NVIDIA, Cray, the Portland Group (PGI), and CAPS present OpenACC, a new open programming standard for parallel computing.
In an effort to make it easier for programmers to take advantage of parallel computing, NVIDIA, Cray Inc., the Portland Group (PGI), and CAPS enterprise announced today a new parallel-programming standard, known as OpenACC.
Initially developed by PGI, Cray, and NVIDIA, with support from CAPS, OpenACC is a new open parallel programming standard designed to enable the millions of scientific and technical programmers to easily take advantage of the transformative power of heterogeneous CPU/GPU computing systems.
OpenACC allows parallel programmers to provide simple hints, known as "directives," to the compiler, identifying which areas of code to accelerate, without requiring programmers to modify or adapt the underlying code itself. By exposing parallelism to the compiler, directives allow the compiler to do the detailed work of mapping the computation onto the accelerator.
OpenACC is anticipated to benefit a broad range of programmers working in chemistry, biology, physics, data analytics, weather and climate, intelligence, and many other fields. Existing compilers from Cray, PGI and CAPS are expected to provide initial support for the OpenACC standard beginning in the first quarter of 2012. The OpenACC standard is fully compatible and interoperable with the NVIDIA CUDA parallel programming architecture, which is designed to allow detailed control over the accelerator for maximum performance tuning.
Directives provide a common code base that is multi-platform and multi-vendor compatible, offering an ideal way to preserve investment in legacy applications by enabling an easy migration path to accelerated computing. Based on recent data, the majority of developers have reported 2x to 10x increases in application performance in as little as two weeks when using existing directive-based compilers.1
"Compiler directives are integral as we continue the build-out and deployment of the Titan GPU-accelerated supercomputer," said Buddy Bland, Titan project director at Oak Ridge National Laboratory, referring to a system expected to be the world's fastest supercomputer. "Our ultimate goal is to have all Titan supercomputing code run on hybrid CPU/GPU nodes, and OpenACC will enable programmers to develop portable applications that maximize the performance and power efficiency benefits of this architecture."
"OpenACC represents a major development for the scientific community," said Jeffrey Vetter, joint professor in the Computational Science and Engineering School of the College of Computing at Georgia institute of Technology. "Programming models for open science by definition need to be flexible, open and portable across multiple platforms; OpenACC is well designed to fill this need. It provides a valuable new tool to empower the vast numbers of domain scientists who could benefit from application acceleration, but who may not have the funding or expertise to port their code to emerging architectures."
"I am enthusiastic about the future of accelerator technologies," said Michael Wong, CEO of the OpenMP Architecture Review Board. "The OpenACC announcement highlights the technically impressive initiative undertaken by members of the OpenMP Working Group on Accelerators. I look forward to working with all four companies within the OpenMP organization to merge OpenACC with other ideas to create a common specification which extends OpenMP to support accelerators. We look forward to incorporating accelerator support with the full support of all OpenMP members in a future version of the OpenMP specification."
More information about OpenACC, as well as the OpenACC specification, is available today at OpenACC-standard.org. OpenACC is based largely on the PGI and Cray accelerator programming models. NVIDIA, Cray, PGI, and CAPS are members of the OpenMP subcommittee on accelerators, and intend to work within that organization to come to a common standard.