Global arrays: A nonuniform memory access programming model for high-performance computers

Nieplocha, Jaroslaw; Harrison, Robert J.; Littlefield, Richard J.

doi:10.1007/bf00130708

Cited by 259 publications

(166 citation statements)

References 23 publications

Supporting

Mentioning

161

Contrasting

Order By: Relevance

“…Particularly, the system consisted of the following three levels in the memory hierarchy: disk, global memory, and the local or device memory. The global memory corresponds to the use of global arrays [27] on the cluster. The device memory is the memory on the GPU, which had a similar size as the local memory on each node.…”

Section: Experimental Evaluationmentioning

confidence: 99%

Practical Loop Transformations for Tensor Contraction Expressions on Multi-level Memory Hierarchies

Krishnamoorthy

Agrawal

2011

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Modern architectures are characterized by deeper levels of memory hierarchy, often explicitly addressable. Optimizing applications for such architectures requires careful management of the data movement across all these levels. In this paper, we focus on the problem of mapping tensor contractions to memory hierarchies with more than two levels, specifically addressing placement of memory allocation and data movement statements, choice of loop fusions, and tile size selection. Existing algorithms to find an integrated solution to this problem even for two-level memory hierarchies have been shown to be expensive. We improve upon this work by focusing on the first-order cost components, simplifying the analysis required and reducing the number of candidates to be evaluated. We have evaluated our framework on a cluster of GPUs. Using five candidate tensor contraction expressions, we show that fusion at multiple levels improves performance, and our framework is effective in determining profitable transformations.

show abstract

Section: Experimental Evaluationmentioning

confidence: 99%

Practical Loop Transformations for Tensor Contraction Expressions on Multi-level Memory Hierarchies

Krishnamoorthy

Agrawal

2011

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…Their stated aim was to produce a portable and scalable parallel program but, at the same time, to minimise the number of di erences between the sequential and parallel codes to aid in maintainability. To ensure portability, the parallelisation was based on use of the Global Arrays [4].…”

Section: Molpromentioning

confidence: 99%

“…As was discussed, programming the VPP series requires consideration of the vector unit e ciency in addition to the normal issues associated with the communication latency and bandwidth for a distributed-memory machine. Computational chemistry applications are normally coded using explicit message passing (such as MPI [2]) or by using a Ôglobal-memoryÕ model (such as Linda [3] or the Global Arrays [4]) which e ectively hides the underlying message passing from the user.…”

Section: Introductionmentioning

confidence: 99%

Computational chemistry on Fujitsu vector–parallel processors: Development and performance of applications software

et al. 2000

View full text Add to dashboard Cite

In this and a preceding paper, we provide an introduction to the Fujitsu VPP range of vector±parallel supercomputers and to some of the computational chemistry software available for the VPP. Here, we consider the implementation and performance of seven popular chemistry application packages. The codes discussed range from classical molecular dynamics to semiempirical and ab initio quantum chemistry. All have evolved from sequential codes, and have typically been parallelised using a replicated data approach. As such they are well suited to the large-memory/fast-processor architecture of the VPP. For one code, CASTEP, a distributed-memory data-driven parallelisation scheme is presented. Ó 2000 Published by Elsevier Science B.V. All rights reserved.

show abstract

“…The development of these tools have been motivated by algorithmic requirements of theoretical chemistry electronic structure computations (dynamic load balancing with up to three orders of magnitude variation in task size, blocked access patterns to distributed dense arrays) and NUMA characteristics of high-performance computers [19,24]. The Global Arrays toolkit implements shared-memory NUMA model for distributed dense arrays.…”

Section: Tools For Explicit Management Of Memory Hierarchymentioning

confidence: 99%

“…The Global Arrays library [19,24] implements a shared-memory programming model in which data locality is managed explicitly by the programmer. This management is achieved by explicit calls to functions that transfer data between a global address space (a distributed array) and local storage.…”

Section: Global Arraysmentioning

confidence: 99%

Explicit Management of Memory Hierarchy

Nieplocha

Harrison

Foster

1997

Advances in High Performance Computing

View full text Add to dashboard Cite

show abstract

Global arrays: A nonuniform memory access programming model for high-performance computers

Cited by 259 publications

References 23 publications

Practical Loop Transformations for Tensor Contraction Expressions on Multi-level Memory Hierarchies

Practical Loop Transformations for Tensor Contraction Expressions on Multi-level Memory Hierarchies

Computational chemistry on Fujitsu vector–parallel processors: Development and performance of applications software

Explicit Management of Memory Hierarchy

Contact Info

Product

Resources

About