|
- Easy to Use MATLAB Toolbox.
- Adds True Fixed-Point Arithmetic
to MATLAB.
- Permits Easy Conversion of Working
Double Precision MATLAB to Fixed-Point Arithmetic.
- Fixed-Point Precision Permits Direct
Mapping to Hardware DSPs.
- Bitwise Identical Results to Fixed-Point
Processors.
- Supports Modulo, and Saturation
Arithmetic; Signed and Unsigned Operands.
The Catalytic Compilers
FxP Toolbox is a MATLAB package which enables users to express
computations using fixed-point arithmetic rather than MATLAB's
default floating-point arithmetic. The FxP Toolbox executes fixed-point
operations precisely: the results are identical in precision and
accuracy to operations executed on typical fixed-point hardware
such as Digital Signal Processors (DSP). The toolbox supports
signed and unsigned operands, saturation and modulo arithmetic,
and rounding modes of round, truncate, floor, and ceiling.
The FxP Toolbox makes
it simple to convert computations. Because the toolbox uses the
class and operator-overloading features of MATLAB, converting
a working program into fixed-point arithmetic requires only the
insertion of a few constructors and assignment statements.
|
|
The FxP Toolbox provides a simple interface
to user-written transcendental and mathematical functions such
as sin, cos, exp, and sqrt. This interface allows your MATLAB
prototypes to use the same functions that will be used in the
final application, thereby guaranteeing identical results between
prototype and implementation as well as permitting early exploration
of transcendental behavior.
The advanced SP&C FxP Toolbox
extends fixed-point support to all functions in both the Signal
Processing and Communications MATLAB1 toolboxes. These toolboxes
can either be executed directly in fixed-point or in floating
point with the results converted to fixed-point.
The fragment in the graphic above
shows how simple it is to convert MATLAB programs into fixed-point
arithmetic using the FxP Toolbox. The highlighted statement is
the only change necessary to run the filter as signed 8-bit fixed-point
with 4 bits of precision to the right of the decimal. The variable
fxp_error holds the error injected in the computation by
the conversion to fixed-point.
|
|