is first normalized and then denormalized before it is stored in the designated floating-
point data register. This method simplifies the handling of all other data formats and types.
If an external operand is an extended precision unnormalized number, the number is
normalized before it is used in an arithmetic operation. If the external operand is an ex-
tended precision unnormalized zero (i.e., with a mantissa of all zeros), the number is
converted to an extended precision normalized zero before the specified operation is per-
formed. This normalization and conversion applies not only to external unnormalized
operands involved in arithmetic operations, but also applies to loading a floating-point
data register. Note that the regular use of unnormalized inputs defeats the purpose of the
IEEE standard and may produce gross inaccuracy in the results.
3.5.2 Conversions to Other Data Formats
Conversion from the extended precision data format to any of the other six data formats
occurs when the contents of an FPCP floating-point data register are stored to memory or
an MPU data register. Since no operation performed by the FPCP can create an unnor-
malized result, the result of moving a floating-point data register to an extended precision
external destination can never be an unnormalized number.
3.6 DATA FORMAT DETAILS
The following paragraphs provide the format specification details for the single (S), double
(D), extended (X) precision binary real, and packed decimal (P) real string data formats.
Refer to Tables 3-1 through 3-4 and Figure 3-11.
MC68881/MC68882 USER'S MANUAL FREESCALE
3-9
To Next Page
Loading...