{"id":13321,"date":"2014-11-18T14:48:15","date_gmt":"2014-11-18T13:48:15","guid":{"rendered":"http:\/\/www.planet3dnow.de\/cms\/?p=13321"},"modified":"2014-11-18T16:21:04","modified_gmt":"2014-11-18T15:21:04","slug":"amd-pathscale-openacc-standards-group","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/13321-amd-pathscale-openacc-standards-group\/","title":{"rendered":"AMD<\/span> und PathScale treten OpenACC Standards Group bei"},"content":{"rendered":"

Im Rah\u00admen der gera\u00adde in New Orleans statt\u00adfin\u00adden\u00adden Super\u00adcom\u00adpu\u00adting Con\u00adfe\u00adrence 2014 (SC14<\/span>) haben AMD<\/span> und der Com\u00adpi\u00adler-Spe\u00adzia\u00adlist PathSca\u00adle ihren Bei\u00adtritt zur Open\u00adACC Stan\u00addards Group<\/a><\/em> bekannt\u00adge\u00adge\u00adben. Open\u00adACC ist eine im HPC-Umfeld sehr belieb\u00adte API<\/span> zur Pro\u00adgram\u00admie\u00adrung von Clus\u00adtern und Super\u00adcom\u00adpu\u00adtern, die \u00e4hn\u00adlich wie OpenMP auf Direk\u00adti\u00adven basiert und urspr\u00fcng\u00adlich 2011 von CAPS<\/span>, Cray, NVIDIA<\/span> und PGI<\/span> spe\u00adzi\u00adell f\u00fcr den Ein\u00adsatz von Beschleu\u00adni\u00adger\u00adkar\u00adten erschaf\u00adfen<\/a> wur\u00adde. Seit\u00adher mau\u00adser\u00adte sich die Open\u00adACC-API auch durch den Bei\u00adtritt von immer mehr For\u00adschungs\u00adein\u00adrich\u00adtun\u00adgen und Uni\u00adver\u00adsi\u00adt\u00e4\u00adten zu einem wich\u00adti\u00adgen Stan\u00addard f\u00fcr die Pro\u00adgram\u00admie\u00adrung von H\u00f6chst\u00adleis\u00adtungs\u00adrech\u00adnern. Ent\u00adwick\u00adler k\u00f6n\u00adnen mit Hil\u00adfe der Direk\u00adti\u00adven par\u00adal\u00adle\u00adli\u00adsier\u00adba\u00adre Tei\u00adle des Quell\u00adcodes f\u00fcr den Com\u00adpi\u00adler mar\u00adkie\u00adren, die f\u00fcr die Aus\u00adla\u00adge\u00adrung auf die Beschleu\u00adni\u00adger geeig\u00adnet sind. Hier\u00adzu geh\u00f6\u00adren vor allem Schlei\u00adfen, bei deren Mar\u00adkie\u00adrung auch wei\u00adte\u00adre Details defi\u00adniert wer\u00adden k\u00f6n\u00adnen, damit die Aus\u00adf\u00fch\u00adrung auf spe\u00adzi\u00adfi\u00adschen Beschleu\u00adni\u00adgern wirk\u00adlich effek\u00adtiv erfolgt. Ent\u00adwick\u00adler k\u00f6n\u00adnen sich somit auf die Algo\u00adrith\u00admen und deren Imple\u00admen\u00adtie\u00adrung in die im HPC-Umfeld gebr\u00e4uch\u00adli\u00adchen Pro\u00adgram\u00admier\u00adspra\u00adchen For\u00adtran, C und C++ kon\u00adzen\u00adtrie\u00adren, w\u00e4h\u00adrend der Com\u00adpi\u00adler die eigent\u00adli\u00adche Par\u00adal\u00adle\u00adli\u00adsie\u00adrung ent\u00adspre\u00adchend der Ziel\u00adhard\u00adware vornimmt.<\/span><\/p>\n

