Last edited by Zolozshura
Tuesday, May 5, 2020 | History

2 edition of Automatic computation and data partitioning on scalable shared memory multiprocessors found in the catalog.

Automatic computation and data partitioning on scalable shared memory multiprocessors

Sudarsan Tandri

Automatic computation and data partitioning on scalable shared memory multiprocessors

by Sudarsan Tandri

  • 223 Want to read
  • 5 Currently reading

Published by University of Toronto, Dept. of Computer Science in Toronto .
Written in English


Edition Notes

Thesis (Ph.D.)--University of Toronto, 1997.

StatementSudarsan Tandri.
The Physical Object
Pagination128 p.
Number of Pages128
ID Numbers
Open LibraryOL17320795M
ISBN 100612277399

Algorithms for Scalable Synchronization on Shared-Memory Multiprocessors.J ohn M.:\IIellor-Cnumney' Michael L. Scott! April Techniques for efficiently coordinating parallel computation on MIMD, shared-memory multiproces­ cess a particular shared data structure at a tirne) and are a baRic building block for synchronization File Size: 3MB. Understanding how off-chip memory bandwidth partitioning in Chip Multiprocessors affects system performance Abstract: Chip Multi-Processor (CMP) architectures have recently become a mainstream computing platform. Recent CMPs allow cores to share expensive resources, such as the last level cache and off-chip pin bandwidth.

Lect. 4: Shared Memory Multiprocessors Obtained by connecting full processors together – Processors have their own connection to memory – Processors are capable of independent execution and control (Thus, by this definition, GPU is not a multiprocessor as the GPU cores are not. Compiler optimizations for scalable parallel systems: languages, compilation techniques, Optimal tiling for minimizing communication in distributed shared-memory multiprocessors. Anant Agarwal, David Data and computation alignment is an important part of compiling sequential programs to architectures with non-uniform memory access times.

Automatic Memory Partitioning and Scheduling for Throughput and Power Optimization Jason Cong, Wei Jiang, Bin Liu and Yi Zou Computer Science Department hardware acceleration are data-intensive or computation-intensive kernels in multimedia processing, which require high throughput. The book presents two approaches to automatic partitioning and scheduling so that the same parallel program can be made to execute efficiently on widely different multiprocessors. The first approach is based on a macro dataflow model in which the program is partitioned into tasks at compile time and the tasks are scheduled on processors at run.


Share this book
You might also like
Cognitive-behavioural therapy for ADHD in adolescents and adults

Cognitive-behavioural therapy for ADHD in adolescents and adults

Seminar on the democratic principle of equal representation - forty years of Council of Europe activity, Strasbourg, 6-7 November 1989

Seminar on the democratic principle of equal representation - forty years of Council of Europe activity, Strasbourg, 6-7 November 1989

Common bushes afire with God

Common bushes afire with God

Christianity in Celtic lands

Christianity in Celtic lands

Civil Code of the Hungarian Peoples Republic

Civil Code of the Hungarian Peoples Republic

Problems and progress in medical care

Problems and progress in medical care

Our broken dreams

Our broken dreams

Conservation information centres

Conservation information centres

Directory of international film and video festivals.

Directory of international film and video festivals.

Urban Development Corporation, award and administration of personal service contracts

Urban Development Corporation, award and administration of personal service contracts

Mental Health Act

Mental Health Act

Parliamentary stages of a Government bill.

Parliamentary stages of a Government bill.

Bible Insights: Esther

Bible Insights: Esther

Illegal alien felons

Illegal alien felons

American National Standard for Metric Practice (Ansi)

American National Standard for Metric Practice (Ansi)

Automatic computation and data partitioning on scalable shared memory multiprocessors by Sudarsan Tandri Download PDF EPUB FB2

Of data and computation partitions; shared memory effects such as cache affinity, false sharing, synchronization and contention must also be taken into account. Furthermore, the presense of shared memory hardware makes the use of the owner-computes rule unnecessary; the performance of some applications benefit from relaxing this rule.

“Computation and data partitioning on scalable shared memory multiprocessors,” Proc. of PDPTA, pp. 41–50, Automatic data and computation partitioning on scalable shared memory multiprocessors.

In: Sehr D., Banerjee U., Gelernter D., Nicolau A., Padua D. (eds) Languages and Compilers for Parallel Computing. Cited by: 5. This paper describes an algorithm for deriving data and computation partitions to improve memory locality on scalable shared memory multiprocessors.

The algorithm first determines computation partitions by establishing affinity between where the computations are performed and where the data is located based on data accesses in the program.

