Function:

FPTAN calculates the ratio between x and y in the following formula:

   x
   -  = TAN(original ST)
   y

The y result replaces the original argument in ST and x is then pushed
onto the stack. On pre-287xl FPUs, the values for y and x may be anything,
the ratio however is correct. On 287xl+ FPUs, x is always 1.
ST(1) represents the fractional value itself there.
To generate the same set of results on all FPUs, the FPTAN should be
followed by FDIV and FLD1. Note that this reproduces the original
results on the 287xl+.

Note that ST(7) must be free or an invalid operation exception may occur
because x is pushed onto the stack.

The 486 bug occurs when a specific set of code is executed with a specific
set of data. There is no way you can anticipate this and the workaround
should always be implemented if code will run on a 486/487.
The bug corrupts the FPU stack without signalling it to either FPU or CPU.
Data corruption is usually the result.
Workaround: FPTAN should always be followed by: FCLEX, FINIT, FLDCW, FSTSW,
FSTSWAX, <FSAVE> or <FSTENV> or by a WAIT and a non-FPU instruction.
Do note that some of these FPU instructions contain bugs themselves.




FRSTOR  Restore FPU state saved to memory by FSAVE
──────────────────────────────────────────────────────────────────────────────