Other Scholarly Content
 

A message-passing system for the large-grain data-flow machine

Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/defaults/p5548063r

Descriptions

Attribute NameValues
Creator
Abstract
  • The Large-Grain Data-Flow (LGDF) Machine is a tightly-coupled multi-processor consisting of several independent nodes, upon each of which may run one or more tasks. These tasks must be able to communicate and synchronize with each other. This document describes a message-passing system developed to meet these requirements. The message-passing system allows any task on an arbitrary node to send information to any other task on an arbitrary, possibly the same, node. The system is based on the use of shared memory. Messages may be of any length, subject only to the constraints of memory size. Except for a header of control information, the system places no interpretation upon the contents of the message. Using performance data automatically provided, the user can easily tune the system to enhance performance. The first section of this paper presents a brief introduction to the concept of large-grain data-now. This is followed by a description of the message-passing system. All components of the system from the hardware to the high level system calls are discussed. The third section is a user's manual. This is a guide to the low level system calls, to system tunable parameters, and to deadlock conditions. The paper is concluded with a discussion of possible problems with and extensions to the system.
Resource Type
Date Issued
Academic Affiliation
Rights Statement
Publisher
Peer Reviewed
Language

Relationships

Parents:

This work has no parents.

Items

Thumbnail Title Date Uploaded Visibility Actions
file details Thieret_Lewis_1086_04_28.pdf 2023-02-14 Public Download