AMD & PathScale Join OpenACC Group

Up to now the OpenACC parallel programming standard has mostly been perceived as a NVIDIA affair along with backing from the likes of Cray and PGI. Now, however, AMD and PathScale are joining the OpenACC Standards Group so hopefully we'll see greater, multi-vendor adoption of it going forward.

From the Super Computing 2014 event in New Orleans, the OpenACC Standards Group announced that AMD and PathScale are the newest members. This standards group has 18 other members.

From the press release, AMD talks up OpenACC 2.0 for AMD APUs and they look forward to evolving the standard going forward. PathScale is the compiler company many Phoronix readers should be familiar with and they're supporting OpenACC 2.0 with their ENZO 2014 product.

OpenACC works in a similar way to OpenMP in that programming directives are used for annotating the code and it's supported by the C, C++, and Fortran languages. There's also various functions for interfacing with OpenACC through its runtime library. OpenACC can run on CPUs but supports GPUs and other accelerator targets as well. As something many Phoronix readers have likely been following through our numerous articles on the matter, NVIDIA for the past year has been funding Mentor Graphics / Code Sourcery for adding OpenACC 2.0 with NVIDIA GPU (NVPTX) support to GCC and that work is coming along for GCC 5.