Exploring GPU Memory Performance Using Digital Image Processing Algorithms

Puya Memarzia, Farshad Khunjush
Department of Computer Science and Engineering, Shiraz University, Shiraz, Iran
Indian Journal of Computer Science and Engineering (IJCSE), Vol. 5 No.6, 2014


   title={Exploring GPU Memory Performance Using Digital Image Processing Algorithms},

   author={Memarzia, Puya and Khunjush, Farshad},



Download Download (PDF)   View View   Source Source   



Leveraging the incredible parallel computational power of graphics processing units (GPUs) is a proven method for accelerating general applications. Efficient utilization of the GPU remains one of the greatest challenges facing programmers. The performance of GPU applications is extremely reliant on memory performance, to the point that it can be considered a critical bottleneck. This is further amplified when working with large amounts of data, which is common. In this paper, we explore several well-known data transfer and memory access methods. Our aim is to find out how they affect the performance of different applications. To do so, we first examine and specify the different techniques; then, we apply these techniques to a variety of digital image processing applications, which serve as the case study. The NVIDIA CUDA parallel programming framework serves as the foundation for our research. Our experimental results highlight the merits of each optimization method. We then use these results to categorize the benchmarks according to their behavior. We demonstrate significantly superior performance including speedups of up to 24x compared to naive implementations and up to 157x compared to serial implementations.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2020 hgpu.org

All rights belong to the respective authors

Contact us: