Mark Harris from Nvidia has posted a quick code example highlighting the power of OpenACC compiler directives for heterogeneous computing.
OpenACC is a promising new standard that I think will be a powerful tool for high-productivity acceleration of existing C and Fortran code. It will also be very valuable for prototyping; programmers can use it to determine if it is worthwhile to spend more time writing lower-level CUDA code for GPUs, for example. I hope you have found this overview interesting—please let me know what you think in the comments! I’ll revisit OpenACC in a future post, possibly to discuss some of the challenges you might face in applying OpenACC to more complicated codes, or to provide some guidelines on getting the best results. If you are interested in trying OpenACC you can download a free trial of PGI Accelerator to try it out.
Read the Full Story.