Parameterized floating-point logarithm and exponential functions for FPGAs

Abstract : As FPGAs are increasingly being used for floating-point computing, the feasibilityof a library of floating-point elementary functions for FPGAs is discussed.An initial implementation of such a library contains parameterizedoperators for the logarithm and exponential functions. In single precision,those operators use a small fraction of the FPGA’s resources, have a smallerlatency than their software equivalent on a high-end processor, and provideabout ten times the throughput in pipelined version. Previous work had shownthat FPGAs could use massive parallelism to balance the poor performanceof their basic floating-point operators compared to the equivalent in processors.As this work shows, when evaluating an elementary function, the flexibilityof FPGAs provides much better performance than the processor withouteven resorting to parallelism. The presented library is freely available fromhttp://www.ens-lyon.fr/LIP/Arenaire/. L’utilisation des FPGA pour des calculs en virgule flottante étant de plusen plus courante, on peut dès lors envisager la réalisation d’une bibliothèquede fonctions élémentaires en virgule flottante pour FPGA.
Document type :
Reports
Complete list of metadatas

Cited literature [19 references]  Display  Hide  Download

https://hal-lara.archives-ouvertes.fr/hal-02102238
Contributor : Colette Orange <>
Submitted on : Wednesday, April 17, 2019 - 10:40:26 AM
Last modification on : Thursday, November 21, 2019 - 2:38:42 AM

File

RR2006-02.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02102238, version 1

Collections

Citation

Jérémie Detrey, Florent de Dinechin. Parameterized floating-point logarithm and exponential functions for FPGAs. [Research Report] LIP RR-2006-02, Laboratoire de l'informatique du parallélisme. 2006, 2+13p. ⟨hal-02102238⟩

Share

Metrics

Record views

21

Files downloads

46