Altera has unveiled new accelerators for Power chips and Nvidia is to introduce compilers to run x86 Linux apps on the processors more easily.
Silicon Valley-based Altera makes field-programmable gate arrays (FPGAs) that are used to speed up processing (pictured). It announced support for IBM’s OpenPower Foundation in November, 2013, and is listed as a platinum (highest level) member. Its main competitor Xilinx – credited with the invention of FPGAs – is a silver-level member.
The announcement at this week’s Supercomputing 2014 conference in New Orleans marks the arrival of the second bolt-on for Power processors from an OpenPower member. Big Blue launched the first Power System to feature new GNU accelerators from Nvidia last month. The Nvidia/IBM link has already earned the two firms a $325m supercomputer deal from the U.S. government.
Altera has connected its FPGAs to Power8 by using IBM’s Coherent Accelerator Processor Interface (CAPI). The two companies have also worked with another OpenPower member, Nallatech, to develop an OpenPower CAPI Development Kit for Power8.
David Gamba, general manager of Alera’s computer and storage business unit, said: “Altera is at the forefront of supplying Power users reconfigurable hardware accelerators based on CAPI that are supported with an OpenCL programming model. The result is highly optimised accelerators that deliver optimal FLOPs/Watt/dollar.”
Meanwhile, Nvidia announced on Tuesday (Nov 18) that it is developing enhanced compilers to let developers run Linux x86-based GPU-accelerated apps on Power hardware with minimal effort.
The new Fortran, C and C++ compilers for Power will provide a user interface, language features, parallel programming features and optimisation capabilities that are identical to those available on PGI Linux x86 compilers.
A statement from Altera said that the new compilers will support IBM’s new GPU-accelerated Power Systems and, perhaps tellingly, “additional systems under development by members of the OpenPower Foundation”.
Douglas Miles, director of PGI compilers & tools at Nvidia, said: “Our goal is to let HPC developers recompile and run their applications on all major CPU and GPU-accelerated platforms with uniformly high performance using a common source code base. We expect most GPU-accelerated x86 applications currently built with PGI compilers will port to GPU-accelerated Power systems with a simple recompile.”