{"id":27270,"date":"2016-11-15T08:09:16","date_gmt":"2016-11-15T07:09:16","guid":{"rendered":"http:\/\/www.planet3dnow.de\/cms\/?p=27270"},"modified":"2016-11-15T08:18:38","modified_gmt":"2016-11-15T07:18:38","slug":"amd-releases-new-version-of-rocm-the-most-versatile-open-source-platform-for-gpu-computing","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/27270-amd-releases-new-version-of-rocm-the-most-versatile-open-source-platform-for-gpu-computing\/","title":{"rendered":"<span class=\"caps\">AMD<\/span> Releases New Version of ROCm, the Most Versatile Open Source Platform for <span class=\"caps\">GPU<\/span> Computing"},"content":{"rendered":"<p style=\"text-align: center;\">\u2014 Sup\u00adport for new <span class=\"caps\">GPU<\/span> hard\u00adware, math libra\u00adri\u00ades and modern pro\u00adgramming lan\u00adguages fur\u00adther unlock the power of <span class=\"caps\">GPU<\/span> computing&nbsp;\u2014<\/p>\n<p>&nbsp;<\/p>\n<p><span class=\"caps\">SALT<\/span> <span class=\"caps\">LAKE<\/span> <span class=\"caps\">CITY<\/span>\u2014 Nov. 14, 2016 \u2014 At <span class=\"caps\">SC16<\/span>, <span class=\"caps\">AMD<\/span> (<span class=\"caps\">NASDAQ<\/span>: <span class=\"caps\">AMD<\/span>) today announ\u00adced a new release of Rade\u00adon Open Com\u00adpu\u00adte Plat\u00adform (ROCm) fea\u00adturing soft\u00adware sup\u00adport of new Rade\u00adon <span class=\"caps\">GPU<\/span> hard\u00adware, new math libra\u00adri\u00ades, and a rich foun\u00adda\u00adti\u00adon of modern pro\u00adgramming lan\u00adguages, desi\u00adgned to speed deve\u00adlo\u00adp\u00adment of high-per\u00adfor\u00admance, ener\u00adgy-effi\u00adci\u00adent hete\u00adro\u00adge\u00adneous com\u00adpu\u00adting sys\u00adtems. <span class=\"caps\">AMD<\/span> also announ\u00adced plan\u00adned sup\u00adport of Open\u00adCL\u2122 and for a wide ran\u00adge of CPUs in upco\u00adming releases of ROCm, inclu\u00adding sup\u00adport for <span class=\"caps\">AMD<\/span>\u2019s upco\u00adming \u201cZen\u201d-based CPUs, Cavi\u00adum Thun\u00adderX CPUs, and <span class=\"caps\">IBM<\/span> Power 8 CPUs. The advan\u00adces fur\u00adther cement ROCm\u2019s posi\u00adti\u00adon as the most ver\u00adsa\u00adti\u00adle open source plat\u00adform for <span class=\"caps\">GPU<\/span> computing.<\/p>\n<p><span class=\"dquo\">\u201c<\/span>Rade\u00adon Open Com\u00adpu\u00adte is a plat\u00adform for a new era of <span class=\"caps\">GPU<\/span> pro\u00adblem-sol\u00adving, desi\u00adgned to harness the power of open source soft\u00adware to unlock new solu\u00adti\u00adons for <span class=\"caps\">HPC<\/span> and hypers\u00adca\u00adle com\u00adpu\u00adting,\u201d said Raja Kodu\u00adri, seni\u00ador vice pre\u00adsi\u00addent and chief archi\u00adtect, Rade\u00adon Tech\u00adno\u00adlo\u00adgies Group, <span class=\"caps\">AMD<\/span>. \u201cToday\u2019s release of ROCm gives deve\u00adlo\u00adpers con\u00adsum\u00adma\u00adte fle\u00adxi\u00adbi\u00adli\u00adty in whe\u00adre and how they use <span class=\"caps\">GPU<\/span> com\u00adpu\u00adte. We\u2019re exci\u00adted to show off some of the\u00adse incre\u00addi\u00adble appli\u00adca\u00adti\u00adons at&nbsp;<span class=\"caps\">SC16<\/span>.