Systolic Optimization on GPU Platforms

ENRIQUE ALBA; PABLO VIDAL

LECTURE NOTES IN COMPUTER SCIENCE

SPRINGER

Año: 2012 vol. 6927 p. 375 - 383

0302-9743

The article presents a systolic algorithm implemented using NVIDIA?s Compute Unified Device Architecture (CUDA). The algorithm works as a general disposition of the elements in a mesh by sinchronously computing basic solutions among processing elements. We have used instances of the Subset Sum Problem for evaluating to study the behavior of the proposed model. The experimental results show that the approach is very efficient, especially for large problem instances and consumes shorter times compared to other algorithms like parallel Genetic Algorithms and Random Search.