Intel i86W Manual page 97

FLOATING-POINT INSTRUCTIONS
6.3.3 Floating-Point Reciprocals
frcp.p fsrc2, fdest (Floating-Point Reciprocal)
fdest
~
1 / fsrc2 with absolute mantissa error
< 2-
7
frsqr.p fsrc2, fdest (Floating-Point Reciprocal Square Root)
fdest
~
1 /
~
with absolute mantissa error
< 2-
7
The
frep
and
frsqr
instructions are intended to be used with algorithtns such as the
Newton-Raphson approximation to compute divide and square root. Assemblers and
compilers must encode fsrcl as
fO.
A Newton-Raphson approximation may produce a
result that is different from the IEEE standard in the two least significant bits of the
mantissa. A library routine supplied by Intel may be used to calculate the correct IEEE-
standard rounded result.
Traps
The instructions
frep
and
frsqr
cause the source-exception trap if fsrc2 is zero. An
frsqr
causes the source-exception trap if fsrc2 < O.
6.4 ADDER INSTRUCTIONS
The adder unit of the floating-point section provides floating-point addition, subtraction,
and comparison, as well as conversion from floating-point to integer formats.
6-9