FAST RADIX-2, RADIX-3, RADIX-4, AND RADIX-5 KERNELS FOR FAST FOURIER TRANSFORMATIONS ON COMPUTERS WITH OVERLAPPING MULTIPLY-ADD INSTRUCTIONS

Authors
GOEDECKER S
Citation
S. Goedecker, FAST RADIX-2, RADIX-3, RADIX-4, AND RADIX-5 KERNELS FOR FAST FOURIER TRANSFORMATIONS ON COMPUTERS WITH OVERLAPPING MULTIPLY-ADD INSTRUCTIONS, SIAM journal on scientific computing, 18(6), 1997, pp. 1605-1611
Citations number
4
Categorie Soggetti
Computer Sciences",Mathematics
Journal title
SIAM journal on scientific computing → ACNP
ISSN journal
10648275
Volume
18
Issue
6
Year of publication
1997
Pages
1605 - 1611
Database
ISI
SICI code
1064-8275(1997)18:6<1605:FRRRAR>2.0.ZU;2-B
Abstract
We present a new formulation of fast Fourier transformation (FFT) kern els for radix 2, 3, 4, and 5, which have a perfect balance of multipli es and adds. These kernels give higher performance on machines that ha ve a single multiply-add (mult-add) instruction. We demonstrate the su periority of this new kernel on IBM and SGI workstations.