|
Title:
|
MAPREDUCE AND OBJECT-ORIENTED PARADIGM APPLIED TO THE RADIO INTERFEROMETER SIGNAL CORRELATION IN GPU ENVIRONMENT |
|
Author(s):
|
Gustavo Poli, José Hiroki Saito, José Roberto Cecatto |
|
ISBN:
|
978-989-8533-14-2 |
|
Editors:
|
Hans Weghorn and Pedro Isaías |
|
Year:
|
2012 |
|
Edition:
|
Single |
|
Keywords:
|
GPU, CUDA, MapReduce, Signal Correlation, Object-Oriented, radio interferometer, antenna, BDA. |
|
Type:
|
Full Paper |
|
First Page:
|
179 |
|
Last Page:
|
186 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
The challenges to scientific computing are complex, with specific needs such as related to the response time of applications, complexity of the algorithms to be processed, as well as the volume of data. In this context, the interferometric array processing in the radio astronomy frequency band are great examples of these needs. Currently, there are radio telescopes under construction which when completed will produce a large daily volume of data that can take you some TB. In this work, it's presented the implementation of the MapReduce algorithm, traditional in the architecture WSC (Warehouse-Scale Computer), which has excellent performance in the execution of tasks on large volume of data, for the correlation processing of signals produced by an interferometric array (Brazilian Decimetric Array BDA) using with platform for running the GPU NVIDIA and an implementation using the Object-Oriented (OO) paradigm. The results obtained show the viability of its use, both for ease of implementation, and simplification of logic , because it is a natural parallel algorithm, and the OO paradigm allows the problem elements better preview, and the execution time for the processing of 64 samples per cycle was smaller than a second, to the BDA of 38 antennas using the GeForce 8800 GTX model of GPU. |
|
|
|
|
|
|
