In die­ser Sek­ti­on wer­den Pres­se­mit­tei­lun­gen der Her­stel­ler 1:1 ver­öf­fent­licht. Pla­net 3DNow! über­nimmt weder für die Rich­tig­keit der Anga­ben noch für die Art der For­mu­lie­rung jed­we­de Ver­ant­wor­tung. Die Anga­ben kön­nen sub­jek­tiv sein und geben nicht not­wen­di­ger­wei­se die Mei­nung der auf Pla­net 3DNow! schrei­ben­den Autoren wieder.

CAPS Announcing Full Support for OpenACC 2.0 in its Compilers

Verfasst vonDr@

DENVER, Colo., Nov. 14 – CAPS ent­re­pri­se, lea­ding pro­vi­der of HPC com­pi­ler tech­no­lo­gy, is plea­sed to announ­ce sup­port for the Open­ACC 2.0 par­al­lel pro­gramming stan­dard for acce­le­ra­tors in its flag­ship CAPS Compilers.

CAPS ent­re­pri­se is a foun­ding mem­ber of the Open­ACC stan­dards group and has been high­ly invol­ved in the deve­lo­p­ment of the Open­ACC 2.0 spe­ci­fi­ca­ti­on. Based on a com­mon set of direc­ti­ves for C and For­tran lan­guages, Open­ACC enables pro­gramm­ers to easi­ly take advan­ta­ge of the pro­ces­sing power of hete­ro­ge­neous many-core architectures.

CAPS com­pi­lers tar­get the who­le ran­ge of today’s acce­le­ra­tors: NVIDIA GPU acce­le­ra­tors, AMD GPUs/APUs, and Intel Xeon Phi, giving users a hard­ware-inde­pen­dent soft­ware solu­ti­on based on a wide­ly used pro­gramming standard.

Accor­ding to Dr. Sté­pha­ne Chau­veau, CAPS Tech­ni­cal Direc­tor, “The Open­ACC 2.0 stan­dard offers seve­ral major impro­ve­ments over Open­ACC 1.0. Among­st them, users should app­re­cia­te a bet­ter asyn­chro­no­us beha­vi­or, a well-defi­ned beha­vi­or for exter­nal func­tion calls within acce­le­ra­ted regi­ons and seve­ral data manage­ment impro­ve­ments (such as mul­ti­di­men­sio­nal C/C++ arrays) and glo­bal varia­bles.” The Open­ACC stan­dards group is alre­a­dy working hard on the spe­ci­fi­ca­ti­on of the next ver­si­on that is the fruit of feed­backs from the Open­ACC community.

“CAPS has an excel­lent histo­ry of sup­port­ing mul­ti­ple CPU archi­tec­tures, inclu­ding x86 and ARM, as well as mul­ti­ple acce­le­ra­tors, in their com­pi­ler solu­ti­ons,” said Dun­can Poo­le, pre­si­dent of the Open­ACC stan­dards group. “Their solu­ti­on pro­du­ces eit­her Open­CL or NVIDIA CUDA, which gives pro­gramm­ers a hig­her-level lan­guage alter­na­ti­ve for pro­gramming HPC systems.”

The Oak Ridge Natio­nal Labo­ra­to­ry is fun­ding the sup­port of the Open­ACC stan­dard in the CAPS Com­pi­lers. Oscar Her­nan­dez, a rese­arch staff at ORNL and secre­ta­ry of the Open­ACC stan­dards group, notes, “This is an important mile­stone for CAPS that will help us to port more appli­ca­ti­ons to Titan or other acce­le­ra­tor-based sys­tems using Open­ACC 2.0. Our users are eager to start using the new Open­ACC 2.0 features.”

Full sup­port for the Open­ACC 2.0 stan­dard is expec­ted in Decem­ber 2013 with par­ti­al sup­port alre­a­dy available in CAPS Compilers.

CAPS will be at Super­com­pu­ting 2013 (SC13), Novem­ber, 18–21, in Den­ver, Colo­ra­do, booth #1127 to show­ca­se its flag­ship com­pi­ler. CAPS fel­lows will also be pre­sent in the Open­ACC booth, #4508.

About CAPS entreprise

CAPS ent­re­pri­se is a lea­ding pro­vi­der of solu­ti­ons for pro­gramming and deploy­ing appli­ca­ti­ons on many-core sys­tems. Its source-to-source com­pi­lers are based on C, C++, and FORTRAN direc­ti­ves and Open­ACC. CAPS com­pi­lers tar­get various archi­tec­tures, inclu­ding NVIDIA GPU acce­le­ra­tors, AMD GPUs and APUs, and Intel Xeon Phi. www.caps-entreprise.com

—–

Source: CAPS entreprise