Heterogeneous Computing For Solving System Of The Linear Equations By The Conjugate Gradient Method
Heterogeneous Computing For Solving System Of The Linear Equations By The Conjugate Gradient Method
Author(s): Eduard BondarenkoSubject(s): ICT Information and Communications Technologies
Published by: Нов български университет
Keywords: Accelerators; GPU; OpenACC; OpenMP; CUBLAS; pinned memory
Summary/Abstract: The main purpose of this work is to show the advantages of using various approaches of heterogeneous programming. The results were received at the example of solving the system of the linear equations by the conjugate gradient method. High-level and low-level technologies (OpenACC and CUDA respectively) were used to accelerate computations on the GPU. The results of the work clearly reflect benefits of using the low-level technology CUDA. Several types of the heterogeneous computing were considered in this work. The main difference of each type is the amount of the data that is processed on the graphic accelerators and central processing units. The OpenMP technology was used to get a clearer comparative overview for the computations acceleration on the CPU. In this paper there was also shown another way to increase performance besides using GPU as acceleration unit.
Journal: Computer Science and Education in Computer Science
- Issue Year: 10/2014
- Issue No: 1
- Page Range: 363-372
- Page Count: 10
- Language: English