ASC3: Algorithmic Strategies for Compiler Controlled Caches

Abstract

The ASC3 effort was a collaboration among research groups. The effort focused on strategies for improving the performance of data-intensive and memory-bound programs through the innovation of algorithmic approaches in the following areas: 1) The management of programmable memory hierarchies, 2) Data remapping and speculative execution for improved cache performance, and 3) Compiler optimizations that are tolerant to memory aliasing. The techniques developed by the ASC3 effort were applied to a number of data-intensive applications, including automatic target recognition, database management, image matching, neural network simulation and scientific computation. The strategies were validated using industry-strength simulation and emulation tools, based on the Trimaran EPIC Research Infrastructure. The optimizations developed in the program exploited trends in microprocessor design and are applicable to current platforms, emerging EPIC architectures, and future generations of COTS software. The ASC3 effort also extended the functionality of the Trimaran Compiler EPIC infrastructure.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 01, 2003
Accession Number
ADA418767

Entities

People

  • Benjamin M. Goldberg
  • Krishna Palem
  • Michael O. Rabin

Organizations

  • New York University

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Air Force
  • Air Force Research Laboratories
  • Algorithms
  • Classification
  • Compilers
  • Computer Programming
  • Computing System Architectures
  • Government Procurement
  • Governments
  • Hierarchies
  • Hot Spots
  • Instruction Set Architecture
  • Neural Networks
  • New York
  • Optimization
  • Simulations
  • Target Recognition

Fields of Study

  • Computer science

Readers

  • Critical Infrastructure Protection in CBRN and WMD Threats.
  • Database Systems and Applications
  • Game Theory.

Technology Areas

  • AI & ML