OpenACC 2.0 With GPU Support Coming To GCC

It looks like GCC will have the first open-source implementation of the OpenACC 2.0 standard and it will support GPU acceleration!

Last year I wrote how OpenACC hasn't been loved by open-source compilers. OpenACC is the parallel computing programming standard out of NVIDIA, PGI, and Cray for heterogeneous CPU/GPU systems. OpenACC is now up to version 2.0 but doesn't even have any mainline support in GCC or LLVM/Clang for the older versions of the parallel CPU/GPU programming specification. OpenACC itself is some directives added to C and Fortran.

The OpenACC news this year has been NVIDIA buying out PGI to work on OpenACC and then OpenACC support for GCC that hasn't yet been merged and was just converting OpenACC 1.1 (not 2.0) code into OpenCL code.

Earlier this month I wrote about Samsung engineers working on OpenACC support for GCC that was also targeting the older OpenACC 1.1 and its implementation was doing OpenACC over OpenMP. But now it looks like a real OpenACC implementation is finally en route for GCC.

Oak Ridge with its Titan super-computer boasting 18,688 NVIDIA GPUs is now partnering with Mentor Graphics (and NVIDIA) to work on adding OpenACC 2.0 support to GCC. The organizations wat OpenACC in GCC complete with GPU acceleration support so that this programming specification can be more widely adopted. "By integrating the OpenACC programming standard into the open-source GCC compiler suite, Mentor Graphics is providing the first open-source implementation of OpenACC 2.0. Because GCC is the default compiler on most Linux distributions and is readily available on other platforms... this implementation will greatly expand access to the language and facilitate the development and testing of OpenACC applications on smaller systems, such as workstations and clusters."

I haven't yet seen any of their OpenACC 2.0 code for GCC land on any public channels to know how they're implementing the GPU acceleration support, etc. However, this partnership information and announcement can be found at olcf.ornl.gov.

For those wondering about the timing of this OpenACC 2.0 GCC news and today's NVIDIA CUDA 6 release along with AMD putting out a new FirePro card, this is all news in the lead-up to next week's SC13 Supercomputing Conference in Denver, Colorado.