C.1 OVERVIEW
ADSP-2100 family processors support 16-bit fixed-point data in hardware.
Special features in the computation units allow you to support other
formats in software. This appendix describes various aspects of the 16-bit
data format. It also describes how to implement a block floating-point
format in software.
C.2 UNSIGNED OR SIGNED: TWOS-COMPLEMENT FORMAT
Unsigned binary numbers may be thought of as positive, having nearly
twice the magnitude of a signed number of the same length. The least
significant words of multiple precision numbers are treated as unsigned
numbers.
Signed numbers supported by the ADSP-2100 family are in
twos-complement format. Signed-magnitude, ones-complement, BCD or
excess-n formats are not supported.
C.3 INTEGER OR FRACTIONAL
The ADSP-2100 family supports both fractional and integer data formats,
with the exception that the ADSP-2100 processor does not perform integer
multiplication. In an integer, the radix point is assumed to lie to the right
of the LSB, so that all magnitude bits have a weight of 1 or greater. This
format is shown in Figure C.1, which can be found on the following page.
Note that in twos-complement format, the sign bit has a negative weight.
CNumeric Formats
C – 1
To Next Page
Loading...
