Vector seeker: A tool for finding vector potential

Graham Carl Evans, Seth Abraham, Bob Kuhn, David A Padua

Research output: Contribution to conferencePaper

Abstract

The importance of vector instructions is growing in modern computers. Almost all architectures include some form of vector instructions and the tendency is for the size of the instructions to grow with newer designs. To take advantage of the performance that these systems offer, it is imperative that programs use these instructions, and yet they do not always do so. The tools to take advantage of these extensions require programmer assistance either by hand coding or providing hints to the compiler. We present Vector Seeker, a tool to help investigate vector parallelism in existing codes. Vector Seeker runs with the execution of a program to optimistically measure the vector parallelism that is present. Besides describing Vector Seeker, the paper also evaluates its effectiveness using two applications from Petascale Application Collaboration Teams (PACT) and eight applications from Media Bench II. These results are compared to known results from manual vectorization studies. Finally, we use the tool to automatically analyze codes from Numerical Recipes and TSVC and then compare the results with the automatic vectorization algorithms of Intel's ICC.

Original languageEnglish (US)
Pages41-48
Number of pages8
DOIs
StatePublished - Jan 1 2014
Event2014 1st ACM SIGPLAN Workshop on Programming Models for SIMD/Vector Processing, WPMVP 2014 - Co-located with PPoPP 2014 - Orlando, FL, United States
Duration: Feb 16 2014Feb 16 2014

Other

Other2014 1st ACM SIGPLAN Workshop on Programming Models for SIMD/Vector Processing, WPMVP 2014 - Co-located with PPoPP 2014
CountryUnited States
CityOrlando, FL
Period2/16/142/16/14

Fingerprint

Vector Potential
Vectorization
Parallelism
Compiler
Coding
Evaluate

Keywords

  • Compilers
  • Dynamic analysis
  • SIMD
  • Vectorization

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Software
  • Modeling and Simulation

Cite this

Evans, G. C., Abraham, S., Kuhn, B., & Padua, D. A. (2014). Vector seeker: A tool for finding vector potential. 41-48. Paper presented at 2014 1st ACM SIGPLAN Workshop on Programming Models for SIMD/Vector Processing, WPMVP 2014 - Co-located with PPoPP 2014, Orlando, FL, United States. https://doi.org/10.1145/2568058.2568069

Vector seeker : A tool for finding vector potential. / Evans, Graham Carl; Abraham, Seth; Kuhn, Bob; Padua, David A.

2014. 41-48 Paper presented at 2014 1st ACM SIGPLAN Workshop on Programming Models for SIMD/Vector Processing, WPMVP 2014 - Co-located with PPoPP 2014, Orlando, FL, United States.

