Loading…
This event has ended. View the official site or create your own event → Check it out
This event has ended. Create your own
All dates, times and locations of tech program events or other scheduled plans are subject to change. Please check back regularly to ensure you view the most up-to-date version of the schedule.
View analytic
Wednesday, July 29 • 10:30am - 11:00am
Autotuning OpenACC Work Distribution via Direct Search

Sign up or log in to save this to your schedule and see who's attending!

OpenACC provides a high-productivity API for programming GPUs and similar accelerator devices. One of the last steps in tuning OpenACC programs is selecting values for the num_gangs and vector_length clauses, which control how a parallel workload is distributed to an accelerator's processing units. In this paper, we present an autotuner that can assist the programmer in selecting high-quality values for these parameters, and we evaluate the effectiveness of two direct search methods in finding solutions. We compared num_gangs and vector_length values found by our autotuner to those found by a bounded exhaustive search. On a suite of 36 OpenACC kernels, one or both of our autotuner's direct search methods identified values within the top 5% for 29 of the kernels; all results were within the top 30%. The majority of the kernels achieved a statistically significant speedup over the compiler's defaults, and the autotuner required only 7-11 runs of the target program, on average.


Wednesday July 29, 2015 10:30am - 11:00am
Majestic B

Attendees (5)