Abstract

To investigate the effects of cache coherency on multiprocessors, it is helpful to first explore coherency issues within uniprocessors, working with a small part of a big problem instead of attacking the big problem from the start. This thesis will investigate the design and implementation of three different cache designs, varying the mapping strategy, replacement algorithm, and size parameters to determine the effects each have on the cache miss ratio, coherency, and average memory access time. VHDL is used to create software models of each cache design investigated, so that parameter values can be easily changed, and so that no money or time is wasted by first prototyping the cache design in actual hardware. These VHDL implementations are presented, along with several test-bench programs that were used to not only validate the performance of the VHDL implementation, but also to explore the program-dependent performance factors and coherency. Several snoopy cache coherence protocols are presented at the end of the thesis, in order to suggest future research into the VHDL implementation of shared-memory multiprocessors.

Library of Congress Subject Headings

Cache memory; Multiprocessors; VHDL (Computer hardware description language)

Publication Date

11-1-1998

Document Type

Thesis

Department, Program, or Center

Computer Engineering (KGCOE)

Advisor

Czernikowski, Roy

Advisor/Committee Member

Shaaban, Muhammad

Advisor/Committee Member

Hsu, Kenneth

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: TK7895.M4 B47 1998

Campus

RIT – Main Campus

Download

Share

COinS