{"id":20750,"date":"2015-10-29T17:06:03","date_gmt":"2015-10-29T16:06:03","guid":{"rendered":"http:\/\/www.planet3dnow.de\/cms\/?p=20750"},"modified":"2015-10-29T17:06:03","modified_gmt":"2015-10-29T16:06:03","slug":"pgi-accelerator-compilers-add-openacc-support-for-x86-multicore-cpus","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/20750-pgi-accelerator-compilers-add-openacc-support-for-x86-multicore-cpus\/","title":{"rendered":"<span class=\"caps\">PGI<\/span> Accelerator Compilers Add OpenACC Support for x86 Multicore CPUs"},"content":{"rendered":"<p><strong><span class=\"caps\">SANTA<\/span> <span class=\"caps\">CLARA<\/span>, Calif.\u2014Oct. 29, 2015\u2014<\/strong><span class=\"caps\">NVIDIA<\/span> today announ\u00adced avai\u00adla\u00adbi\u00adli\u00adty of ver\u00adsi\u00adon 15.10 of the <span class=\"caps\">PGI<\/span> Acce\u00adle\u00adra\u00adtor\u2122 For\u00adtran, C and C++ com\u00adpi\u00adlers, adding sup\u00adport for the Open\u00adACC\u00ae direc\u00adti\u00adves-based par\u00adal\u00adlel pro\u00adgramming stan\u00addard on x86 archi\u00adtec\u00adtu\u00adre mul\u00adti\u00adco\u00adre microprocessors.<\/p>\n<p>The new <span class=\"caps\">PGI<\/span> com\u00adpi\u00adlers deli\u00adver per\u00adfor\u00admance por\u00adta\u00adbi\u00adli\u00adty, allo\u00adwing Open\u00adACC-enab\u00adled source code to be com\u00adpi\u00adled for par\u00adal\u00adlel exe\u00adcu\u00adti\u00adon on a mul\u00adti\u00adco\u00adre <span class=\"caps\">CPU<\/span> or a <span class=\"caps\">GPU<\/span> acce\u00adle\u00adra\u00adtor. This capa\u00adbi\u00adli\u00adty pro\u00advi\u00addes tre\u00admen\u00addous fle\u00adxi\u00adbi\u00adli\u00adty for pro\u00adgramm\u00aders, enab\u00adling them to deve\u00adlop appli\u00adca\u00adti\u00adons that can take advan\u00adta\u00adge of mul\u00adti\u00adple sys\u00adtem archi\u00adtec\u00adtures with a sin\u00adgle ver\u00adsi\u00adon of their source code.<\/p>\n<p><span class=\"dquo\">\u201c<\/span>Our goal is to enable <span class=\"caps\">HPC<\/span> deve\u00adlo\u00adpers to easi\u00adly port appli\u00adca\u00adti\u00adons across all major <span class=\"caps\">CPU<\/span> and acce\u00adle\u00adra\u00adtor plat\u00adforms with uni\u00adform\u00adly high per\u00adfor\u00admance using a com\u00admon source code base,\u201d said Dou\u00adglas Miles, direc\u00adtor of <span class=\"caps\">PGI<\/span> Com\u00adpi\u00adlers <span class=\"amp\">&amp;<\/span> Tools at <span class=\"caps\">NVIDIA<\/span>. \u201cThis capa\u00adbi\u00adli\u00adty will be par\u00adti\u00adcu\u00adlar\u00adly important in the race towards exas\u00adca\u00adle com\u00adpu\u00adting in which the\u00adre will be a varie\u00adty of sys\u00adtem archi\u00adtec\u00adtures requi\u00adring a more fle\u00adxi\u00adble appli\u00adca\u00adti\u00adon pro\u00adgramming approach.\u201d<\/p>\n<p>This new <span class=\"caps\">PGI<\/span> fea\u00adture com\u00adpi\u00adles Open\u00adACC com\u00adpu\u00adte regi\u00adons for par\u00adal\u00adlel exe\u00adcu\u00adti\u00adon across all of the cores in an x86 pro\u00adces\u00adsor or mul\u00adti-socket ser\u00adver. The cores are trea\u00adted in aggre\u00adga\u00adte as a shared-memo\u00adry acce\u00adle\u00adra\u00adtor, eli\u00admi\u00adna\u00adting all data move\u00adment over\u00adhead in the resul\u00adting Open\u00adACC pro\u00adgrams. By default the com\u00adpi\u00adler gene\u00adra\u00adtes code that uses all the available cores in the sys\u00adtem, and seve\u00adral methods exist for pro\u00adgramm\u00aders to con\u00adtrol and fine-tune this behavior.