— Sup­port for new GPU hard­ware, math libra­ri­es and modern pro­gramming lan­guages fur­ther unlock the power of GPU computing —

 

SALT LAKE CITY— Nov. 14, 2016 — At SC16, AMD (NASDAQ: AMD) today announ­ced a new release of Rade­on Open Com­pu­te Plat­form (ROCm) fea­turing soft­ware sup­port of new Rade­on GPU hard­ware, new math libra­ri­es, and a rich foun­da­ti­on of modern pro­gramming lan­guages, desi­gned to speed deve­lo­p­ment of high-per­for­mance, ener­gy-effi­ci­ent hete­ro­ge­neous com­pu­ting sys­tems. AMD also announ­ced plan­ned sup­port of Open­CL™ and for a wide ran­ge of CPUs in upco­ming releases of ROCm, inclu­ding sup­port for AMD’s upco­ming “Zen”-based CPUs, Cavi­um Thun­derX CPUs, and IBM Power 8 CPUs. The advan­ces fur­ther cement ROCm’s posi­ti­on as the most ver­sa­ti­le open source plat­form for GPU computing.

“Rade­on Open Com­pu­te is a plat­form for a new era of GPU pro­blem-sol­ving, desi­gned to harness the power of open source soft­ware to unlock new solu­ti­ons for HPC and hypers­ca­le com­pu­ting,” said Raja Kodu­ri, seni­or vice pre­si­dent and chief archi­tect, Rade­on Tech­no­lo­gies Group, AMD. “Today’s release of ROCm gives deve­lo­pers con­sum­ma­te fle­xi­bi­li­ty in whe­re and how they use GPU com­pu­te. We’re exci­ted to show off some of the­se incre­di­ble appli­ca­ti­ons at SC16.”

“An open source approach to hete­ro­ge­neous com­pu­ting can help bring the bene­fits of high-per­for­mance com­pu­ting direct­ly to deve­lo­pers, pro­vi­ding the fle­xi­bi­li­ty for them to use available com­pu­te resour­ces and extra­ct the best pos­si­ble per­for­mance from their available hard­ware,” said Deni­se Dumas, vice pre­si­dent, Plat­form Engi­nee­ring, Red Hat. “As a lea­der in hel­ping to nur­tu­re and main­tain open source com­mu­ni­ties, we are plea­sed to see that AMD has elec­ted to take a com­mu­ni­ty-focu­sed approach to ROCm.”

“The com­bi­na­ti­on of AMD’s GPU sup­port of Thun­derX® along with the breadth of fea­tures of the ROCm soft­ware release is a valuable addi­ti­on to Cavium’s over­all solu­ti­on for the HPC and Hypers­ca­le ser­ver mar­ket,” said Lar­ry Wikeli­us, Vice Pre­si­dent of the Soft­ware Eco­sys­tem and Solu­ti­ons Group at Cavi­um, Inc. “Today’s announce­ment by AMD is an excel­lent com­ple­ment to the inno­va­ti­ve fea­tures of Thun­derX® that include 48 ARMv8 cores, inte­gra­ted I/O and dual socket con­fi­gu­ra­ti­on sup­port that our cus­to­mers and part­ners are deploy­ing today in the­se markets”

The new release of ROCm intro­du­ces a wide ran­ge of updates, including:

• Expan­ded GPU sup­port – ROCm now sup­ports all Pola­ris archi­tec­tu­re-based gra­phics pro­ducts, inclu­ding the Rade­on™ RX 460, 470 and 480 gra­phics cards, and the Rade­on™ Pro WX 7100, 5100 and 4100 GPUs, gro­wing the rich port­fo­lio of devices sup­port­ed by ROCm devices. The Pola­ris archi­tec­tu­re is spe­ci­fi­cal­ly desi­gned to bene­fit low-level pro­gramming, hel­ping deve­lo­pers to extra­ct the most from the hardware.
• ROCm Vir­tua­liza­ti­on of the GPU hard­ware via OS Con­tai­ners and Linux®’s Ker­nel Vir­tu­al Machi­ne (KVM) — ROCm now sup­ports Docker con­tai­ne­riza­ti­on, allo­wing end-users to sim­pli­fy the deploy­ment of an appli­ca­ti­on in ROCm-enab­led Linux ser­ver envi­ron­ments. ROCm also sup­ports GPU Hard­ware Vir­tua­liza­ti­on via KVM pass-through to allow the bene­fits of hard­ware-acce­le­ra­ted GPU com­pu­ting in vir­tua­li­zed solutions.
  · Hete­ro­ge­neous Com­pu­te Com­pi­ler (HCC) – HCC is a sin­gle source ISO C++ 11/14 com­pi­ler for both CPU and GPU, with sup­port for the C++17 “Par­al­lel Stan­dard Tem­p­la­te Libra­ry”. It is built on a rich com­pi­ler infra­struc­tu­re inclu­ding LLVM-based GCN ISA code gene­ra­ti­on with assem­bler and dis­as­sem­bler support.
• Hete­ro­ge­neous-Com­pu­te Inter­face for Por­ta­bi­li­ty (HIP) – HIP enables deve­lo­pers to port CUDA appli­ca­ti­ons to ROCm using HIPIFY which auto­ma­tes the con­ver­si­on to the HIP ker­nel lan­guage and run­time API, crea­ting por­ta­ble appli­ca­ti­ons that can run on vir­tual­ly any GPU using eit­her NVIDIA’s CUDA Com­pi­ler or HCC.
• New Math Acce­le­ra­ti­on Libra­ri­es – ROCm intro­du­ces sup­port for new advan­ced math acce­le­ra­ti­on libra­ri­es with sup­port for BLAS, FFT and N‑dimensional ten­sor contractions.

