November 14, 2013

CAPS Announcing Full Support for OpenACC 2.0 in its Compilers

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.


Source: CAPS entreprise