new implementation
started with doubbas library and begin integration of new olson
single code.
so far done:
fcmp.asm
ftst.asm
fabs.asm
fneg.asm
fmul.asm	needs more work but it should function properly
alpha1 release Oct 17,1988
Corrected Autodoc, removed reference to more than 32bits of precision.
Changed the 881 mul/div code to use FSGNMUL,FSGNDIV . This should
make it a bit faster.

5/26/89
Decide the 68881 is flawed in single precision mode.

5/26/89
Fixed bug in floor calculation. It was not working at all.

5/27/89
Fixed rounding in fmul, fixed underflow in cxadj4.asm by
removing rounding of unnormalized numbers. May not be the
right thing todo.

5/28/89
Fixed rounding in fdiv

5/29/89
faddsub was trashing d6/d7, fixed.
made all branches into short branches that were possible
changed the rounding of unnormalized numbers to rund up
on 3/4. It was just chopping the bits.

Now passes Paranoia!
