Back to search

Cornell University

Logical Accelerators on Manycore Processors

Abstract

dc:description.abstract

Traditional general-purpose processors emphasize finding parallelism using hardware control logic, but this logic is energy inefficient and will limit future performance improvements. Manycore processors offer a contrasting approach for general-purpose computation: find parallelism in software and remove energy inefficient hardware control logic. Manycores offer up to thousands of programmable threads that can efficiently target applications with thread-level parallelism (TLP). However, manycores have not yet gained market adoption due to the difficulty of programming and lack of performance on applications containing data-level parallelism (DLP) and instruction-level parallelism (ILP). This thesis proposes to use software reconfiguration to improve manycore performance on various classes of applications with little hardware overhead. We propose two designs where software can group together existing hardware resources of adjacent cores to form logical accelerators. Rockcress adds lightweight hardware to enable the formation of a logical vector unit to accelerate applications with DLP. Watercress adds lightweight hardware to enable the formation of a logical CGRA to accelerate applications with ILP. These techniques will expand the set of applications that can enjoy the performance and computational density of manycore architectures.

Degree

thesis:*
Name thesis:degree_name
Ph. D., Electrical and Computer Engineering
Level thesis:degree_level
Doctor of Philosophy
Discipline thesis:degree_discipline
Electrical and Computer Engineering
Grantor
Cornell University
Year dc:date.issued
2023

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Bedoukian, Philip
Chair dc:contributor.chair
  • Sampson, Adrian
Committee members dc:contributor.committeemember
  • Batten, Christopher
  • Zhang, Zhiru

Rights

dc:rights
Statement dc:rights
  • Attribution 4.0 International
Language dc:language.iso
en

Identifiers

dc:identifier.*
Dc Identifier Bibid
16219344
Dc Identifier Other
ProQuest Submission ID: 13973
ProQuest Publication ID: 30631571

Chain of custody

source
Harvested from
Cornell University
Base URL
ecommons.cornell.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Bedoukian, Philip. Logical Accelerators on Manycore Processors. Doctor of Philosophy thesis, Cornell University, 2023. https://hdl.handle.net/1813/114575