Department of Computer Science at UH

University of Houston

Department of Computer Science

In Partial Fulfillment of the Requirements for the Degree of
Master of Science

Pooja Chitral

Will defend her thesis


Developing Accelerator Benchmark suite Using OpenACC Directives

Abstract

In recent years GPU computing has been very popular for the scientific applications, especially after the release of programming languages like CUDA, OpenCL and OpenACC. The growing popularity of GPU computation for commercial and scientific fields is attributed to high compute power of GPU cores. Accelerator benchmark suite using OpenACC2.0 is a combination of very popular benchmarks - Parboil and NAS Parallel benchmark. These benchmarks contain a wide range of throughput commuting applications which are useful for studying the performance of computing architecture and the compilers. Parboil benchmark includes applications from different scientific and commercial fields including image processing, bimolecular simulation and astronomy. NAS Parallel benchmark includes set of applications which target different areas of Computation Fluid dynamics.

Designed accelerator benchmark suite exploits the GPU architecture and the power of computation provided by its high computational environment using the new directives and clauses provided by OpenACC2.0. This benchmark can act as a reference point for the new programmer in GPU computing, reducing the time taken to understand one of the most powerful parallel programming paradigms.

Finally the goal of the accelerator benchmark is to serve the OpenACC community with some real time code to learn quickly and achieve the good performance for the scientific computing applications.

 

Date: Tuesday, December 2, 2014
Time: 9:00 AM
Place: PGH 550

Faculty, students, and the general public are invited.
Advisor: Prof. Barbara Chapman