AMD and Pathscale announced at SC14 that they have joined the OpenACC standards committee. OpenACC provides an efficient and performance-portable path for developing massively parallel programs across a wide range of accelerators, including GPUs, many core coprocessors and multi-core CPUs. OpenACC has been gaining traction for parallel programming. Such a move appears necessary as we have been told that caps-enterprise is no longer in business.
The recent paper, “NAS Parallel Benchmarks for GPGPUs using a Directive-based Programming Model” by Rengan Xu, Xiaonan Tian, Sunita Chandrasekaran, Yonghong Yan, and Barbara Chapman at the University of Houston Department of Computer Science discusses issues important to OpenACC programmers including array privatization, memory coalescing, cache optimization and examine their impact on the performance of the benchmarks. The authors note that right choice or combination of techniques/hints are crucial for compilers to generate highly efficient codes tuned to a particular type of accelerator.
Leave a Reply