph including backward branches is unfolded by the proposed scheme.