A Memory Design for the Message-Driven Processor

Abstract

The Message-Driven Processor (MDP) is a low-latency processing node for a scalable fine-grain MIMD concurrent computer, the Jellybean Machine. Programs are executed by passing messages through a low-latency network. Each MDP integrates a processor, a memory, and a communication network. On top of this message-passing model, the MDP supports a global virtual address space. This thesis involves the design and implementation of a memory for the Message- Driven Processor. The memory array can be accessed by index, by row, or as a set-associative cache. Index operations are used to read and write memory. Row operations reduce the latency in message-handling by providing special purpose buffers, Row Buffers that access four words (a row) of memory simultaneously. Two Queue Row Buffers enable buffering message at two different priority levels as soon as they arrive from the network. An Instruction Row Buffer acts as a small instruction cache. Set-associative operations provide a translation mechanism to enable translating any object to its associated item. MDP operating system routines use this cache to translate virtual identifiers into global addresses. The microarchitecture and the circuit design of the memory is developed. A test chip is fabricated to verify the design. Evaluation of the row operations is presented.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 01, 1988
Accession Number
ADA200781

Entities

People

  • Soha M. Hassoun

Organizations

  • Massachusetts Institute of Technology

Tags

Communities of Interest

  • Advanced Electronics

DTIC Thesaurus Topics

  • Alpha Particles
  • Amplifiers
  • Cell Size
  • Computer Architecture
  • Computer Science
  • Computers
  • Computing System Architectures
  • Decoding
  • Diagrams
  • Electrical Engineering
  • Engineering
  • Generators
  • Microarchitecture
  • Semiconductors
  • Simulations
  • Test And Evaluation
  • Waveforms

Fields of Study

  • Computer science

Readers

  • Database Systems and Applications
  • Parallel and Distributed Computing.

Technology Areas

  • Space