\u201d<\/p>\n<p><span class=\"dquo\">\u201c<\/span>An open source approach to hete\u00adro\u00adge\u00adneous com\u00adpu\u00adting can help bring the bene\u00adfits of high-per\u00adfor\u00admance com\u00adpu\u00adting direct\u00adly to deve\u00adlo\u00adpers, pro\u00advi\u00adding the fle\u00adxi\u00adbi\u00adli\u00adty for them to use available com\u00adpu\u00adte resour\u00adces and extra\u00adct the best pos\u00adsi\u00adble per\u00adfor\u00admance from their available hard\u00adware,\u201d said Deni\u00adse Dumas, vice pre\u00adsi\u00addent, Plat\u00adform Engi\u00adnee\u00adring, Red Hat. \u201cAs a lea\u00adder in hel\u00adping to nur\u00adtu\u00adre and main\u00adtain open source com\u00admu\u00adni\u00adties, we are plea\u00adsed to see that <span class=\"caps\">AMD<\/span> has elec\u00adted to take a com\u00admu\u00adni\u00adty-focu\u00adsed approach to&nbsp;ROCm.\u201d<\/p>\n<p><span class=\"dquo\">\u201c<\/span>The com\u00adbi\u00adna\u00adti\u00adon of <span class=\"caps\">AMD<\/span>\u2019s <span class=\"caps\">GPU<\/span> sup\u00adport of Thun\u00adderX\u00ae along with the breadth of fea\u00adtures of the ROCm soft\u00adware release is a valuable addi\u00adti\u00adon to Cavium\u2019s over\u00adall solu\u00adti\u00adon for the <span class=\"caps\">HPC<\/span> and Hypers\u00adca\u00adle ser\u00adver mar\u00adket,\u201d said Lar\u00adry Wikeli\u00adus, Vice Pre\u00adsi\u00addent of the Soft\u00adware Eco\u00adsys\u00adtem and Solu\u00adti\u00adons Group at Cavi\u00adum, Inc. \u201cToday\u2019s announce\u00adment by <span class=\"caps\">AMD<\/span> is an excel\u00adlent com\u00adple\u00adment to the inno\u00adva\u00adti\u00adve fea\u00adtures of Thun\u00adderX\u00ae that include 48 ARMv8 cores, inte\u00adgra\u00adted I\/O and dual socket con\u00adfi\u00adgu\u00adra\u00adti\u00adon sup\u00adport that our cus\u00adto\u00admers and part\u00adners are deploy\u00ading today in the\u00adse markets\u201d<\/p>\n<p>The new release of ROCm intro\u00addu\u00adces a wide ran\u00adge of updates, including:<\/p>\n<ul>\n<li>Expan\u00added <span class=\"caps\">GPU<\/span> sup\u00adport \u2013 ROCm now sup\u00adports all Pola\u00adris archi\u00adtec\u00adtu\u00adre-based gra\u00adphics pro\u00adducts, inclu\u00adding the Rade\u00adon\u2122 <span class=\"caps\">RX<\/span> 460, 470 and 480 gra\u00adphics cards, and the Rade\u00adon\u2122 Pro <span class=\"caps\">WX<\/span> 7100, 5100 and 4100 GPUs, gro\u00adwing the rich port\u00adfo\u00adlio of devices sup\u00adport\u00aded by ROCm devices. The Pola\u00adris archi\u00adtec\u00adtu\u00adre is spe\u00adci\u00adfi\u00adcal\u00adly desi\u00adgned to bene\u00adfit low-level pro\u00adgramming, hel\u00adping deve\u00adlo\u00adpers to extra\u00adct the most from the hardware.<\/li>\n<li>ROCm Vir\u00adtua\u00adliza\u00adti\u00adon of the <span class=\"caps\">GPU<\/span> hard\u00adware via <span class=\"caps\">OS<\/span> Con\u00adtai\u00adners and Linux\u00ae\u2019s Ker\u00adnel Vir\u00adtu\u00adal Machi\u00adne (<span class=\"caps\">KVM<\/span>) \u2014 ROCm now sup\u00adports Docker con\u00adtai\u00adne\u00adriza\u00adti\u00adon, allo\u00adwing end-users to sim\u00adpli\u00adfy the deploy\u00adment of an appli\u00adca\u00adti\u00adon in ROCm-enab\u00adled Linux ser\u00adver envi\u00adron\u00adments. ROCm also sup\u00adports <span class=\"caps\">GPU<\/span> Hard\u00adware Vir\u00adtua\u00adliza\u00adti\u00adon via <span class=\"caps\">KVM<\/span> pass-through to allow the bene\u00adfits of hard\u00adware-acce\u00adle\u00adra\u00adted <span class=\"caps\">GPU<\/span> com\u00adpu\u00adting in vir\u00adtua\u00adli\u00adzed solutions.