In addi­ti­on to today’s release of ROCm, AMD also announ­ced that it is working to expand the ROCm eco­sys­tem through increased ser­ver CPU sup­port, and plan­ned sup­port of Open­CL. Upco­ming releases of ROCm are expec­ted to sup­port AMD “Zen”-based x86 CPUs, ARM AArch64 CPU archi­tec­tu­re start­ing with Cavi­um Thun­derX pro­ces­sors, as well as IBM Power 8 CPUs and ser­vers to meet cus­to­mers’ gro­wing hypers­ca­le and HPC needs. Open­CL is being built on a ROCm run­time and com­pi­ler foun­da­ti­on which give much grea­ter low-level con­trol of the GPU via its direct-to-metal access.

The new ver­si­on of Rade­on Open Com­pu­te is available now at https://github.com/RadeonOpenCompute/ROCm.

Demons­tra­ti­ons at SC16

At SC16 in booth 1431, AMD will showcase:

• ROCm Tech­no­lo­gy Clus­ter run­ning Machi­ne Lear­ning Code on Super­mi­cro® servers
• Port­ing the CUDA appli­ca­ti­on Caf­fe via HIP Port­ing Tool
• Ray-tra­cing and VR visua­liza­ti­on for HPC with AMD Fire­Pro™ S9300 X2 & Rade­on™ R9 Nano GPUs
• OpenMP 4.5 Inter­ope­ra­bi­li­ty tar­ge­ting mul­ti­ple GPUs & platforms
• IBM Power8 ser­ver with AMD Fire­Pro™ S9170 Ser­ver GPU run­ning ROCm
• Pen­gu­in Com­pu­ting Tun­dra™ Extre­me ARMv8 Thun­derX based ser­ver with Rade­on™ RX 460 run­ning ROCm
• In-situ ren­de­ring with Head­less OpenGL/EGL Interop Open­CL on ROCm

Sup­port­ing Resources

• AMD Deve­lo­per Central
• Learn more about Hete­ro­ge­neous Sys­tem Architecture
• Learn more about GPUOpen
• Learn more about ROCm
• Beco­me a fan of AMD on Facebook
• Fol­low AMD on Twit­ter @AMD
• Fol­low Rade­onTM Pro on Twitter
• Fol­low Rade­on™ gra­phics on Twitter

About AMD

For more than 45 years AMD has dri­ven inno­va­ti­on in high-per­for­mance com­pu­ting, gra­phics, and visua­liza­ti­on tech­no­lo­gies ― the buil­ding blocks for gam­ing, immersi­ve plat­forms, and the dat­a­cen­ter. Hundreds of mil­li­ons of con­su­mers, lea­ding For­tu­ne 500 busi­nesses, and cut­ting-edge sci­en­ti­fic rese­arch faci­li­ties around the world rely on AMD tech­no­lo­gy dai­ly to impro­ve how they live, work, and play. AMD employees around the world are focu­sed on buil­ding gre­at pro­ducts that push the boun­da­ries of what is pos­si­ble. For more infor­ma­ti­on about how AMD is enab­ling today and inspi­ring tomor­row, visit the AMD (NASDAQ: AMD) web­site, blog, Face­book and Twit­ter pages.

—30—

AMD, the AMD Arrow logo, Rade­on and com­bi­na­ti­ons the­reof, are trade­marks of Advan­ced Micro Devices, Inc. Open­CL is a trade­mark of Apple Inc. used by per­mis­si­on by Khro­nos Group, Inc. ARM is a regis­tered trade­mark of ARM Limi­t­ed in the UK and other count­ries. Other names are for infor­ma­tio­nal pur­po­ses only and may be trade­marks of their respec­ti­ve owners.