Parallel Performance of Fast Wavelet Transforms

Small-scale parallel implementation of fast wavelet transforms

We look at the use of parallel techniques for Fast Wavelet Transforms, focusing particularly on how to make best use of a very limited (and, perhaps, variable) number of available processors. Efficient algorithms for parallel FWT have been established in the case where the lengths of the vectors to be transformed and the number of processors are both powers of 2. We show that a näıve extension ...

Design of the Scienti c Subroutine Library for the Fujitsu VPP 300

A research agreement between the Australian National University and Fujitsu Japan has led to the development of a library of parallel mathematical subrou-tines and the extension of the library of single processor routines for the Fujitsu VPP300. The Fujitsu VPP300 provides a very sophisticated architecture combining vector processors in parallel by a crossbar switch. Very high performance can b...

Parallel Implementation of Fast Wavelet Transforms

Computing discrete transforms on the Cell Broadband Engine

Discrete transforms are of primary importance and fundamental kernels in many computationally intensive scientific applications. In this paper, we investigate the performance of two such algorithms; Fast Fourier Transform (FFT) and Discrete Wavelet Transform (DWT), on the Sony–Toshiba–IBM Cell Broadband Engine (Cell/B.E.), a heterogeneous multicore chip architected for intensive gaming applicat...

Vector-parallel Fast Wavelet Transforms

Parallel 2D fast (tensor) wavelet transform algorithms have been implemented using the transpose-split idea. They run successfully on the Fujitsu VPP 300 and will form part of the Scientiic Software Library of the VP and the VPP series. An alternative strategy based on parallelisation of 1D transforms yielding longer vector lengths is proposed and the merits of these approaches are compared. Fi...