<br>\n\u00b7 Hete\u00adro\u00adge\u00adneous Com\u00adpu\u00adte Com\u00adpi\u00adler (<span class=\"caps\">HCC<\/span>) \u2013 <span class=\"caps\">HCC<\/span> is a sin\u00adgle source <span class=\"caps\">ISO<\/span> C++ 11\/14 com\u00adpi\u00adler for both <span class=\"caps\">CPU<\/span> and <span class=\"caps\">GPU<\/span>, with sup\u00adport for the C++17 \u201cPar\u00adal\u00adlel Stan\u00addard Tem\u00adp\u00adla\u00adte Libra\u00adry\u201d. It is built on a rich com\u00adpi\u00adler infra\u00adstruc\u00adtu\u00adre inclu\u00adding LLVM-based <span class=\"caps\">GCN<\/span> <span class=\"caps\">ISA<\/span> code gene\u00adra\u00adti\u00adon with assem\u00adbler and dis\u00adas\u00adsem\u00adbler support.<\/li>\n<li>Hete\u00adro\u00adge\u00adneous-Com\u00adpu\u00adte Inter\u00adface for Por\u00adta\u00adbi\u00adli\u00adty (<span class=\"caps\">HIP<\/span>) \u2013 <span class=\"caps\">HIP<\/span> enables deve\u00adlo\u00adpers to port <span class=\"caps\">CUDA<\/span> appli\u00adca\u00adti\u00adons to ROCm using <span class=\"caps\">HIPIFY<\/span> which auto\u00adma\u00adtes the con\u00adver\u00adsi\u00adon to the <span class=\"caps\">HIP<\/span> ker\u00adnel lan\u00adguage and run\u00adtime <span class=\"caps\">API<\/span>, crea\u00adting por\u00adta\u00adble appli\u00adca\u00adti\u00adons that can run on vir\u00adtual\u00adly any <span class=\"caps\">GPU<\/span> using eit\u00adher <span class=\"caps\">NVIDIA<\/span>\u2019s <span class=\"caps\">CUDA<\/span> Com\u00adpi\u00adler or&nbsp;<span class=\"caps\">HCC<\/span>.<\/li>\n<li>New Math Acce\u00adle\u00adra\u00adti\u00adon Libra\u00adri\u00ades \u2013 ROCm intro\u00addu\u00adces sup\u00adport for new advan\u00adced math acce\u00adle\u00adra\u00adti\u00adon libra\u00adri\u00ades with sup\u00adport for <span class=\"caps\">BLAS<\/span>, <span class=\"caps\">FFT<\/span> and N\u2011dimensional ten\u00adsor contractions.<\/li>\n<\/ul>\n<p>In addi\u00adti\u00adon to today\u2019s release of ROCm, <span class=\"caps\">AMD<\/span> also announ\u00adced that it is working to expand the ROCm eco\u00adsys\u00adtem through increased ser\u00adver <span class=\"caps\">CPU<\/span> sup\u00adport, and plan\u00adned sup\u00adport of Open\u00adCL. Upco\u00adming releases of ROCm are expec\u00adted to sup\u00adport <span class=\"caps\">AMD<\/span> \u201cZen\u201d-based x86 CPUs, <span class=\"caps\">ARM<\/span> AArch64 <span class=\"caps\">CPU<\/span> archi\u00adtec\u00adtu\u00adre start\u00ading with Cavi\u00adum Thun\u00adderX pro\u00adces\u00adsors, as well as <span class=\"caps\">IBM<\/span> Power 8 CPUs and ser\u00advers to meet cus\u00adto\u00admers\u2019 gro\u00adwing hypers\u00adca\u00adle and <span class=\"caps\">HPC<\/span> needs. Open\u00adCL is being built on a ROCm run\u00adtime and com\u00adpi\u00adler foun\u00adda\u00adti\u00adon which give much grea\u00adter low-level con\u00adtrol of the <span class=\"caps\">GPU<\/span> via its direct-to-metal access.