<\/p>\n<p>Key bene\u00adfits of run\u00adning Open\u00adACC on mul\u00adti\u00adco\u00adre CPUs include:<\/p>\n<ul>\n<li>Effec\u00adti\u00adve uti\u00adliza\u00adti\u00adon of all cores of a mul\u00adti\u00adco\u00adre <span class=\"caps\">CPU<\/span> or mul\u00adti-socket ser\u00adver for par\u00adal\u00adlel execution<\/li>\n<li>Com\u00admon pro\u00adgramming model across CPUs and GPUs in For\u00adtran, C and&nbsp;C++<\/li>\n<li>Rapid explo\u00adita\u00adti\u00adon of exis\u00adting mul\u00adti\u00adco\u00adre par\u00adal\u00adle\u00adlism in a pro\u00adgram using the <span class=\"caps\">KERNELS<\/span> direc\u00adti\u00adve, which enables incre\u00admen\u00adtal opti\u00admiza\u00adti\u00adon for par\u00adal\u00adlel execution<\/li>\n<li>Sca\u00adlable per\u00adfor\u00admance across mul\u00adti\u00adco\u00adre CPUs and&nbsp;GPUs<\/li>\n<\/ul>\n<p><span class=\"dquo\">\u201c<\/span>Port\u00ading <span class=\"caps\">HPC<\/span> appli\u00adca\u00adti\u00adons from one plat\u00adform to ano\u00adther is one of the most signi\u00adfi\u00adcant cos\u00adts in the adop\u00adti\u00adon of breakth\u00adrough hard\u00adware tech\u00adno\u00adlo\u00adgies,\u201d said Bud\u00addy Bland, pro\u00adject direc\u00adtor at Oak Ridge Natio\u00adnal Labo\u00adra\u00adto\u00adry. \u201cOpen\u00adACC for mul\u00adti\u00adco\u00adre x86 CPUs pro\u00advi\u00addes con\u00adti\u00adnui\u00adty and code por\u00adta\u00adbi\u00adli\u00adty from exis\u00adting CPU-only and GPU-enab\u00adled appli\u00adca\u00adti\u00adons from machi\u00adnes like Titan to all of <span class=\"caps\">DOE<\/span>\u2019s upco\u00adming major sys\u00adtems as well as por\u00adta\u00adbi\u00adli\u00adty among tho\u00adse systems.\u201d<\/p>\n<p><strong>Gro\u00adwing Momen\u00adtum for OpenACC<\/strong><\/p>\n<p>The\u00adre are more than 10,000 deve\u00adlo\u00adpers using Open\u00adACC today, and seve\u00adral recent deve\u00adlo\u00adp\u00adments unders\u00adcore the con\u00adti\u00adnu\u00adal\u00adly gro\u00adwing adop\u00adti\u00adon of Open\u00adACC in high per\u00adfor\u00admance com\u00adpu\u00adting. At recent hacka\u00adthons con\u00adduc\u00adted world\u00adwi\u00adde, experts across a varie\u00adty of sci\u00aden\u00adti\u00adfic domains have been acce\u00adle\u00adra\u00adting their sci\u00aden\u00adti\u00adfic appli\u00adca\u00adti\u00adons with acce\u00adle\u00adra\u00adtors and Open\u00adACC. The\u00adse include appli\u00adca\u00adti\u00adons in such diver\u00adse fields as <span class=\"caps\">MRI<\/span> image recon\u00ads\u00adtruc\u00adtion (PowerG\u00adrid), com\u00adpu\u00adta\u00adtio\u00adnal flu\u00adid dyna\u00admics (<span class=\"caps\">INCOMP3D<\/span>, HiP\u00adSTAR and Nume\u00adca), cos\u00admo\u00adlo\u00adgy and astro\u00adphy\u00adsics (<span class=\"caps\">RAMSES<\/span>, <span class=\"caps\">CASTRO<\/span> and <span class=\"caps\">MAESTRO<\/span>), quan\u00adtum che\u00admis\u00adtry (<span class=\"caps\">LSDALTON<\/span>), com\u00adpu\u00adta\u00adtio\u00adnal phy\u00adsics (Nek\u00adCEM) and&nbsp;more.