This paper describes an algorithm for deriving data and computation partitions to improve locality on scalable shared memory multiprocessors.

The algorithm first determines a computation partitioning by establishing affinity between where the computations are performed and where the data. BibTeX @MISC{Tandri96automaticdata, author = {Sudarsan Tandri and Tarek S.

Abdelrahman}, title = {Automatic Data and Computation Partitioning on Scalable Shared Memory Multiprocessors}, year =. ing data and computation partitions on scalable shared memory multiprocessors.

The algorithm establishes affin-ity relationships between where computations are per-formed and where data is located based on array accesses in the program. The algorithm then uses these affinity re-lationshipsto determine bothstaticand dynamic partitions.

The problem of deriving optimal partitions is NP-hard [2]. Additionally, the suitability of data and/or computation partitions depends not only on data access patterns, but also on the characteristics of the target multiprocessor.

A number of approaches have been proposed to tackle this problem. sor must exclusively reside in the processor’s local portion of the shared memory. Hardware automatically migrates data to the processor that requests the data in units of subpages.

Hence, the computation partitioning of a loop dictates the residence of a data item and hence the distribution of the arrays in the loop. This paper describes an algorithm for deriving data and computation partitions to improve memory locality on scalable shared memory multiprocessors.

The algorithm first determines computation partitions by establishing affinity between where the computations are performed and where the data is located based on data accesses in the : Sudarsan Tandri and Tarek S. Abdelrahman. to piace data carefully in the distributed shared memory.

In particular. relying on only the native operating system page placement policies to manage data often results in poor performance of applications. Computation and data partitioning is necessary for the management of corn pu tation and data on shared memory by: 2.

Furthermore, the presence of shared memory hardware makes the use of the owner-computes rule unnecessary; the performance of some applications benefit from relaxing this rule. We described an algorithm for deriving data and computation partitions on SSMMs taking shared memory effects.

This paper describes an algorithm for deriving data and computation partitions to improve locality on scalable shared memory multiprocessors. The algorithm first determines a computation partitioning by establishing affinity between where the computations are performed and where the data is located.

Automatic Data Partitioning on Distributed Memory Multiprocessors Conference Paper (PDF Available) March with 17 Reads How we measure 'reads'.

S. Tandri and T. Abdelrahman. Automatic partitioning of data and computation on scalable shared memory multiprocessors. In Proc. of the Int’l Conference on Parallel Processing, pages 64–73, Google ScholarCited by: 2. Automatic Data and Computation Partitioning on Scalable Shared Memory Multiprocessors.

By Sudarsan Tandri And, Sudarsan T and Tarek S. Abdelrahman. Abstract. this paper, we address the problem of automatically deriving data and computation partitions for scientific applications on SSMMs.

The problem of deriving optimal partitions is NP-hard [2]. Automatic Data and Computation Partitioning on Scalable Shared Memory Multiprocessors Sudarsan Tandri and Tarek Abdelrahman University of Toronto.

This paper describes an algorithm for deriving data and computation partitions to improve memory locality on scalable shared memory multiprocessors.

Automatic Computation and Data Partitioning By Sudarsan Tandri. Abstract. Scalable shared memory multiprocessors are becoming increasingly popular platforms for high-performance scientific computing because they both scale to large numbers of processors and support the familiar shared memory abstraction.

In order to improve application Author: Sudarsan Tandri. On shared-memory multiprocessors these reductions are typically parallelized by computing partial results into replicated buffers, then combining the values into shared data using synchronization.

Existing automatic parallelizers for distributed-memory machines require the user to explicitly specify how the data of the sequential program is mapped to the processors of the target machine. In this paper, we outline the features of a software tool to provide automatic Cited by:   Abstract.

In this paper we present a unified approach for compiling programs for Distributed-Memory Multiprocessors (DMM). Parallelization of sequential programs for DMM is much more difficult to achieve than for shared memory systems due to the exclusive local memory of each Virtual Processor (VP).Cited by: 1.

A shared-memory multiprocessor is a computer system composed of multiple independent processors that execute different instruction streams. Using Flynns’s classification [ 1], an SMP is a multiple-instruction multiple-data (MIMD) processors share a common memory address space and communicate with each other via memory.Automatic tuning of iterative computation on heterogeneous multiprocessors with ADITHE.

for shared memory systems. Data partitioning algorithms aiming to minimize the execution time and.In previous works [3, 4] we have shown that the DWA-LIP method performs well for typical irregular codes on ccNUMA share memory machines, specially for large data sets.

In all these tested cases.