An efficient thread partition policy for secure functional language

A parallel computer, which is loaded with processor in numerous commercial chips, is dropped performance due to the delayed time of seeing the increasing memories and to a burden for synchronization. To solve this, multithreaded is being used. Thread is partition by searching for a part available for performing sequentially when executing Non-Strict program in multithreaded. However, the existing partition divided the conditional expression by having predicate, true expression and false expression as the basic blocks, and applied partition to each. Therefore, the context switching frequency between synchronization and thread grew. This study proposes the secure thread partition policy in the conditional expression of dividing Non-Strict program into thread. Thread partition policy can lead to knowing the sequence of performance in the time of translation, thereby being what collects it with thread by distinguishing a part of program available for the static scheduling. As a result of experiment, the context switching frequency and the synchronization were reduced by generating big thread in the conditional expression compared to the existing partition. Accordingly, the performance of program in multithreaded model was improved.