<\/p>\n<p>In addi\u00adti\u00adon, Gaus\u00adsi\u00adan, Inc. has announ\u00adced that it is using Open\u00adACC to port the <span class=\"caps\">GAUSSIAN<\/span> com\u00adpu\u00adta\u00adtio\u00adnal che\u00admis\u00adtry appli\u00adca\u00adti\u00adon to acce\u00adle\u00adra\u00adtors. At the recent iCAS2 con\u00adfe\u00adrence on cli\u00adma\u00adte and wea\u00adther in Anne\u00adcy, France, Meteo\u00adsu\u00adis\u00adse, the Swiss Fede\u00adral Office of Meteo\u00adro\u00adlo\u00adgy and Cli\u00adma\u00adto\u00adlo\u00adgy, announ\u00adced the deploy\u00adment of a <a href=\"http:\/\/blogs.nvidia.com\/blog\/2015\/09\/15\/gpus-weather\/\">GPU-acce\u00adle\u00adra\u00adted ver\u00adsi\u00adon of <span class=\"caps\">COSMO<\/span><\/a>, the world\u2019s first pro\u00adduc\u00adtion wea\u00adther fore\u00adcas\u00adting appli\u00adca\u00adti\u00adon run\u00adning on <span class=\"caps\">GPU<\/span> accelerators.<\/p>\n<p>In a recent poll of 150 Open\u00adACC deve\u00adlo\u00adpers, 94 per\u00adcent of the respond\u00adents repor\u00adted get\u00adting a spee\u00addup when run\u00adning on an acce\u00adle\u00adra\u00adtor, and over 90 per\u00adcent of the users would recom\u00admend OpenACC.<\/p>\n<p>More infor\u00adma\u00adti\u00adon on the <span class=\"caps\">PGI<\/span> Acce\u00adle\u00adra\u00adtor com\u00adpi\u00adlers with Open\u00adACC sup\u00adport is available at <a href=\"http:\/\/www.pgroup.com\/accel\">www.pgroup.com\/accel<\/a>. More infor\u00adma\u00adti\u00adon on the Open\u00adACC <span class=\"caps\">API<\/span> and stan\u00addard is at <a href=\"http:\/\/www.openacc.org\">www.openacc.org<\/a>.<\/p>\n<p>In addi\u00adti\u00adon to sup\u00adport for Open\u00adACC on mul\u00adti\u00adco\u00adre CPUs, <span class=\"caps\">PGI<\/span> ver\u00adsi\u00adon 15.10 includes a pre-pro\u00adduc\u00adtion release of the <span class=\"caps\">PGI<\/span> For\u00adtran, C and C++ com\u00adpi\u00adlers for Open\u00adPOWER CPUs with sup\u00adport for Open\u00adACC on <span class=\"caps\">NVIDIA<\/span> GPUs.<\/p>\n<p><strong>Avai\u00adla\u00adbi\u00adli\u00adty and Free&nbsp;Trial<\/strong><\/p>\n<p><span class=\"caps\">PGI<\/span> 15.10 with sup\u00adport for Open\u00adACC on mul\u00adti\u00adco\u00adre CPUs is expec\u00adted to be available this month direct\u00adly from <span class=\"caps\">PGI<\/span> and aut\u00adho\u00adri\u00adzed resel\u00adlers. New users can regis\u00adter for a free 90-day tri\u00adal as part of the <span class=\"caps\">NVIDIA<\/span> Open\u00adACC Tool\u00adkit. Uni\u00adver\u00adsi\u00adty stu\u00addents and facul\u00adty can app\u00adly for a free <span class=\"caps\">PGI<\/span> license.<\/p>\n<p><strong>About <span class=\"caps\">PGI<\/span> Software<\/strong><\/p>\n<p>An <span class=\"caps\">NVIDIA<\/span> Cor\u00adpo\u00adra\u00adti\u00adon brand, <span class=\"caps\">PGI<\/span> soft\u00adware includes high-per\u00adfor\u00admance par\u00adal\u00adlel For\u00adtran, C and C++ com\u00adpi\u00adlers and tools for work\u00adsta\u00adtions, ser\u00advers and clus\u00adters based on x64 pro\u00adces\u00adsors from Intel and <span class=\"caps\">AMD<\/span>, and <span class=\"caps\">HPC<\/span> acce\u00adle\u00adra\u00adtors from <span class=\"caps\">NVIDIA<\/span> and <span class=\"caps\">AMD<\/span>. More infor\u00adma\u00adti\u00adon is available at <a href=\"http:\/\/www.pgroup.com\">www.pgroup.com<\/a>, <a href=\"mailto:sales@pgroup.com\">sales@pgroup.com<\/a> or by cal\u00adling (503) 682\u20112806.