Research output: Contribution to conferencePaper

Evans, GC, Abraham, S, Kuhn, B & Padua, DA 2014, 'Vector seeker: A tool for finding vector potential', Paper presented at 2014 1st ACM SIGPLAN Workshop on Programming Models for SIMD/Vector Processing, WPMVP 2014 - Co-located with PPoPP 2014, Orlando, FL, United States, 2/16/14 - 2/16/14 pp. 41-48. https://doi.org/10.1145/2568058.2568069
Evans GC, Abraham S, Kuhn B, Padua DA. Vector seeker: A tool for finding vector potential. 2014. Paper presented at 2014 1st ACM SIGPLAN Workshop on Programming Models for SIMD/Vector Processing, WPMVP 2014 - Co-located with PPoPP 2014, Orlando, FL, United States. https://doi.org/10.1145/2568058.2568069
Evans, Graham Carl ; Abraham, Seth ; Kuhn, Bob ; Padua, David A. / Vector seeker : A tool for finding vector potential. Paper presented at 2014 1st ACM SIGPLAN Workshop on Programming Models for SIMD/Vector Processing, WPMVP 2014 - Co-located with PPoPP 2014, Orlando, FL, United States.8 p.
@conference{900c765c8d8740bca00b6330a5e1f654,
title = "Vector seeker: A tool for finding vector potential",
abstract = "The importance of vector instructions is growing in modern computers. Almost all architectures include some form of vector instructions and the tendency is for the size of the instructions to grow with newer designs. To take advantage of the performance that these systems offer, it is imperative that programs use these instructions, and yet they do not always do so. The tools to take advantage of these extensions require programmer assistance either by hand coding or providing hints to the compiler. We present Vector Seeker, a tool to help investigate vector parallelism in existing codes. Vector Seeker runs with the execution of a program to optimistically measure the vector parallelism that is present. Besides describing Vector Seeker, the paper also evaluates its effectiveness using two applications from Petascale Application Collaboration Teams (PACT) and eight applications from Media Bench II. These results are compared to known results from manual vectorization studies. Finally, we use the tool to automatically analyze codes from Numerical Recipes and TSVC and then compare the results with the automatic vectorization algorithms of Intel's ICC.",
keywords = "Compilers, Dynamic analysis, SIMD, Vectorization",
author = "Evans, {Graham Carl} and Seth Abraham and Bob Kuhn and Padua, {David A}",
year = "2014",
month = "1",
day = "1",
doi = "10.1145/2568058.2568069",
language = "English (US)",
pages = "41--48",
note = "2014 1st ACM SIGPLAN Workshop on Programming Models for SIMD/Vector Processing, WPMVP 2014 - Co-located with PPoPP 2014 ; Conference date: 16-02-2014 Through 16-02-2014",

}

TY - CONF

T1 - Vector seeker

T2 - A tool for finding vector potential

AU - Evans, Graham Carl

AU - Abraham, Seth

AU - Kuhn, Bob

AU - Padua, David A

PY - 2014/1/1

Y1 - 2014/1/1

N2 - The importance of vector instructions is growing in modern computers. Almost all architectures include some form of vector instructions and the tendency is for the size of the instructions to grow with newer designs. To take advantage of the performance that these systems offer, it is imperative that programs use these instructions, and yet they do not always do so. The tools to take advantage of these extensions require programmer assistance either by hand coding or providing hints to the compiler. We present Vector Seeker, a tool to help investigate vector parallelism in existing codes. Vector Seeker runs with the execution of a program to optimistically measure the vector parallelism that is present. Besides describing Vector Seeker, the paper also evaluates its effectiveness using two applications from Petascale Application Collaboration Teams (PACT) and eight applications from Media Bench II. These results are compared to known results from manual vectorization studies. Finally, we use the tool to automatically analyze codes from Numerical Recipes and TSVC and then compare the results with the automatic vectorization algorithms of Intel's ICC.

AB - The importance of vector instructions is growing in modern computers. Almost all architectures include some form of vector instructions and the tendency is for the size of the instructions to grow with newer designs. To take advantage of the performance that these systems offer, it is imperative that programs use these instructions, and yet they do not always do so. The tools to take advantage of these extensions require programmer assistance either by hand coding or providing hints to the compiler. We present Vector Seeker, a tool to help investigate vector parallelism in existing codes. Vector Seeker runs with the execution of a program to optimistically measure the vector parallelism that is present. Besides describing Vector Seeker, the paper also evaluates its effectiveness using two applications from Petascale Application Collaboration Teams (PACT) and eight applications from Media Bench II. These results are compared to known results from manual vectorization studies. Finally, we use the tool to automatically analyze codes from Numerical Recipes and TSVC and then compare the results with the automatic vectorization algorithms of Intel's ICC.

KW - Compilers

KW - Dynamic analysis

KW - SIMD

KW - Vectorization

UR - http://www.scopus.com/inward/record.url?scp=84899764947&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84899764947&partnerID=8YFLogxK

U2 - 10.1145/2568058.2568069

DO - 10.1145/2568058.2568069

M3 - Paper

AN - SCOPUS:84899764947

SP - 41

EP - 48

ER -