| | |
Summary: Numerical Error Minimizing Floating-Point to Fixed-Point
ANSI C Compilation
Tor Aamodt
Paul Chow
{aamodt,pc}@eecg.utoronto.ca
Dept. of Electrical and Computer Engineering
University of Toronto, Toronto, Ontario,
M5S 3G4, Canada
Abstract
This paper presents an ANSI C floating-point to
fixed-point conversion capability currently being inte-
grated within an application specific processor archi-
tecture/compiler co-development project at the Univer-
sity of Toronto. The conversion process utilizes profil-
ing data to capture the dynamic range of floating-point
variables and intermediate calculations to guide in the
generation of scaling operations. An algorithm for gen-
erating shift operations resulting in a minimization of
numerical error due to truncation, rounding and over-
flow is presented along with a novel DSP-ISA operation:
|
