CC78K4 Datasheet, PDF (43/523 Pages) NEC – CC78K4 Ver.2.30 or Later, C Compiler Language

CHAPTER 2 CONSTRUCTS OF C LANGUAGE
(i) Floating-point number (float type) specifications
• Format
The floating-point number format is shown below.
(Higher address) s
e
m
31 30
23 22
The numerical values in this format are as follows.
(Lower address)
0
(Value of sign)
(Value of exponent)
(–1)
* (Value of mantissa) *2
s: Sign (1 bit)
0 for a positive number and 1 for a negative number.
e: Exponent (8 bits)
An exponent with a base of 2 is expressed as a 1-byte integer (expressed by two’s complement
in the case of a negative), and used after having a further bias of 7FH added. These
relationships are shown in Table 2-4 below.
Table 2-4. Exponent Relationships
Exponent (Hexadecimal)
FE
•
•
•
81
80
7F
7E
•
•
•
01
Value of Exponent
127
•
•
•
2
1
0
–1
•
•
•
–126
m: Mantissa (23 bits)
The mantissa is expressed as an absolute value, with bit positions 22 to 0 equivalent to the 1st to
23rd places of a binary number. Except for when the value of the floating point is 0, the value of
the exponent is always adjusted so that the mantissa is within the range of 1 to 2 (normalization).
The result is that the position of 1 (i.e. the value of 1) is always 1, and is thus represented by
omission in this format.
• Zero expression
When exponent = 0 and mantissa = 0, ±0 is expressed as follows.
(Value of sign)
(–1)
*0
User’s Manual U15556EJ1V0UM
43