<?xml version="1.0" encoding="utf-8"?>
<TEI xmlns="http://www.tei-c.org/ns/1.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:hal="http://hal.archives-ouvertes.fr/" xmlns:gml="http://www.opengis.net/gml/3.3/" xmlns:gmlce="http://www.opengis.net/gml/3.3/ce" version="1.1" xsi:schemaLocation="http://www.tei-c.org/ns/1.0 http://api.archives-ouvertes.fr/documents/aofr-sword.xsd">
  <teiHeader>
    <fileDesc>
      <titleStmt>
        <title>HAL TEI export of hal-00165020</title>
      </titleStmt>
      <publicationStmt>
        <distributor>CCSD</distributor>
        <availability status="restricted">
          <licence target="https://creativecommons.org/publicdomain/zero/1.0/">CC0 1.0 - Universal</licence>
        </availability>
        <date when="2026-05-18T16:58:30+02:00"/>
      </publicationStmt>
      <sourceDesc>
        <p part="N">HAL API Platform</p>
      </sourceDesc>
    </fileDesc>
  </teiHeader>
  <text>
    <body>
      <listBibl>
        <biblFull>
          <titleStmt>
            <title xml:lang="en">More Instruction Level Parallelism Explains the Actual Efficiency of Compensated Algorithms</title>
            <author role="aut">
              <persName>
                <forename type="first">Philippe</forename>
                <surname>Langlois</surname>
              </persName>
              <email type="md5">6053c70a9c4900f81e9af50f10d1844c</email>
              <email type="domain">univ-perp.fr</email>
              <idno type="idhal" notation="string">philippe-langlois</idno>
              <idno type="idhal" notation="numeric">3635</idno>
              <idno type="halauthorid" notation="string">2230-3635</idno>
              <idno type="IDREF">https://www.idref.fr/104061731</idno>
              <affiliation ref="#struct-3191"/>
            </author>
            <author role="aut">
              <persName>
                <forename type="first">Nicolas</forename>
                <surname>Louvet</surname>
              </persName>
              <idno type="halauthorid">10509-0</idno>
              <affiliation ref="#struct-3191"/>
            </author>
            <editor role="depositor">
              <persName>
                <forename>Philippe</forename>
                <surname>Langlois</surname>
              </persName>
              <email type="md5">6053c70a9c4900f81e9af50f10d1844c</email>
              <email type="domain">univ-perp.fr</email>
            </editor>
          </titleStmt>
          <editionStmt>
            <edition n="v1" type="current">
              <date type="whenSubmitted">2007-07-24 14:57:29</date>
              <date type="whenModified">2024-01-26 12:55:51</date>
              <date type="whenReleased">2007-07-24 15:44:37</date>
              <date type="whenProduced">2007-07-24</date>
              <date type="whenEndEmbargoed">2007-07-24</date>
              <ref type="file" target="https://hal.science/hal-00165020v1/document">
                <date notBefore="2007-07-24"/>
              </ref>
              <ref type="file" subtype="author" n="1" target="https://hal.science/hal-00165020v1/file/hal.pdf" id="file-165020-495506">
                <date notBefore="2007-07-24"/>
              </ref>
            </edition>
            <respStmt>
              <resp>contributor</resp>
              <name key="113752">
                <persName>
                  <forename>Philippe</forename>
                  <surname>Langlois</surname>
                </persName>
                <email type="md5">6053c70a9c4900f81e9af50f10d1844c</email>
                <email type="domain">univ-perp.fr</email>
              </name>
            </respStmt>
          </editionStmt>
          <publicationStmt>
            <distributor>CCSD</distributor>
            <idno type="halId">hal-00165020</idno>
            <idno type="halUri">https://hal.science/hal-00165020</idno>
            <idno type="halBibtex">langlois:hal-00165020</idno>
            <idno type="halRefHtml">2007</idno>
            <idno type="halRef">2007</idno>
            <availability status="restricted">
              <licence target="https://about.hal.science/hal-authorisation-v1/">HAL Authorization<ref corresp="#file-165020-495506"/></licence>
            </availability>
          </publicationStmt>
          <seriesStmt>
            <idno type="stamp" n="UNIV-PERP">Université Perpignan Via Domitia</idno>
          </seriesStmt>
          <notesStmt>
            <note type="commentary">11 pages</note>
            <note type="audience" n="1">Not set</note>
          </notesStmt>
          <sourceDesc>
            <biblStruct>
              <analytic>
                <title xml:lang="en">More Instruction Level Parallelism Explains the Actual Efficiency of Compensated Algorithms</title>