<\/p>\n<p>The new ver\u00adsi\u00adon of Rade\u00adon Open Com\u00adpu\u00adte is available now at https:\/\/github.com\/RadeonOpenCompute\/ROCm.<\/p>\n<p><b>Demons\u00adtra\u00adti\u00adons at&nbsp;<span class=\"caps\">SC16<\/span><\/b><\/p>\n<p>At <span class=\"caps\">SC16<\/span> in booth 1431, <span class=\"caps\">AMD<\/span> will showcase:<\/p>\n<ul>\n<li>ROCm Tech\u00adno\u00adlo\u00adgy Clus\u00adter run\u00adning Machi\u00adne Lear\u00adning Code on Super\u00admi\u00adcro\u00ae servers<\/li>\n<li>Port\u00ading the <span class=\"caps\">CUDA<\/span> appli\u00adca\u00adti\u00adon Caf\u00adfe via <span class=\"caps\">HIP<\/span> Port\u00ading&nbsp;Tool<\/li>\n<li>Ray-tra\u00adcing and <span class=\"caps\">VR<\/span> visua\u00adliza\u00adti\u00adon for <span class=\"caps\">HPC<\/span> with <span class=\"caps\">AMD<\/span> Fire\u00adPro\u2122 <span class=\"caps\">S9300<\/span> <span class=\"caps\">X2<\/span> <span class=\"amp\">&amp;<\/span> Rade\u00adon\u2122 <span class=\"caps\">R9<\/span> Nano&nbsp;GPUs<\/li>\n<li>OpenMP 4.5 Inter\u00adope\u00adra\u00adbi\u00adli\u00adty tar\u00adge\u00adting mul\u00adti\u00adple GPUs <span class=\"amp\">&amp;<\/span> platforms<\/li>\n<li><span class=\"caps\">IBM<\/span> Power8 ser\u00adver with <span class=\"caps\">AMD<\/span> Fire\u00adPro\u2122 <span class=\"caps\">S9170<\/span> Ser\u00adver <span class=\"caps\">GPU<\/span> run\u00adning&nbsp;ROCm<\/li>\n<li>Pen\u00adgu\u00adin Com\u00adpu\u00adting Tun\u00addra\u2122 Extre\u00adme ARMv8 Thun\u00adderX based ser\u00adver with Rade\u00adon\u2122 <span class=\"caps\">RX<\/span> 460 run\u00adning&nbsp;ROCm<\/li>\n<li>In-situ ren\u00adde\u00adring with Head\u00adless OpenGL\/<span class=\"caps\">EGL<\/span> Interop Open\u00adCL on&nbsp;ROCm<\/li>\n<\/ul>\n<p><b>Sup\u00adport\u00ading Resources<\/b><\/p>\n<ul>\n<li><span class=\"caps\">AMD<\/span> Deve\u00adlo\u00adper Central<\/li>\n<li>Learn more about Hete\u00adro\u00adge\u00adneous Sys\u00adtem Architecture<\/li>\n<li>Learn more about GPUOpen<\/li>\n<li>Learn more about&nbsp;ROCm<\/li>\n<li>Beco\u00adme a fan of <span class=\"caps\">AMD<\/span> on Facebook<\/li>\n<li>Fol\u00adlow <span class=\"caps\">AMD<\/span> on Twit\u00adter @<span class=\"caps\">AMD<\/span><\/li>\n<li>Fol\u00adlow Rade\u00adonTM Pro on Twitter<\/li>\n<li>Fol\u00adlow Rade\u00adon\u2122 gra\u00adphics on Twitter<\/li>\n<\/ul>\n<p><b>About <span class=\"caps\">AMD<\/span><\/b><\/p>\n<p>For more than 45 years <span class=\"caps\">AMD<\/span> has dri\u00adven inno\u00adva\u00adti\u00adon in high-per\u00adfor\u00admance com\u00adpu\u00adting, gra\u00adphics, and visua\u00adliza\u00adti\u00adon tech\u00adno\u00adlo\u00adgies \u2015 the buil\u00adding blocks for gam\u00ading, immersi\u00adve plat\u00adforms, and the dat\u00ada\u00adcen\u00adter. Hundreds of mil\u00adli\u00adons of con\u00adsu\u00admers, lea\u00adding For\u00adtu\u00adne 500 busi\u00adnesses, and cut\u00adting-edge sci\u00aden\u00adti\u00adfic rese\u00adarch faci\u00adli\u00adties around the world rely on <span class=\"caps\">AMD<\/span> tech\u00adno\u00adlo\u00adgy dai\u00adly to impro\u00adve how they live, work, and play. <span class=\"caps\">AMD<\/span> employees around the world are focu\u00adsed on buil\u00adding gre\u00adat pro\u00adducts that push the boun\u00adda\u00adries of what is pos\u00adsi\u00adble. For more infor\u00adma\u00adti\u00adon about how <span class=\"caps\">AMD<\/span> is enab\u00adling today and inspi\u00adring tomor\u00adrow, visit the <span class=\"caps\">AMD<\/span> (<span class=\"caps\">NASDAQ<\/span>: <span class=\"caps\">AMD<\/span>) web\u00adsite, blog, Face\u00adbook and Twit\u00adter&nbsp;pages.