Time Sharing Systems That Use a Partitioning Algorithm on Mesh-connected Parallel Computers

نویسندگان

  • J. Ding
  • A. Hori
  • H. Tezuka
  • P. G. Sobalvarro
  • S. Pakin
  • W. E. Weihl
چکیده

computers. Scheduling here is in principle moderate co-scheduling, with which the order of priority of the specic parallel processes is controlled per xed time. In addition, as a method to change the order of priority of parallel processes simultaneously for conducting co-scheduling, the authors proposed and installed internal synchronization which takes advantage of the synchronous clock that is a hardware property of AP1000+. Together with the communication library for polling/signal switching, this method allowed ecient execution of both ne grain parallel processes and coarse grain parallel processes. It was conrmed that co-scheduling allows ecient operation of busy wait communications of ne grain parallel processes, which wait for communication by busy wait, even when several of those are input. In addition , it was conrmed that the overall processing eciency of coarse grain parallel processes, which wait for communication by signal, could be improved by issuing context switching to execute another parallel process. Although the problem with internal synchronization is the co-scheduling skew caused by the failure of synchronization in the respective processors when conducting rescheduling , it was concluded that the time lag was negligible since it accounted for around 2.5% of 200 ms where priority switching takes place. The important issue of a parallel OS is to establish both a communication library and a scheduler in consideration of the properties of applications and the structures of the hardware available. A parallel OS designer must decide whether to use waiting by busy wait or context switching based on the communication patterns of the applications. If co-scheduling can eciently be installed on usable hardware, it is sucient to consider its introduction for ne grain parallel processes. At present, AP/Linux is compatible with both ne grain and coarse grain parallel processes, thus allowing ecient processing of a wide variety of application programs.

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 Fast Algorithm for Image Component Labeling with Local Operators on Mesh Connected Computers

A new parallel algorithm for image component labeling with local operators on SIMD mesh connected computers is presented. This algorithm provides a positive answer to the open question whether there exists an O(n)–time and O(log n)–space local labeling algorithm on SIMD mesh connected computers. The algorithm uses a pipeline mechanism with stack–like data structures to achieve the lower bound o...

متن کامل

Design and Implementation of an Adaptive Meta-Partitioner for SAMR Grid Hierarchies

In this paper we present a pilot implementation of the Meta-Partitioner, a partitioning framework that automatically selects, configures, and invokes suitable partitioning algorithms for Structured Adaptive Mesh Refinement (SAMR) applications. Efficient use of SAMR on parallel computers requires that the dynamic grid hierarchy is repeatedly repartitioned and redistributed. The partitioning proc...

متن کامل

Algorithms Minimizing Peak Energy on Mesh-Connected Systems

There are many situations in parallel computing in which reducing power consumption is an important goal. Here we consider mesh-connected systems where each processor is connected to its neighbors in a regular grid. Such systems have been studied and constructed at a wide range of scales, from cellular automata to sensor networks to supercomputer interconnection networks. We are especially inte...

متن کامل

Parallel Mesh Partitioning on Distributed Memory Systems

We discuss the problem of deriving parallel mesh partitioning algorithms for mapping unstructured meshes to parallel computers. In itself this raises a paradox – we seek to find a high quality partition of the mesh, but to compute it in parallel we require a partition of the mesh. In fact, we overcome this difficulty by deriving an optimisation strategy which can find a high quality partition e...

متن کامل

Heuristic Algorithms for Automatic Graph Partitioning

Practical implementations of the Finite Element method on distributed memory multi-computer systems necessitate the use of partitioning tools to subdivide the mesh into sub-meshes of roughly equal size. Graph partitioning algorithms are mandatory when implementing distributed sparse matrix methods or domain decomposition techniques for irregularly structured problems, on parallel computers. We ...

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 1999