                <author role="aut">
                  <persName>
                    <forename type="first">Philippe</forename>
                    <surname>Langlois</surname>
                  </persName>
                  <email type="md5">6053c70a9c4900f81e9af50f10d1844c</email>
                  <email type="domain">univ-perp.fr</email>
                  <idno type="idhal" notation="string">philippe-langlois</idno>
                  <idno type="idhal" notation="numeric">3635</idno>
                  <idno type="halauthorid" notation="string">2230-3635</idno>
                  <idno type="IDREF">https://www.idref.fr/104061731</idno>
                  <affiliation ref="#struct-3191"/>
                </author>
                <author role="aut">
                  <persName>
                    <forename type="first">Nicolas</forename>
                    <surname>Louvet</surname>
                  </persName>
                  <idno type="halauthorid">10509-0</idno>
                  <affiliation ref="#struct-3191"/>
                </author>
              </analytic>
              <monogr>
                <imprint/>
              </monogr>
            </biblStruct>
          </sourceDesc>
          <profileDesc>
            <langUsage>
              <language ident="en">English</language>
            </langUsage>
            <textClass>
              <keywords scheme="author">
                <term xml:lang="en">performance evaluation</term>
                <term xml:lang="en">Accurate polynomial evaluation</term>
                <term xml:lang="en">Horner algorithm</term>
                <term xml:lang="en">compensated Horner algorithm</term>
                <term xml:lang="en">floating point arithmetic</term>
                <term xml:lang="en">IEEE-754 standard</term>
                <term xml:lang="en">instruction level parallelism</term>
                <term xml:lang="en">performance evaluation.</term>
              </keywords>
              <classCode scheme="classification">ACM G.4</classCode>
              <classCode scheme="halDomain" n="info.info-ms">Computer Science [cs]/Mathematical Software [cs.MS]</classCode>
              <classCode scheme="halTypology" n="UNDEFINED">Preprints, Working Papers, ...</classCode>
              <classCode scheme="halOldTypology" n="UNDEFINED">Preprints, Working Papers, ...</classCode>
              <classCode scheme="halTreeTypology" n="UNDEFINED">Preprints, Working Papers, ...</classCode>
            </textClass>
            <abstract xml:lang="en">
              <p>The compensated Horner algorithm and the Horner algorithm with double-double arithmetic improve the accuracy of polynomial evaluation in IEEE-754 floating point arithmetic. Both yield a polynomial evaluation as accurate as if it was computed with the classic Horner algorithm in twice the working precision. Both algorithms also share the same low-level computation of the floating point rounding errors and cost a similar number of floating point operations. We report numerical experiments to exhibit that the compensated algorithm runs at least twice as fast as the double-double one on modern processors. We propose to explain such efficiency by identifying more instruction level parallelism in the compensated implementation. Such property also applies to other compensated algorithms for summation, dot product and triangular linear system solving. More generally this paper illustrates how this kind of performance analysis may be useful to highlight the actual efficiency of numerical algorithms.</p>
            </abstract>
          </profileDesc>
        </biblFull>
      </listBibl>
    </body>
    <back>
      <listOrg type="structures">
        <org type="laboratory" xml:id="struct-3191" status="OLD">
          <orgName>Laboratoire de Physique Appliquée et d'Automatique</orgName>
          <orgName type="acronym">LP2A</orgName>
          <desc>
            <address>
              <addrLine>Universite de Perpignan 52 avenue Paul Alduy 66860 PERPIGNAN</addrLine>
              <country key="FR"/>
            </address>
          </desc>
          <listRelation>
            <relation name="EA3679" active="#struct-101475" type="direct"/>
          </listRelation>
        </org>
        <org type="institution" xml:id="struct-101475" status="VALID">
          <idno type="ROR">https://ror.org/03am2jy38</idno>
          <orgName>Université de Perpignan Via Domitia</orgName>
          <orgName type="acronym">UPVD</orgName>
          <desc>
            <address>
              <addrLine>52 avenue Paul Alduy - 66860 Perpignan Cedex 9</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.univ-perp.fr/</ref>
          </desc>
        </org>
      </listOrg>
    </back>
  </text>
</TEI>