<\/p>\n<p>\u201430\u2014<\/p>\n<p><span class=\"caps\">AMD<\/span>, the <span class=\"caps\">AMD<\/span> Arrow logo, Rade\u00adon and com\u00adbi\u00adna\u00adti\u00adons the\u00adreof, are trade\u00admarks of Advan\u00adced Micro Devices, Inc. Open\u00adCL is a trade\u00admark of Apple Inc. used by per\u00admis\u00adsi\u00adon by Khro\u00adnos Group, Inc. <span class=\"caps\">ARM<\/span> is a regis\u00adtered trade\u00admark of <span class=\"caps\">ARM<\/span> Limi\u00adt\u00aded in the <span class=\"caps\">UK<\/span> and other count\u00adries. Other names are for infor\u00adma\u00adtio\u00adnal pur\u00adpo\u00adses only and may be trade\u00admarks of their respec\u00adti\u00adve owners.<\/p>\n","protected":false},"excerpt":{"rendered":"<p><span class=\"caps\">SALT<\/span> <span class=\"caps\">LAKE<\/span> <span class=\"caps\">CITY<\/span>\u2014 Nov. 14, 2016 \u2014 At <span class=\"caps\">SC16<\/span>, <span class=\"caps\">AMD<\/span> (<span class=\"caps\">NASDAQ<\/span>: <span class=\"caps\">AMD<\/span>) today announ\u00adced a new release of Rade\u00adon Open Com\u00adpu\u00adte Plat\u00adform (ROCm) fea\u00adturing soft\u00adware sup\u00adport of new Rade\u00adon <span class=\"caps\">GPU<\/span> hard\u00adware, new math libra\u00adri\u00ades, and a rich foun\u00adda\u00adti\u00adon of modern pro\u00adgramming lan\u00adguages, desi\u00adgned to speed deve\u00adlo\u00adp\u00adment of high-per\u00adfor\u00admance, ener\u00adgy-effi\u00adci\u00adent hete\u00adro\u00adge\u00adneous com\u00adpu\u00adting sys\u00adtems. <span class=\"caps\">AMD<\/span> also announ\u00adced plan\u00adned sup\u00adport of Open\u00adCL\u2122 and for a wide ran\u00adge of CPUs in upco\u00adming releases of ROCm, inclu\u00adding sup\u00adport for <span class=\"caps\">AMD<\/span>\u2019s upco\u00adming \u201cZen\u201d-based CPUs, Cavi\u00adum Thun\u00adderX CPUs, and <span class=\"caps\">IBM<\/span> Power 8 CPUs. The advan\u00adces fur\u00adther cement ROCm\u2019s posi\u00adti\u00adon as the most ver\u00adsa\u00adti\u00adle open source plat\u00adform for <span class=\"caps\">GPU<\/span> com\u00adpu\u00adting. (\u2026) <a class=\"moretag\" href=\"https:\/\/www.planet3dnow.de\/cms\/27270-amd-releases-new-version-of-rocm-the-most-versatile-open-source-platform-for-gpu-computing\/\">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-27270","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\/27270","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=27270"}],"version-history":[{"count":6,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/27270\/revisions"}],"predecessor-version":[{"id":27276,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/27270\/revisions\/27276"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=27270"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=27270"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=27270"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}