Author: Dan Strnad (rev. Jim Luther)
Year: 1990
... describes the format of Fixed and Frac data types and operations performed on numerical data types in the Integer Math Toolset.
Apple II
Technical Notes
_____________________________________________________________________________
Developer Technical Support
Apple IIGS
#79: Integer Math Data Types
Revised by: Jim Luther May 1990
Written by: Dan Strnad March 1990
This Technical Note describes the format of Fixed and Frac data types used by
the Integer Math tool set and operations performed on the Integer Math
numerical data types.
Revised since March 1990: Fixed original date, bit numbering of diagrams, and
a multiplication sign in the equation.
_____________________________________________________________________________
As stated in Volume 1 of the Apple IIgs Technical Reference, the Integer Math
tool set provides the following numerical data types:
Integers
Longints
Fixed
Frac
Extended
The precise format of the Fixed and Frac data types is not provided in the
reference manual, so this Note details these formats.
The format for the Fixed data type is stated in the manual as being a 32-bit
signed value with 16 bits of fraction. This means that the low-order 16 bits
of the Fixed format data value are considered as a fraction of 2^16, which is
the binary number represented by a one followed by 16 zeroes ($10000). In
other words, a Fixed value is the same as a long integer value whose binary
point has been moved to the left 16 places. In this representation, if the
low-order part of the Fixed format data value were $8000, the fractional value
would be equal to 1/2. A low-order part of $C000 would represent a fractional
part equal to 3/4. Therefore the highest value that a Fixed can contain is
32,767 and 65,535/65,536; the least value is equal to -32768.
31 30 29 18 17 16
__________ __________ __________ ___________ __________ __________ __________
| | | | | | | |
| -32768 | 16384 | 8192 | ... | 4 | 2 | 1 |
| | | | | | | |
|__________|__________|__________|___________|__________|__________|__________|
high-order word
15 14 13 2 1 0
__________ __________ __________ ___________ __________ __________ __________
| 1 | 1 | 1 | | 1 | 1 | 1 |
| - | - | - | ... | ----- | ----- | ----- |
| 2 | 4 | 8 | | 16384 | 32786 | 65536 |
|__________|__________|__________|___________|__________|__________|__________|
low-order word
Figure 1-Fixed Data Type
The format for the Frac data type is stated in the manual as being a 32-bit
signed value with 30 bits of fraction. This means that the low-order 30 bits
of the Frac format data value are considered as a fraction of 2^30, which is
the binary number represented by a one followed by 30 zeroes ($40000000). In
other words, a Frac value is the same as a long integer value whose binary
point has been moved to the left 30 places. The high-order 2 bits of the Frac
format data value are treated as follows. The high bit has a value of -2 and
the low bit has a value of 1. Therefore the highest value that a Frac can
contain is 1 and ((2^30)-1)/2^30; the least value is equal to -2.
31 30 29 18 17 16
__________ __________ __________ ___________ __________ __________ __________
| | | 1 | | 1 | 1 | 1 |
| -2 | 1 | - | ... | ---- | ---- | ----- |
| | | 2 | | 4096 | 8192 | 16384 |
|__________|__________|__________|___________|__________|__________|__________|
high-order word
15 14 13 2 1 0
__________ __________ __________ ___________ __________ __________ __________
| 1 | 1 | 1 | | 1 | 1 | 1 |
| ----- | ----- | ------ | ... |--------- |--------- |----------|
| 32768 | 65536 | 131072 | |268435456 |536870912 |1073741824|
|__________|__________|__________|___________|__________|__________|__________|
low-order word
Figure 2-Frac Data Type
Note that for Longints, Fixed, and Frac values, the hex representations of the
largest and smallest data values are $7FFFFFFF and $80000000, respectively.
A property of the Fixed and Frac data types is that two Fixed or two Frac
values may be added or subtracted just as if they were 32-bit integers. To
demonstrate this, imagine scaling the numbers by a given factor to make them
integers. After adding the numbers, the sum could be scaled back down by the
same factor. This follows from the distributive property of multiplication
over addition, which allows one to make the inference shown in the equations
which follow. In these equations, V1 and V2 are both either Fixed or Frac
values. The value for C being discussed, which illustrates the ability to
scale Fixed and Frac values, is 2^16 for Fixed values of V1 and V2, or 2^30
for Frac values of V1 and V2.
(C * V1) + (C * V2) C * (V1 + V2)
___________________ = _____________ = V1 + V2
C C
Similarly, two Fixed or two Frac values may be compared, as Longints are
compared, with one another. In general, the comparison, addition, and
subtraction operations used for long integers may also be performed on any
two Fixed or any two Frac values.
Further Reference
_____________________________________________________________________________
o Apple IIGS Technical Reference Manual
o Apple Numerics Manual, Second Edition