The Quest for a Zero Overhead Shared Memory Parallel Machine

نویسندگان

  • Gautam Shah
  • Aman Singla
  • Umakishore Ramachandran
چکیده

In this paper we present a new approach to benchmark the performance of shared memory systems. This approach focuses on recognizing how far off is the performance of a given memory system from a realistic ideal parallel machine. We define such a realistic machine model called the z-machine, which accounts for the inherent communication costs in an application by tracking the data flow in the application. The z-machine is incorporated into an execution-driven simulation framework to be used as a reference for benchmarking for different memory systems. The components of the overheads in these memory systems are identified and quantified for four applications. Using the z-machine performance as the standard to strive for we discuss the implications of the performance results and suggest architectural trends to pursue for realizing a zero overhead shared memory machine.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

A Brief Introduction to UPC

UPC is a parallel extension of the C programming language which provides programmers with a shared global address space. A descendant of Split-C [1], AC [2], and PCP [3], UPC has two primary objectives: 1) to provide efficient access to the underlying machine, and 2) to establish a common syntax and semantics for explicitly parallel programming in C. The quest for high performance means in part...

متن کامل

Memory Utilization in Software DSM for Embedded Systems

Software Distributed Shared Memory (S-DSM) systems support parallel programming by implementing a shared memory on top of distributed system. It frees programmer from communication complexities to concentrate to parallel algorithms. However, there is a drawback: current S-DSM systems waist memory on all processors. Therefore it is hard to be implemented to embedded systems, which have small res...

متن کامل

Cost Models for Partitioning Parallel Computationsin Two Tiered ArchitecturesCsaba

The requirements to reduce the price/performance ratio and to provide scalability force multicomputer vendors to market a new type of multicomputer system , built with commodity components. The processors are grouped into hypernodes (or subsystems), which are connected through scalable high-speed interconnects. The system concept is somewhat hybrid: within a hypernode the machine is eeectively ...

متن کامل

Compilation Techniques for Fair Execution of Shared Memory Parallel Programs over a Network of Workstations

Compiler technologies are crucial for the eecient execution of sequential programs. This is not yet true for parallel programs, where the operating system performs most of the work, resulting in increased overhead for scheduling and distributed shared memory simulation. In this work we suggest simple compilation techniques that can be used to guarantee eecient execution of shared memory paralle...

متن کامل

Parallel algorithms for geometric shortest path problems

The original goal of this project was to investigate and compare the experimental performance and ease of programming of algorithms for geometric shortest path finding using shared memory and message passing programming styles on a shared memory machine. However, due to the extended unavailability of a suitable shared memory machine, this goal was only partially met, though a system suitable fo...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 1995