The Performance Analysis of Fast DCT Algorithms onParallel Cluster Architecture
Atri Sanyal · International Journal of Information and Electronics Engineering · 2012
Clustered processor, having 16 clusters with 4 PE's in each cluster and each PE having two ALU, which can perform simultaneously, is taken. The PE's operate in a SIMD manner. The cluster and PE can transfer data from one another (1). 4 FDCT algorithms: - namely i>Vetterli ii> Arai's iii> Loeffler's iv> Chen's are taken into consideration. Our proposed 1D FDCT algorithm, parallel DCT is introduced. Examining all five algorithms using Timing Table it is concluded that percentage utilization is highest in Parallel DCT than other algorithms making it the most suitable to use in a parallel Cluster Architecture.