Back to search

Cornell University

EXTENDING DIOSPYROS AUTO-VECTORIZATION TO LLVM

Abstract

dc:description.abstract

Short vector instruction sets are ubiquitous on computers. Auto-vectorizingcompilers help programmers bridge the gap by automatically mapping domain specific codes to efficient sequences of vector instructions on domain specific architectures. This thesis builds on the Diospyros project to implement a Diospyros LLVM pass translating linear algebraic kernels to SIMD vector architectures. The pass uses the idea from the Diospyros project to apply a search-based equality saturation method to find vector instructions. Results show that the Diospyros LLVM pass generates vector code with good runtime performance on stencil and convolution kernels, but poor results on other kernels indicate future work is required.

Degree

thesis:*
Name thesis:degree_name
M.S., Computer Science
Level thesis:degree_level
Master of Science
Discipline thesis:degree_discipline
Computer Science
Grantor
Cornell University
Year dc:date.issued
2023

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Tran, Jonathan
Chair dc:contributor.chair
  • Sampson, Adrian
Committee member dc:contributor.committeemember
  • Zhang, Zhiru

Subjects

dc:subject × 4

Rights

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

Identifiers

dc:identifier.*
Dc Identifier Bibid
16454641
Dc Identifier Other
ProQuest Submission ID: 11979
ProQuest Publication ID: 30640414

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

Tran, Jonathan. EXTENDING DIOSPYROS AUTO-VECTORIZATION TO LLVM. Master of Science thesis, Cornell University, 2023. https://hdl.handle.net/1813/115639