A library for parallel thread-level speculation

Abstract

In this paper we describe the design and implementation of a library (based on pthreads and C + +) to use explicit parallelism thread-level speculation. The library consists of several models proposed for speculation, guaranteeing sequential consistency: decision structures speculation (where the branches are executed in parallel to the previous section), loops speculation (where all iterations are run in parallel), critical sections speculation, and speculation by cooperating threads. The library was evaluated through case studies taken from SPEC CPU 2000, allowing acceleration of about 1.5 x for a multicore architecture (four cores) with coarse-grained multithreaded.