Reducing task creation and termination overhead in explicitly parallel programs

Abstract

There has been a proliferation of task-parallel programming systems to address the requirements of multicore programmers. Current production task-parallel systems include Cilk++, Intel Threading Building Blocks, Java Concurrency, .Net Task Parallel Library, OpenMP 3.0, and current research task-parallel languages include Cilk, Chapel, Fortress, X10, and Habanero-Java (HJ). It is desirable for the programmer to express all the parallelism intrinsic to their algorithm in their code for forward scalability and portability, but the overhead incurred by doing so can be prohibitively large in today's systems. In this paper, we address the problem of reducing the total amount of overhead incurred by a program due to excessive task creation and termination. We introduce a transformation framework to optimize task-parallel programs with finish, forall and next statements. Our approach includes elimination of redundant task creation and termination operations as well as strength reduction of termination operations (finish) to lighter-weight synchronizations (next). Experimental results were obtained on three platforms: a dual-socket 128-thread (16-core) Niagara T2 system, a quad-socket 16-way Intel Xeon SMP and a quad-socket 32-way Power7 SMP. The results showed maximum speedup of 66.7×, 11.25× and 23.1× respectively on each platform and 4.6×, 2.1× and 6.4× performance improvements respectively in geometric mean related to non-optimized parallel codes. The original benchmarks in this study were written with medium-grained parallelism; a larger relative improvement can be expected for programs written with finer-grained parallelism. However, even for the medium-grained parallel benchmarks studied in this paper, the significant improvement obtained by the transformation framework underscores the importance of the compiler optimizations introduced in this paper.

Improving the Performance of X10 Programs by Clock Removal

“…al. [20] in which an algorithm, which amount to moving parallel loops outside sequential loops with barrier insertion is proposed. Our contention here is that moving in the opposite direction may be beneficial in some cases.…”

A Transformation Framework for Optimizing Task-Parallel Programs

“…In this section, we present experimental results for evaluating the transformation framework described in this article using the HJ compiler and runtime system [Habanero 2009;Zhao et al 2010]. All transformations were performed using the rules in Section 5, which assume the possibility of exceptions.…”

“…In a previous conference submission[Zhao et al 2010], we referred to the forall-coarsening phase as forall-distillation.ACM Transactions on Programming Languages and Systems, Vol. 35, No.…”

Intermediate language extensions for parallelism

“…The High level PIR (HPIR) has a hierarchical structure that's akin to the AST for many concurrency constructs, but flattens non-loop sequential constructs such as if-then-else statements. This level is suitable for efficient high level analyses such as May Happen in Parallel (MHP) [3], and to high level transformations such as forall chunking with phasers [34], and elimination of redundant async operations and strength reduction of termination finish operations to lighter-weight barrier (next) synchronizations [41].…”

“…A new methodology for reducing the overhead of both task creation and termination by SPMDizing parallel loops (e.g., forall or finish + foreach) was introduced in [42]. The algorithm encountered the RST in a bottom-up manner; for each parallel loop, it applies transformation rules in a top-down manner to expose redundant termination operations and enlarge the granularity of parallelism.…”