AMD<\/span> hat\u00adte bereits 2012 die Bedeu\u00adtung von Open\u00adACC erkannt und gab Ende Okto\u00adber 2012 eine Zusam\u00admen\u00adar\u00adbeit mit The Port\u00adland Group<\/em> (PGI<\/span>)<\/a> bekannt, damit deren Open\u00adACC-Com\u00adpi\u00adler auch Code f\u00fcr AMDs APUs und dedi\u00adzier\u00adte Gra\u00adfik\u00adkar\u00adten erzeu\u00adgen kann. Aller\u00addings wur\u00adde die\u00adse Initia\u00adti\u00adve durch den Auf\u00adkauf von PGI<\/span> durch NVIDIA<\/span> kon\u00adter\u00adka\u00adriert. Den\u00adnoch ran\u00adgen sich die Unter\u00adneh\u00admen dazu durch, das begon\u00adne\u00adne Pro\u00adjekt zu einem erfolg\u00adrei\u00adchen Ende zu f\u00fch\u00adren. Im Febru\u00adar 2014 erschien<\/a> dann die ers\u00adte sta\u00adbi\u00adle Ver\u00adsi\u00adon des Open\u00adACC-Com\u00adpi\u00adlers von PGI<\/span>, die Unter\u00adst\u00fct\u00adzung f\u00fcr AMD-GPUs bie\u00adtet. Hier\u00adzu erzeugt der Open\u00adACC-Com\u00adpi\u00adler basie\u00adrend auf den Direk\u00adti\u00adven Open\u00adCL-Code, der dann \u00fcber AMDs Open\u00adCL-Lauf\u00adzeit\u00adum\u00adge\u00adbung aus\u00adge\u00adf\u00fchrt wird. Zwi\u00adschen\u00adzeit\u00adlich trieb AMD<\/span> in Koope\u00adra\u00adti\u00adon mit SUSE<\/span> und der Open-Source-Gemein\u00adde die Inte\u00adgra\u00adti\u00adon eines HSA-Back-Ends in die GNU<\/span> Com\u00adpi\u00adler Coll\u00adec\u00adtion (GCC<\/span>)<\/a> vor\u00adan, wel\u00adches mit Hil\u00adfe von OpenMP-Direk\u00adti\u00adven in der Ver\u00adsi\u00adon 4.0, die bereits seit vie\u00adlen Jah\u00adren zur Par\u00adal\u00adle\u00adli\u00adsie\u00adrung auf Mehr\u00adkern- und Mul\u00adti-Pro\u00adzes\u00adsor\u00adsys\u00adte\u00admen zum Ein\u00adsatz kom\u00admen, eben\u00adfalls die Aus\u00adla\u00adge\u00adrung von Berech\u00adnun\u00adgen von der CPU<\/span> auf Beschleu\u00adni\u00adger unter\u00adst\u00fct\u00adzen soll. AMD<\/span> und SUSE<\/span> ori\u00aden\u00adtie\u00adren sich hier\u00adbei an den von Open\u00adACC genutz\u00adten Direktiven.<\/p>\n

Greg Stoner, Sr. Direc\u00adtor of Deve\u00adlo\u00adper Tech\u00adno\u00adlo\u00adgy, Pro\u00adfes\u00adsio\u00adnal Gra\u00adphics bei AMD<\/span> betont in sei\u00adnem State\u00adment die Vor\u00adtei\u00adle von Open\u00adACC bei der Pro\u00adgram\u00admie\u00adrung von hete\u00adro\u00adge\u00adnen Rechen\u00adsys\u00adte\u00admen. Das ver\u00adein\u00adfach\u00adte, auf Direk\u00adti\u00adven basie\u00adren\u00adde Pro\u00adgram\u00admier\u00admo\u00addell mit nur einem ein\u00adzi\u00adgen Source-Code lege eine soli\u00adde Grund\u00adla\u00adge, um die Vor\u00adtei\u00adle hete\u00adro\u00adge\u00adner Rechen\u00adsys\u00adte\u00adme nutz\u00adbar zu machen. Zudem ist Stoner neu\u00ades Mit\u00adglied des Open\u00adACC Board of Direc\u00adtors gewor\u00adden und will in die\u00adser Funk\u00adti\u00adon bei der Wei\u00adter\u00adent\u00adwick\u00adlung des Stan\u00addards behilf\u00adlich sein.<\/p>\n

