A Parallel Approach to Cosine Calculation Using OpenCL

Faris Cakaric; Azra Ibrahimovic; Ajdin Kahrovic

Fast calculation of high-precision cosine function becomes increasingly important in many areas of computer science. This paper proposes an algorithm for high-performance calculation of cosine function using Maclaurin series and by exploiting benefits of parallelisation. It presents several parallel implementations of this algorithm in OpenCL framework, improving them from naïve to the optimised implementation. The paper shows comparison of time of execution when the algorithm is executed sequentially on CPU and in parallel on GPU, confirming enormous decrease in time of execution when algorithm executes in parallel. Finally, the paper draws conclusions about the scalability of algorithm and percentage of total time of execution wasted on communication overheads.

Select Volume / Issues:

• Volume 6 Issue 2, Mar. 2015

Year:

2015

Type of Publication:

Article

Keywords:

Cosine Computation; Maclaurin Series; Opencl; Parallelization; Communication Overheads

Journal:

IJECCE

Volume:

6

Number:

2

Pages:

221-226

Month:

March

ISSN:

2278-4209