Function:

FPREM calculates the modulus remainder of ST divided by ST(1) and stores
the result into ST. The procedure can also be seen as a repeated
subtraction of ST by ST(1).

There are several interesting things about this instruction:

The exponent magnitude difference should be no more than 63 or else the
instruction cannot reduce the ST properly in one execution. This means
you would have to execute the instruction several times to get a correct
result for large magnitude differences.
If this is the case, condition code bit C2 is set until the result in ST
is ok. Storing the Status Word and checking C2 should be done if the
condition could occur in your data set.

In addition to that, if the instruction is done, the least-significant
three bits of the quotient are stored in C3,C1 and C0.
If arguments to the tangent function are reduced by PI/4 the codes
represent one of the eight octants of a radius for which the tangent is
to be calculated.

FPREM does not operate according to the IEEE 754 standard, FPREM1
with opcode d9 f5 does, but is about 15-25 clocks slower than FPREM.

The bug appears on the 87 and 287 when 64 is performed with a>=1
and b==1 or 2. In that case the condition code bits represent an
incorrect value. There is no FP workaround. Test to prevent the situation.
Apparently this bug does not appear in the FPREM1 instruction.




FPTAN  Calculate tangent of ST
──────────────────────────────────────────────────────────────────────────────