Fixed-Point Instructions - RCA 70/46 Reference Manual

FIXED-POINT
INSTRUCTIONS
INTRODUCTION
DATA FORMAT
Halfword Fixed-Point
Number
Full-word Fixed-Point
Number
REPRESENTATION
OF NUMBERS
• Using fixed-point instructions, binary arithmetic is performed on
operands used as addresses, index quantities, counts, and fixed-point data.
Generally, the operands involved are 32 bits long and signed. One of the
general registers always holds one operand. The other operand is in either
main memory or in a general register. Negative quantities are in the two's-
complement form.
This instruction set performs the following functions:
1. loading.
2. storing.
3. comparing.
4. shifting.
5. sign control.
6. radix conversion of fixed-point operands.
7. adding.
8. subtracting.
9. mUltiplying.
10. dividing.
The result of all sign control, compare, shift, add, and subtract opera-
tions is reflected in the condition code.
• A fixed-length format of a one-bit sign followed by the integer field
makes up fixed-point numbers. In one of the general registers, the number
is a 31-bit integer field. The complete 32-bit register is occupied by the
fixed-point quantity and sign. A 64-bit operand, with a 63-bit integer
field, is used by some shift, multiply, and divide instructions. A pair of
adjacent registers, addressed by the even address of the leftmost register,
contains these longer operands. The sign-bit of the rightmost register
becomes part of the integer field. The same register can be specified for
both operands in register-to-register operations (except for the Divide
instructions). In main memory, fixed-point operands are in either a 32-bit
word or a 16-bit halfword. The integer fields are then either 31 bits or 15
bits. Radix conversion operations always use a 64-bit decimal field. Integral
storage boundaries for these units of data must be observed. Halfword,
full-word, or double-word operands are addressed with one, two, or three
low-order address bits set to zero. Half-word operands are extended to full
words when they are fetched from main memory and used as a full-word
operand.
I
SIGN
15-hit
Integer
0 1 15
I
SIGN 31-hit
Integer
0
1 31
• All fixed-point operands are treated as signed integers. True binary
notation with a sign bit of zero is the representation of positive numbers.
Two's-complement notation with a sign bit of one is the representation of
negative numbers. To obtain the two's complement of a number, the value
of each bit is changed and a one is added to the low-order hit.
123