Automatically assigned DDC number: 004

Manually assigned DDC number: 00435

Number of references: 0

Title: Using Genetic Algorithms to Fine-Tune Instruction-Scheduling Heuristics

Author:

Author:

Author:

Subject: Steven J. Beaty,Scott Colcord,Philip H. Sweany Using Genetic Algorithms to Fine-Tune Instruction-Scheduling Heuristics

Description: Instruction scheduling is an NP-complete problem that involves finding the fastest sequence of machine instructions from an abstract program representation. List scheduling is a method often used for instruction scheduling when producing code for instruction-level parallel processors and can produce excellent results when appropriate heuristics are chosen. We have investigated a method of experimentally determining good scheduling heuristics and found that it does indeed provide an easy way to tune instructionscheduling heuristics. 1 Introduction Computer manufacturers are continually striving to make faster computers with a combination of faster circuitry and increasing the amount of simultaneous computation (parallelism) in their architectures. One popular method of increasing the degree of simultaneous computation is instruction-level parallelism (ILP.) ILP computers exploit the implicit parallelism that most programs contain [25]. They overlap the execution of operations 1 that ...

Contributor: The Pennsylvania State University CiteSeer Archives

Publisher: unknown

Date: 1999-02-05

Pubyear: 1996

Format: ps

Identifier: http://citeseer.ist.psu.edu/140647.html

Source: http://emess.mscd.edu/~beaty/Dossier/Papers/MPCS96.ps

Language: en

Rights: unrestricted