Fermat Number Transform Hardware Configuration,
Abstract
Based upon the theory presented in the earlier report, 'Exact Convolutions by Number-Theoretic Transforms', (T-55-703), of May, 1975, the present report describes recommended hardware configurations. The advantages from using such transforms to compute convolutions are as follows: There is no round-off error in the computations; Only N multiplications are required, where N is the block length; In complex convolutions it is possible to process the real and imaginary parts independently, which permits more parallelism and higher speed; It will be shown that complex multiplication can be performed with only two, rather than four, real multiplications.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 18, 1975
- Accession Number
- ADA013394
Entities
People
- Philip J. Erdelsky