Transforming Floating-Point Algorithms to Fixed-Point Implementations

Many digital signal processing and communication
algorithms are first simulated using floating-point
arithmetic and are later transformed into fixed-point
arithmetic to reduce implementation
complexity. This transformation process may take
most of the design time for complex designs and may
involve a long series of manual ad-hoc design
choices. This book provides methods to find optimum
word lengths efficiently, implement low-power
fixed-point arithmetic by word length reduction
techniques, and automate the transformation process
from floating-point to fixed-point arithmetic. The
automation step provides a design tradeoff curve of
signal quality vs. implementation complexity for the
system, which allows the designer to pick any
operating point on the tradeoff curve. The book
should help in developing fixed-point hardware or
software implementations from floating-point
representations, and should be especially useful to
professionals who are developing high-speed or
low-power hardware or software.

Autorentext

Kyungtae Han is a hardware engineer at Intel Labs. He received
his Ph.D. degree from The University of Texas at Austin in
electrical engineering.

Brian L. Evans is Professor of Electrical and Computer
Engineering at The University of Texas at Austin. He received
his Ph.D. degree from the Georgia Institute of Technology in
electrical engineering.

Klappentext

Many digital signal processing and communicationalgorithms are first simulated using floating-pointarithmetic and are later transformed into fixed-pointarithmetic to reduce implementationcomplexity. This transformation process may takemost of the design time for complex designs and mayinvolve a long series of manual ad-hoc designchoices. This book provides methods to find optimumword lengths efficiently, implement low-powerfixed-point arithmetic by word length reductiontechniques, and automate the transformation processfrom floating-point to fixed-point arithmetic. Theautomation step provides a design tradeoff curve ofsignal quality vs. implementation complexity for thesystem, which allows the designer to pick anyoperating point on the tradeoff curve. The bookshould help in developing fixed-point hardware orsoftware implementations from floating-pointrepresentations, and should be especially useful toprofessionals who are developing high-speed orlow-power hardware or software.