Exploiting historical data: pruning autotuning spaces and
estimating the number of tuning steps

D 2019

Exploiting historical data: pruning autotuning spaces and estimating the number of tuning steps

OĽHA, Jaroslav, Jana HOZZOVÁ, Jan FOUSEK and Jiří FILIPOVIČ

Basic information

Original name

Exploiting historical data: pruning autotuning spaces and estimating the number of tuning steps

Name in Czech

Využití historických dat: prořezávání prostorů v autotuningu a odhad počtu tunících kroků

Authors

OĽHA, Jaroslav (703 Slovakia, belonging to the institution), Jana HOZZOVÁ (703 Slovakia, belonging to the institution), Jan FOUSEK (203 Czech Republic, belonging to the institution) and Jiří FILIPOVIČ (203 Czech Republic, guarantor, belonging to the institution)

Edition

Cham, Switzerland, Lecture Notes in Computer Science, p. 295-307, 13 pp. 2019

Publisher

Springer, Cham

Other information

Language

English

Type of outcome

Stať ve sborníku

Field of Study

10201 Computer sciences, information science, bioinformatics

Country of publisher

Switzerland

Confidentiality degree

není předmětem státního či obchodního tajemství

Publication form

printed version "print"

Impact factor

Impact factor: 0.402 in 2005

RIV identification code

RIV/00216224:14610/19:00115100

Organization unit

Institute of Computer Science

ISBN

978-3-030-48339-5

ISSN

DOI

http://dx.doi.org/10.1007/978-3-030-48340-1_23

UT WoS

000850928600023

Keywords (in Czech)

autotuning; odhad ceny tuningu; prořezávání prostoru tuningu; analýza senzitivity

Keywords in English

Autotuning; prediction of tuning cost; tuning space pruning; sensitivity analysis

Abstract

V originále

Autotuning, the practice of automatic tuning of code to provide performance portability, has received increased attention in the research community, especially in high performance computing. Ensuring high performance on a variety of hardware usually means modifications to the code, often via different values of a selected set of parameters, such as tiling size, loop unrolling factor or data layout. However, the search space of all possible combinations of these parameters can be enormous. Traditional search methods often fail to find a well-performing set of parameter values quickly. We have found that certain properties of tuning spaces do not vary much when hardware is changed. In this paper, we demonstrate that it is possible to use historical data to reliably predict the number of tuning steps necessary to find a well-performing configuration, and to reduce the size of the tuning space. We evaluate our hypotheses on a number of GPU-accelerated benchmarks written in CUDA and OpenCL.

Links

EF16_013/0001802, research and development project

Name: CERIT Scientific Cloud

Citovat

OĽHA, Jaroslav, Jana HOZZOVÁ, Jan FOUSEK and Jiří FILIPOVIČ. Exploiting historical data: pruning autotuning spaces and estimating the number of tuning steps. In Ulrich Schwardmann, Christian Boehme, Dora B. Heras, Valeria Cardellini, Emmanuel Jeannot, Antonio Salis, Claudio Schifanella, Ravi Reddy Manumachu, Dieter Schwamborn, Laura Ricci, Oh Sangyoon, Thomas Gruber, Laura Antonelli, Stephen L. Scott. Lecture Notes in Computer Science. Cham, Switzerland: Springer, Cham, 2019, p. 295-307. ISBN 978-3-030-48339-5. Available from: https://dx.doi.org/10.1007/978-3-030-48340-1_23.

@inproceedings{1551680,
   author = {Oľha, Jaroslav and Hozzová, Jana and Fousek, Jan and Filipovič, Jiří},
   address = {Cham, Switzerland},
   booktitle = {Lecture Notes in Computer Science},
   doi = {http://dx.doi.org/10.1007/978-3-030-48340-1_23},
   editor = {Ulrich Schwardmann, Christian Boehme, Dora B. Heras, Valeria Cardellini, Emmanuel Jeannot, Antonio Salis, Claudio Schifanella, Ravi Reddy Manumachu, Dieter Schwamborn, Laura Ricci, Oh Sangyoon, Thomas Gruber, Laura Antonelli, Stephen L. Scott},
   keywords = {Autotuning; prediction of tuning cost; tuning space pruning; sensitivity analysis},
   howpublished = {tištěná verze "print"},
   language = {eng},
   location = {Cham, Switzerland},
   isbn = {978-3-030-48339-5},
   pages = {295-307},
   publisher = {Springer, Cham},
   title = {Exploiting historical data: pruning autotuning spaces and estimating the number of tuning steps},
   year = {2019}
}

TY  - JOUR
ID  - 1551680
AU  - Oľha, Jaroslav - Hozzová, Jana - Fousek, Jan - Filipovič, Jiří
PY  - 2019
TI  - Exploiting historical data: pruning autotuning spaces and estimating the number of tuning steps
PB  - Springer, Cham
CY  - Cham, Switzerland
SN  - 9783030483395
KW  - Autotuning
KW  - prediction of tuning cost
KW  - tuning space pruning
KW  - sensitivity analysis
N2  - Autotuning, the practice of automatic tuning of code to provide performance portability, has received increased attention in the research community, especially in high performance computing. Ensuring high performance on a variety of hardware usually means modifications to the code, often via different values of a selected set of parameters, such as tiling size, loop unrolling factor or data layout. However, the search space of all possible combinations of these parameters can be enormous. Traditional search methods often fail to find a well-performing set of parameter values quickly. We have found that certain properties of tuning spaces do not vary much when hardware is changed. In this paper, we demonstrate that it is possible to use historical data to reliably predict the number of tuning steps necessary to find a well-performing configuration, and to reduce the size of the tuning space. We evaluate our hypotheses on a number of GPU-accelerated benchmarks written in CUDA and OpenCL.
ER  -

OĽHA, Jaroslav, Jana HOZZOVÁ, Jan FOUSEK and Jiří FILIPOVIČ. Exploiting historical data: pruning autotuning spaces and estimating the number of tuning steps. In Ulrich Schwardmann, Christian Boehme, Dora B. Heras, Valeria Cardellini, Emmanuel Jeannot, Antonio Salis, Claudio Schifanella, Ravi Reddy Manumachu, Dieter Schwamborn, Laura Ricci, Oh Sangyoon, Thomas Gruber, Laura Antonelli, Stephen L. Scott. \textit{Lecture Notes in Computer Science}. Cham, Switzerland: Springer, Cham, 2019, p.~295-307. ISBN~978-3-030-48339-5. Available from: https://dx.doi.org/10.1007/978-3-030-48340-1\_{}23.

Detailed Information on Publication Record