<\/p>\n<p><strong>Keep Cur\u00adrent on <span class=\"caps\">NVIDIA<\/span><\/strong><\/p>\n<p>Sub\u00adscri\u00adbe to the <a href=\"http:\/\/blogs.nvidia.com\/\"><span class=\"caps\">NVIDIA<\/span> blog<\/a>, fol\u00adlow us on <a href=\"https:\/\/www.facebook.com\/NVIDIA\">Face\u00adbook<\/a>, <a href=\"https:\/\/plus.google.com\/+NVIDIA\">Goog\u00adle+<\/a>, <a href=\"https:\/\/twitter.com\/nvidia\">Twit\u00adter<\/a>, <a href=\"http:\/\/www.linkedin.com\/company\/nvidia\">Lin\u00adke\u00addIn<\/a> and <a href=\"http:\/\/instagram.com\/nvidia\">Insta\u00adgram<\/a>, and view <span class=\"caps\">NVIDIA<\/span> vide\u00ados on <a href=\"http:\/\/www.youtube.com\/user\/nvidia\">You\u00adTube<\/a> and images on <a href=\"https:\/\/www.flickr.com\/photos\/nvidia\">Flickr<\/a>.<\/p>\n<p><span style=\"font-size: 8pt;\"><strong>About <span class=\"caps\">NVIDIA<\/span><em><br>\n<\/em><\/strong>Sin\u00adce 1993, <a href=\"http:\/\/www.nvidia.com\/#source=pr\"><span class=\"caps\">NVIDIA<\/span><\/a> (<span class=\"caps\">NASDAQ<\/span>: <span class=\"caps\">NVDA<\/span>) has pio\u00adnee\u00adred the art and sci\u00adence of <a href=\"http:\/\/www.nvidia.com\/object\/visual-computing.html#source=pr\">visu\u00adal com\u00adpu\u00adting<\/a>. The company\u2019s tech\u00adno\u00adlo\u00adgies are trans\u00adforming a world of dis\u00adplays into a world of inter\u00adac\u00adti\u00adve dis\u00adco\u00advery \u2014 for ever\u00adyo\u00adne from gamers to sci\u00aden\u00adtists, and con\u00adsu\u00admers to enter\u00adpri\u00adse cus\u00adto\u00admers. More infor\u00adma\u00adti\u00adon at <a href=\"http:\/\/nvidianews.nvidia.com#source=pr\">http:\/\/nvidianews.nvidia.com\/<\/a> and <a href=\"http:\/\/blogs.nvidia.com#source=pr\">http:\/\/blogs.nvidia.com\/<\/a>.<\/span><\/p>\n<p><span style=\"font-size: 8pt;\"># # #<\/span><\/p>\n<p><span style=\"font-size: 8pt;\">Cer\u00adtain state\u00adments in this press release inclu\u00adding, but not limi\u00adt\u00aded to, state\u00adments as to: the fea\u00adtures, bene\u00adfits and avai\u00adla\u00adbi\u00adli\u00adty of <span class=\"caps\">PGI<\/span> Acce\u00adle\u00adra\u00adtor com\u00adpi\u00adlers with added sup\u00adport for Open\u00adACC are for\u00adward-loo\u00adking state\u00adments that are sub\u00adject to risks and uncer\u00adtain\u00adties that could cau\u00adse results to be mate\u00adri\u00adal\u00adly dif\u00adfe\u00adrent than expec\u00adta\u00adti\u00adons. Important fac\u00adtors that could cau\u00adse actu\u00adal results to dif\u00adfer mate\u00adri\u00adal\u00adly include: glo\u00adbal eco\u00adno\u00admic con\u00addi\u00adti\u00adons; our reli\u00adance on third par\u00adties to manu\u00adfac\u00adtu\u00adre, assem\u00adble, packa\u00adge and test our pro\u00adducts; the impact of tech\u00adno\u00adlo\u00adgi\u00adcal deve\u00adlo\u00adp\u00adment and com\u00adpe\u00adti\u00adti\u00adon; deve\u00adlo\u00adp\u00adment of new pro\u00adducts and tech\u00adno\u00adlo\u00adgies or enhance\u00adments to our exis\u00adting pro\u00adduct and tech\u00adno\u00adlo\u00adgies; mar\u00adket accep\u00adtance of our pro\u00adducts or our part\u00adners\u2019 pro\u00adducts; design, manu\u00adfac\u00adtu\u00adring or soft\u00adware defects; chan\u00adges in con\u00adsu\u00admer pre\u00adfe\u00adren\u00adces or demands; chan\u00adges in indus\u00adtry stan\u00addards and inter\u00adfaces; unex\u00adpec\u00adted loss of per\u00adfor\u00admance of our pro\u00adducts or tech\u00adno\u00adlo\u00adgies when inte\u00adgra\u00adted into sys\u00adtems; as well as other fac\u00adtors detail\u00aded from time to time in the reports <span class=\"caps\">NVIDIA<\/span> files with the Secu\u00adri\u00adties and Exch\u00adan\u00adge Com\u00admis\u00adsi\u00adon, or <span class=\"caps\">SEC<\/span>, inclu\u00adding its Form 10\u2011Q for the fis\u00adcal peri\u00adod ended July 26, 2015. Copies of reports filed with the <span class=\"caps\">SEC<\/span> are pos\u00adted on the company\u2019s web\u00adsite and are available from <span class=\"caps\">NVIDIA<\/span> wit\u00adhout char\u00adge. The\u00adse for\u00adward-loo\u00adking state\u00adments are not gua\u00adran\u00adtees of future per\u00adfor\u00admance and speak only as of the date hereof, and, except as requi\u00adred by law, <span class=\"caps\">NVIDIA<\/span> dis\u00adclaims any obli\u00adga\u00adti\u00adon to update the\u00adse for\u00adward-loo\u00adking state\u00adments to reflect future events or circumstances.<\/span><\/p>\n<p><span style=\"font-size: 8pt;\">\u00a9 2015 <span class=\"caps\">NVIDIA<\/span> Cor\u00adpo\u00adra\u00adti\u00adon. All rights reser\u00adved. <span class=\"caps\">NVIDIA<\/span>, the <span class=\"caps\">NVIDIA<\/span> logo, <span class=\"caps\">PGI<\/span> Acce\u00adle\u00adra\u00adtor and Open\u00adACC are trade\u00admarks and\/or regis\u00adtered trade\u00admarks of <span class=\"caps\">NVIDIA<\/span> Cor\u00adpo\u00adra\u00adti\u00adon in the U.S. and other count\u00adries. Other com\u00adpa\u00adny and pro\u00adduct names may be trade\u00admarks of the respec\u00adti\u00adve com\u00adpa\u00adnies with which they are asso\u00adcia\u00adted. Fea\u00adtures, pri\u00adcing, avai\u00adla\u00adbi\u00adli\u00adty and spe\u00adci\u00adfi\u00adca\u00adti\u00adons are sub\u00adject to chan\u00adge wit\u00adhout notice.<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"<p><span class=\"caps\">SANTA<\/span> <span class=\"caps\">CLARA<\/span>, Calif.\u2014Oct. 29, 2015\u2014<span class=\"caps\">NVIDIA<\/span> today announ\u00adced avai\u00adla\u00adbi\u00adli\u00adty of ver\u00adsi\u00adon 15.10 of the <span class=\"caps\">PGI<\/span> Acce\u00adle\u00adra\u00adtor\u2122 For\u00adtran, C and C++ com\u00adpi\u00adlers, adding sup\u00adport for the Open\u00adACC\u00ae direc\u00adti\u00adves-based par\u00adal\u00adlel pro\u00adgramming stan\u00addard on x86 archi\u00adtec\u00adtu\u00adre mul\u00adti\u00adco\u00adre micro\u00adpro\u00adces\u00adsors. (\u2026) <a class=\"moretag\" href=\"https:\/\/www.planet3dnow.de\/cms\/20750-pgi-accelerator-compilers-add-openacc-support-for-x86-multicore-cpus\/\">Wei\u00adter\u00adle\u00adsen&nbsp;\u00bb<\/a><\/p>\n","protected":false},"author":8,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"wp_typography_post_enhancements_disabled":false,"ngg_post_thumbnail":0,"footnotes":""},"categories":[22],"tags":[],"class_list":["post-20750","post","type-post","status-publish","format-standard","hentry","category-pressemitteilungen","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/20750","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=20750"}],"version-history":[{"count":1,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/20750\/revisions"}],"predecessor-version":[{"id":20751,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/20750\/revisions\/20751"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=20750"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=20750"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=20750"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}