Evaluation of a Speculative Multithreading Compiler by Characterizing Program Dependences

Speculative multithreading (SpMT) promises to be an eflective mechanism for paralleliring non-numeric pro- grams. Proper thread formation is crucial for obtain- ing good speedup in an SpMT system. We have de- veloped an SpMT compiler framework for partitioning sequential programs into multiple threads. Since con- trol and data speculations are the essence of SpMT execution model, inter-thread data dependences and inter-thread control predictions at run-time play cm- cia1 roles in aflecting the performance of the SpMT system. Therefore, to evaluate existing SpMT tom- piter or hardware systems, and to design more efi- cient systems it is necessary to characterize the dy- namzc program dependences carefully. In this paper, we have studied the run-time behaviors of inter-thread data and control dependences of the threads generated by our compiler in detail and used that for analyzing the performance. The analyses reveal that our com- piler has successfully modeled the dnter-threod data and control dependences of non-numeric applicataons and minimized them while generating the threads.