Freescale Semiconductor MCF5480 - Manual (Page 195)

Operand Data Formats and Types
MCF548x Reference Manual, Rev. 3
Freescale Semiconductor 6-5
6.2.3.4 Not-A-Number
When created by the FPU, NANs represent the results of operations having no mathematical interpretation,
such as infinity divided by infinity. Operations using a NAN operand as an input return a NAN result.
User-created NANs can protect against uninitialized variables and arrays or can represent user-defined
data types. See Figure 6-6 .
Figure 6-6. Not-a-Number Format
If an input operand to an operation is a NAN, the result is an FPU-created default NAN. When the FPU
creates a NAN, the NAN always contains the same bit pattern in the fraction: all fraction bits are ones and
the sign bit is zero. When the user creates a NAN, any nonzero bit pattern can be stored in the fraction and
the sign bit.
6.2.3.5 Denormalized Numbers
Denormalized numbers represent real values near the underflow threshold. Denormalized numbers can be
positive or negative. For denormalized numbers in single- and double-precision, the implied integer bit is
a zero. See Figure 6-7 .
Figure 6-7. Denormalized Number Format
Traditionally, the detection of underflow causes floating-point number systems to perform a flush-to-zero.
The IEEE-754 standard implements gradual underflow: the result mantissa is shifted right (denormalized)
while the result exponent is incremented until reaching the minimum value. If all the mantissa bits of the
result are shifted off to the right during this denormalization, the result becomes zero.
Denormalized numbers are not supported directly in the hardware of this implementation but can be
handled in software if needed (software for the input denorm exception could be written to handle
denormalized input operands, and software for the underflow exception could create denormalized
numbers). If the input denorm exception is disabled, all denormalized numbers are treated as zeros.
Table 6-3 summarizes the data type specifications for byte, word, longword, single- and double-precision
data formats.
Table 6-3. Real Format Summary
Parameter Single-Precision Double-Precision
Data Format
Field Size in Bits
Sign (s) 1 1
Exponent = Maximum Fraction = Any nonzero bit pattern
Sign of Mantissa, 0 or 1
Exponent = 0 Fraction = Any nonzero bit pattern
Sign of Mantissa, 0 or 1
s e f
3130 23 22 0
s e f
63 62 52 51 0