NUMA Optimizations for Algorithmic Skeletons

Abstract

To address NUMA performance anomalies, programmers often resort to application specific optimizations that are not transferable to other programs, or to generic optimizations that do not perform well in all cases. Skeleton based programming models allow NUMA optimizations to be abstracted on a pattern-by-pattern basis, freeing programmers from this complexity. As a case study, we investigate computations that can be implemented with stencil skeletons. We present an analysis of the behavior of a range of simple and complex stencil programs from the NAS and Rodinia benchmark suites, under state-ofthe-art NUMA aware page placement (PP) schemes. We show that even though an application (or skeleton) may have implemented the correct, intuitive scheduling of data and work to threads, the resulting performance can be disrupted by an inappropriate PP scheme. In contrast, we show that a NUMA PP-aware stencil implementation scheme can achieve speed ups of up to 12x over a similar scheme which uses the Linux default PP, and that this works across a set of complex stencil applications. Furthermore, we show that a supposed PP performance optimization in the Linux kernel never improves and in some cases degrades stencil performance by up to 0.27x and should therefore be deactivated by stencil skeleton implementations. Finally, we show that further speed ups of up to 1.1x can be achieved by addressing a work imbalance issue caused by poor conventional understanding of NUMA PP.

Enforcing Deadlines for Skeleton-based Parallel Programming

“…Firstly, we exploit inter-job parallelism whereas the authors exploit parallelism within jobs. Secondly, we are the first to demonstrate that structural information encoded in skeletons can be used for performance improvements in the context of real-time systems as demonstrated before in other fields [16], [30], [31], [32]. Thirdly, we are the first to show that skeletons can be used for fine grained analytical application models.…”

“…Benefits of skeletons are more efficient implementations and increased programmer productivity. Structural information encoded in skeletons can be used for implementations with lower execution time, lower energy consumption or higher throughput [30], [31], [32]. Skeletons improve programmer productivity by off-loading the error prone task of writing and tuning low-level parallel code to library or compiler developers [9].…”

PsmArena: Partitioned shared memory for NUMA-awareness in multithreaded scientific applications

“…However, these OSbased approaches can suffer from false page-sharing and fragmentation, as well as imperfect OS optimization as suggested by this paper and in algorithmic skeletons by Ref. [18].…”

JArena: Partitioned Shared Memory for NUMA-awareness in Multi-threaded Scientific Applications

“…Beyond first-touch, advanced page placements such as the numactl API [13] and automatic page migration [8] and balancing [9] are introduced in the Linux kernel. However, these OS-based approaches can suffer from false page-sharing and fragmentation, as well as imperfect OS optimization as suggested by this paper and in algorithmic skeletons by [18].…”