Title: Developments in memory management in OpenMP

Authors: Jason Sewall; S. John Pennycook; Alejandro Duran; Christian Terboven; Xinmin Tian; Ravi Narayanaswamy

Addresses: Intel Corporation, 3600 Mission College Boulevard, Santa Clara, California 95050, USA ' Intel Corporation, 3600 Mission College Boulevard, Santa Clara, California 95050, USA ' Intel Corporation, 3600 Mission College Boulevard, Santa Clara, California 95050, USA ' High Performance Computing (I12), RWTH Aachen University, Seffenter Weg 23 52074 Aachen, Germany ' Intel Corporation, 3600 Mission College Boulevard, Santa Clara, California 95050, USA ' Intel Corporation, 3600 Mission College Boulevard, Santa Clara, California 95050, USA

Abstract: Modern computers with multi-/many-core processors and accelerators feature a sophisticated and deep memory hierarchy, potentially including distinct main memory, high-bandwidth memory, texture memory and scratchpad memory. The performance characteristics of these memories are varied, and studies have demonstrated the importance of using them effectively. In this article, we explore some of the major issues in developing software to effectively and portably implement these technologies and describe enhancements being added to the OpenMP language to bridge this software-hardware gap. Our proposal separately exposes the characteristics of memory resources (such as kind) and the characteristics of allocations (such as alignment), and is fully compatible with existing OpenMP constructs.

Keywords: memory management; programming languages; compiler directives; heap allocation.

DOI: 10.1504/IJHPCN.2019.097052

International Journal of High Performance Computing and Networking, 2019 Vol.13 No.1, pp.70 - 85

Available online: 11 Dec 2018 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article