The OpenACC standards group today announced several major milestones including the addition of new member, the National Supercomputing Center in Wuxi, the adoption of OpenACC by several major HPC applications, the addition of support for new target platforms and expanded implementations. Also, the organization announced several new training initiatives including a series of multi-day workshops, the 2017 ORNL hackathon schedule and a summary of 3rd party training opportunities worldwide.
The OpenACC specification simplifies the programming of massively parallel computing systems using compiler directives. It enables scientists and engineers to efficiently parallelize applications on high performance computing systems using accelerators, many-core processors, and multi-core CPUs, while maintaining code portability.
New Member and Application Momentum
National Supercomputing Center (NSCC) in Wuxi is the newest OpenACC member. Their TaihuLight Sunway system was ranked #1 in the June 2016 Top500 List of computer systems worldwide. NSCC has developed their own OpenACC implementation including several extensions suitable for the unique architecture of the Sunway system.
We were thrilled to follow the OpenACC specification to create a customized parallel compilation tool, Sunway OpenACC that supports OpenACC 2.0 syntax and targets home-grown SW26010 many-core processor with shared memory. The OpenACC paradigm was chosen for its better fit to our many-core processor, with a few extensions to better support the efficient utilization of the new hardware features, such as the Scratch Pad Memory for each core and DMA instructions,” said Dr. Haohuan Fu, Deputy Director of the National Supercomputing Center in Wuxi and Associate Professor Center for Earth System Science at Tsinghua University. “OpenACC has been used to parallelize and tune CAM-SE,that is 530,000 lines of code, as well as WRF and many other real-world applications for the #1 system on the Top 500.”
Several major HPC applications are reported to be releasing or working on GPU-enabled updates using OpenACC including Gaussian, one of the world’s most widely used quantum chemistry program, ANSYS Fluent, the popular commercial Computation Fluid Dynamics program, and five of the thirteen applications the Oak Ridge Leadership Facility has selected for their Center for Accelerated Applications Readiness (CAAR) program. CAAR is focused on redesigning, porting, and optimizing application codes for OLCF’s next generation Summit hybrid CPU–GPU architecture.
We build and support Fluent on a wide variety of parallel computing systems, and we need to be able to write a single version of our source code that runs efficiently on all of those systems.” said Sunil Sathe, Lead Software Developer. “With OpenACC, we were able to quickly enable a key solver for GPU acceleration while keeping the same code base for CPU execution. The OpenACC performance was excellent on NVIDIA GPUs and very competitive on CPUs. The ability to compile OpenACC based source code for parallel execution on a variety of CPU/GPU hardware platforms makes it a great choice for our future programming model.”
More Platforms and Compiler Options
In addition to OpenACC for the Sunway SW26010 processor, OpenACC support will be announced for OpenPOWER and OpenPOWER + GPU at SC16. OpenACC running on an OpenPower processor will also be demonstrated at SC16 in PGI Booth 2131.
OpenACC was introduced in GCC 5 in 2015, and improved this year in GCC 6 (https://gcc.gnu.org/gcc-6/changes.html). The latest GCC OpenACC implementation is available in the GCC gomp4 branch. OpenACC members will be demonstrating enhancements for GCC 7 at the OpenACC booth #634.
Hackathons Continue in 2017 Along with New Workshops, More Training Courses
Once again next year, the Oak Ridge National Laboratory in conjunction with other OpenACC members will be conducting a series of hackathons worldwide. Hackathons are intensive five day events where teams of experienced applications developers are paired up with OpenACC mentors in an effort to accelerate code development. The current schedule is online.
To educate less experienced developers, Oak Ridge and OpenACC are teaming up to offer a series of two and half day workshops next year. The workshops will include one day of OpenACC training plus a day and a half of actual hands-on work compiling and analyzing code to determine if it is suitable for acceleration and if so, where to begin.
A list of additional OpenACC trainings being conducted worldwide is available on the OpenACC website.
Additional OpenACC activities at SC16:
- OpenACC Birds of a Feather, Wed. Nov. 16th 5:30–7:00PM in room 155-C. Join us for a beer and lively discussion, such as “Should OpenACC and OpenMP ever merge”. Best paper award for WACCPD16, and exciting new OpenACC compiler options for developers.
- Free “Parallel Programming with OpenACC” books will be signed by author Rob Farber Monday, November 14 from 7:00 to 9:00 PM in the OpenACC booth #634
- Bringing About HPC Open-Standards World Peace, Nov. 16th, 10:30AM 255-B
Members will be available for questions in the OpenACC booth #634.
See our complete coverage of SC16, which takes place Nov. 13-18 in Salt Lake City.