FPU Arithmetic Commands 

fadd   Fd,Fa,Fb
fadds  Fd,Fa,Fb
fsub   Fd,Fa,Fb
fsubs  Fd,Fa,Fb
fmul   Fd,Fa,Fb
fmuls  Fd,Fa,Fb
fdiv   Fd,Fa,Fb
fdivs  Fd,Fa,Fb
fsqrt  Fd,Fa
fsqrts Fd,Fa

These commands add, subtract, multiply, divide and calculate the square root of
FPU numbers. The "normal" one is always the Double Precision Command, the one with
the s at the end the Single Precision Command. The Substraction Commands do
Fd=Fa-Fb, the Division Commands do Fd=Fa/Fb. No remainder is calculated.

fres Fd,Fa

This command calculates a single precision estimate of the reciprocal of the
floating point operand in Fa. The estimate is correct to a precision of one part
in 256 of the reciprocal of Fa. According to Sam Jordan this command produces
an exact result, not an estimation, on the PPC 604e.

frsqrte Fd,Fa

This command calculates a double precision estimate of the reciprocal of the
square root of the floating point operand in Fa. The estimate is correct to a
precision of one part in 32 of the reciprocal of the square root of Fa. This
command usually needs only 1 (One !!!) CPU Cycle.

fsel Fd,Fa,Fc,Fb

Fa is compared to 0.  If Fa>=0, Fd is set to Fc. If it is <0 or not a number,
Fd is set to Fb. The comparision ignores the sign of 0. This commands is the
same like in C: Fd=Fa?Fc:Fb; Combined with an fsub, fsel can for example be
used for simple if-then-else, or to calculate the Minimum/Maximum.

Examples:

fsub f3,f2,f1
fsel f4,f3,f1,f2 ; f4=min(f1,f2)

fsub f3,f2,f1
fsel f4,f3,f2,f1 ; f4=max(f1,f2)

Note: These Examples give a wrong result, if one of f1,f2 is NaN (not a number).

With the exception of fsel all FPU Arithmetic Commands support the . Notation
like described in Important things. They don't support the o
Notation.