Limits of task-based parallelism in irregular applications

Abstract

Traditional parallel compilers do not effectively parallelize ir-regular applications because they contain little loop-level parallelism. We explore Speculative Task Parallelism (STP), where tasks are full proce-dures and entire natural loops. Through profiling and compiler analysis, we find tasks that are speculatively memory-and control-independent of their neighboring code. Via speculative futures, these tasks may be executed in parallel with preceding code when there is a high probability of independence. We estimate the amount of STP in irregular appli-cations by measuring the number of memory-independent instructions these tasks expose. We find that 7 to 22% of dynamic instructions are within memory-independent tasks, depending on assumptions.