The Foundations for Scalable Multicore Software in Intel Threading Building Blocks

Abstract

This paper describes two features of Intel ® Threading Building Blocks (Intel ® TBB) [1] that provide the foundation for its robust performance: a work-stealing task scheduler and a scalable memory allocator. Work-stealing task schedulers efficiently balance load while maintaining the natural data locality found in many applications. The Intel TBB task scheduler is available to users directly through an API and is also used in the implementation of the algorithms included in the library. In this paper, we provide an overview of the TBB task scheduler and discuss three manual optimizations that users can make to improve its performance: continuation passing, scheduler bypass, and task recycling. In the Experimental Results section of this paper, we provide performance results for several benchmarks that demonstrate the potential scalability of applications threaded with TBB, as well as the positive impact of these manual optimizations on the performance of fine-grain tasks. The task scheduler is complemented by the Intel TBB scalable memory allocator. Memory allocation can often be a limiting bottleneck in parallel applications. Using the TBB scalable memory allocator eliminates this bottleneck and also improves cache behavior. We discuss details of the design and implementation of the TBB scalable allocator and evaluate its performance relative to several commercial and non-commercial allocators, showing that the TBB allocator is competitive with these other allocators. Intel Corporation uses the Palm OS ® Ready mark under license from Palm, Inc.

Comparison of OpenMP 3.0 and Other Task Parallel Frameworks on Unbalanced Task Graphs

“…In addition to OpenMP 3.0, there are currently several other task parallel languages and libraries available to developers, Microsoft Task Parallel Library [19] for Windows, Intel Thread Bulding Blocks (TBB) [5], and Intel Cilk++ [4], a younger sibling of Cilk based on C++ rather than C. We will use TBB and Cilk++, along with Cilk, as comparison points for our performance evaluation with OpenMP 3.0.…”

“…In this paper, we compare the performance and scalability of the UTS benchmark on several different OpenMP 3.0 compiler and run time implementations (Intel icc 11, gcc 4.4, Mercurium 1.2, SunStudio 12). For comparison we also examine the performance of the UTS benchmark using Cilk [3] , Intel Cilk++ [4], Intel Thread Building Blocks [5]. We also compare with an OpenMP implementation without tasks that performs all scheduling, load balancing, and termination detection explicitly [6].…”

A mean field model of work stealing in large-scale systems

“…This shows that x t j ≤ 1. Since x t j is increasing in t, it converges to some x ∞ j that satisfies (11)- (12).…”

Mechanisms that separate algorithms from implementations for parallel patterns

“…There are many different libraries that address parallelism in different ways [18,26,21]. For the purposes of this evaluation, we choose STAPL, the Structured Template Adaptive Parallel Library [4], to serve as an exemplar.…”