\u201c<\/span>The HPC<\/span> com\u00admu\u00adni\u00adty is dri\u00adving the need for rich hete\u00adro\u00adge\u00adneous com\u00adpu\u00adting solu\u00adti\u00adons to bet\u00adter meet cus\u00adto\u00admer per\u00adfor\u00admance and power needs. With this tran\u00adsi\u00adti\u00adon the\u00adre is a strong need for a rich, sim\u00adpli\u00adfied pro\u00adg\u00adra\u00adming model to sup\u00adport new hete\u00adro\u00adge\u00adneous appli\u00adca\u00adti\u00adon deve\u00adlo\u00adp\u00adment and pull in lega\u00adcy C\/C++\/Fortran appli\u00adca\u00adti\u00adons. Open\u00adACC 2.0 via its sin\u00adgle source direc\u00adti\u00adves-based pro\u00adgramming model puts in place a solid foun\u00adda\u00adti\u00adon to dri\u00adve key bene\u00adfits of a hete\u00adro\u00adge\u00adneous hard\u00adware plat\u00adform. AMD<\/span> is loo\u00adking for\u00adward to working with the Open\u00adACC orga\u00adniza\u00adti\u00adon to help it to con\u00adti\u00adnue to evol\u00adve this important pro\u00adgramming standard.\u201d<\/p><\/blockquote>\n

Oscar Her\u00adnan\u00addez, Prin\u00adci\u00adple Inves\u00adti\u00adga\u00adtor am Oak Ridge Natio\u00adnal Labo\u00adra\u00adto\u00adry (ORNL<\/span>), sieht durch den Bei\u00adtritt von PathSca\u00adle zur Open\u00adACC Stan\u00addards Group einen gro\u00ad\u00dfen Vor\u00adteil in der Diver\u00adsi\u00adfi\u00adzie\u00adrung der ver\u00adf\u00fcg\u00adba\u00adren Com\u00adpi\u00adler, wel\u00adche Unter\u00adst\u00fct\u00adzung f\u00fcr Open\u00adACC bie\u00adten. Zudem beab\u00adsich\u00adti\u00adge das ORNL<\/span> den PathSca\u00adle-Com\u00adpi\u00adler f\u00fcr die Nut\u00adzer des eige\u00adnen Cray-XK7-Sys\u00adtem (Titan) ver\u00adf\u00fcg\u00adbar zu machen.<\/p>\n

\u201c<\/span>It is important in any HPC<\/span> deve\u00adlo\u00adp\u00adment envi\u00adron\u00adment to have access to a varie\u00adty of com\u00adpi\u00adlers. Some\u00adti\u00admes one com\u00adpi\u00adler will exhi\u00adbit bet\u00adter per\u00adfor\u00admance on a par\u00adti\u00adcu\u00adlar code than ano\u00adther, or behave in a more desi\u00adra\u00adble way. We ful\u00adly intend to make PathSca\u00adle available to deve\u00adlo\u00adpers crea\u00adting appli\u00adca\u00adti\u00adons for ORNL<\/span>\u2019s Titan Cray XK7<\/span> supercomputer.\u201d<\/p><\/blockquote>\n

Quel\u00adle:<\/strong> Pres\u00adse\u00admit\u00adtei\u00adlung<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

Im Rah\u00admen der gera\u00adde in New Orleans statt\u00adfin\u00adden\u00adden Super\u00adcom\u00adpu\u00adting Con\u00adfe\u00adrence 2014 (SC14<\/span>) haben AMD<\/span> und der Com\u00adpi\u00adler-Spe\u00adzia\u00adlist PathSca\u00adle ihren Bei\u00adtritt zur Open\u00adACC Stan\u00addards Group bekannt\u00adge\u00adge\u00adben. Open\u00adACC ist eine im HPC-Umfeld sehr belieb\u00adte API<\/span> zur Pro\u00adgram\u00admie\u00adrung von Clus\u00adtern und Super\u00adcom\u00adpu\u00adtern, die \u00e4hn\u00adlich wie OpenMP auf Direk\u00adti\u00adven basiert und urspr\u00fcng\u00adlich 2011 von CAPS<\/span>, Cray, NVIDIA<\/span> und PGI<\/span> spe\u00adzi\u00adell f\u00fcr den Ein\u00adsatz von Beschleu\u00adni\u00adger\u00adkar\u00adten erschaf\u00adfen wur\u00adde. (\u2026) Wei\u00adter\u00adle\u00adsen \u00bb<\/a><\/p>\n","protected":false},"author":8,"featured_media":13322,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"wp_typography_post_enhancements_disabled":false,"ngg_post_thumbnail":0,"footnotes":""},"categories":[12,11],"tags":[966,539,540,538,749],"class_list":["post-13321","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-aktuelles","category-news","tag-amd","tag-compiler","tag-hpc","tag-openacc","tag-pathscale","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/13321","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/comments?post=13321"}],"version-history":[{"count":9,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/13321\/revisions"}],"predecessor-version":[{"id":13344,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/13321\/revisions\/13344"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media\/13322"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=13321"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=13321"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=13321"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}