skip to content
Performance/accuracy trade-offs of floating-point arithmetic on NVidia GPUs : from a characterization to an auto-tuner Preview this item
ClosePreview this item
Checking...

Performance/accuracy trade-offs of floating-point arithmetic on NVidia GPUs : from a characterization to an auto-tuner

Author: Sruthikesh Surineni; Michela Becchi
Publisher: [Columbia, Missouri] : [University of Missouri--Columbia], December 2017.
Dissertation: M.S. University of Missouri--Columbia 2017. Thesis
Edition/Format:   Thesis/dissertation : Document : Thesis/dissertation : State or province government publication : eBook   Computer File : English
Summary:
Floating-point computations produce approximate results, possibly leading to inaccuracy and reproducibility problems. Existing work addresses two issues: first, the design of high precision floating-point representations, and second, the study of methods to support a trade-off between accuracy and performance of central processing unit (CPU) applications. However, a comprehensive study of trade-offs between accuracy  Read more...
Rating:

(not yet rated) 0 with reviews - Be the first.

Find a copy online

Links to this item

Find a copy in the library

&AllPage.SpinnerRetrieving; Finding libraries that hold this item...

Details

Genre/Form: Academic theses
Material Type: Document, Thesis/dissertation, Government publication, State or province government publication, Internet resource
Document Type: Internet Resource, Computer File
All Authors / Contributors: Sruthikesh Surineni; Michela Becchi
OCLC Number: 1099281561
Notes: Field of study: Electrical engineering.
Dr. Michela Becchi, Thesis Supervisor.
Description: 1 online resource (xii, 81 pages) : illustrations (chiefly color)
Responsibility: by Sruthikesh Surineni.

Abstract:

Floating-point computations produce approximate results, possibly leading to inaccuracy and reproducibility problems. Existing work addresses two issues: first, the design of high precision floating-point representations, and second, the study of methods to support a trade-off between accuracy and performance of central processing unit (CPU) applications. However, a comprehensive study of trade-offs between accuracy and performance on modern graphic processing units (GPUs) is missing. This thesis covers the use of different floating-point precisions (i.e., single and double floating-point precision) in the IEEE 754 standard, the GNU Multiple Precision Arithmetic Library (GMP), and composite floating-point precision on a GPU using a variety of synthetic and real-world benchmark applications. First, we analyze the support for a single and double precision floating-point arithmetic on the considered GPU architectures, and we characterize the latencies of all floating-point instructions on GPU. Second, a study is presented on the performance/accuracy tradeoffs related to the use of different arithmetic precisions on addition, multiplication, division, and natural exponential function. Third, an analysis is given on the combined use of different arithmetic operations on three benchmark applications characterized by different instruction mixes and arithmetic intensities. As a result of this analysis, a novel auto tuner was designed in order to select the arithmetic precision of a GPU program leading to a better performance and accuracy tradeoff depending on the arithmetic operations and math functions used in the program and the degree of multithreading of the code.

Reviews

User-contributed reviews
Retrieving GoodReads reviews...
Retrieving DOGObooks reviews...

Tags

Be the first.
Confirm this request

You may have already requested this item. Please select Ok if you would like to proceed with this request anyway.

Linked Data


Primary Entity

<http://www.worldcat.org/oclc/1099281561> # Performance/accuracy trade-offs of floating-point arithmetic on NVidia GPUs : from a characterization to an auto-tuner
    a schema:Book, schema:CreativeWork, bgn:Thesis, pto:Web_document, schema:MediaObject ;
    bgn:inSupportOf "" ;
    library:oclcnum "1099281561" ;
    library:placeOfPublication <http://id.loc.gov/vocabulary/countries/mou> ;
    rdfs:comment "Unknown 'gen' value: sgp" ;
    schema:author <http://experiment.worldcat.org/entity/work/data/9074321751#Person/surineni_sruthikesh> ; # Sruthikesh Surineni
    schema:contributor <http://experiment.worldcat.org/entity/work/data/9074321751#Person/becchi_michela> ; # Michela Becchi
    schema:datePublished "2017" ;
    schema:description "Floating-point computations produce approximate results, possibly leading to inaccuracy and reproducibility problems. Existing work addresses two issues: first, the design of high precision floating-point representations, and second, the study of methods to support a trade-off between accuracy and performance of central processing unit (CPU) applications. However, a comprehensive study of trade-offs between accuracy and performance on modern graphic processing units (GPUs) is missing. This thesis covers the use of different floating-point precisions (i.e., single and double floating-point precision) in the IEEE 754 standard, the GNU Multiple Precision Arithmetic Library (GMP), and composite floating-point precision on a GPU using a variety of synthetic and real-world benchmark applications. First, we analyze the support for a single and double precision floating-point arithmetic on the considered GPU architectures, and we characterize the latencies of all floating-point instructions on GPU. Second, a study is presented on the performance/accuracy tradeoffs related to the use of different arithmetic precisions on addition, multiplication, division, and natural exponential function. Third, an analysis is given on the combined use of different arithmetic operations on three benchmark applications characterized by different instruction mixes and arithmetic intensities. As a result of this analysis, a novel auto tuner was designed in order to select the arithmetic precision of a GPU program leading to a better performance and accuracy tradeoff depending on the arithmetic operations and math functions used in the program and the degree of multithreading of the code."@en ;
    schema:exampleOfWork <http://worldcat.org/entity/work/id/9074321751> ;
    schema:genre "Government publication"@en ;
    schema:genre "Academic theses"@en ;
    schema:inLanguage "en" ;
    schema:name "Performance/accuracy trade-offs of floating-point arithmetic on NVidia GPUs : from a characterization to an auto-tuner"@en ;
    schema:productID "1099281561" ;
    schema:url <https://hdl.handle.net/10355/66752> ;
    wdrs:describedby <http://www.worldcat.org/title/-/oclc/1099281561> ;
    .


Related Entities

<http://experiment.worldcat.org/entity/work/data/9074321751#Person/becchi_michela> # Michela Becchi
    a schema:Person ;
    schema:familyName "Becchi" ;
    schema:givenName "Michela" ;
    schema:name "Michela Becchi" ;
    .

<http://experiment.worldcat.org/entity/work/data/9074321751#Person/surineni_sruthikesh> # Sruthikesh Surineni
    a schema:Person ;
    schema:familyName "Surineni" ;
    schema:givenName "Sruthikesh" ;
    schema:name "Sruthikesh Surineni" ;
    .

<http://www.worldcat.org/title/-/oclc/1099281561>
    a genont:InformationResource, genont:ContentTypeGenericResource ;
    schema:about <http://www.worldcat.org/oclc/1099281561> ; # Performance/accuracy trade-offs of floating-point arithmetic on NVidia GPUs : from a characterization to an auto-tuner
    schema:dateModified "2019-04-26" ;
    void:inDataset <http://purl.oclc.org/dataset/WorldCat> ;
    .

<https://hdl.handle.net/10355/66752>
    rdfs:comment "Freely available online" ;
    .


Content-negotiable representations

Close Window

Please sign in to WorldCat 

Don't have an account? You can easily create a free account.