I have written a 24 bit mantisa, 8 bit exponent floating point package for the 8031 in about 1989. I still have the printed source code, however the actual assembly code is on a disk that works on a 640KB 5.25 disk readable on an old INTEL iPDS. The source code is minus most of the comments origionally entered because the 640KB disk-drive had corupted the diskette before I finished the project, and I had to re-enter all the assembler code quickly.

Please email me if you still have a need for this package. It can be extended for any length (i.e. 32 bit and on up). I used interpolated sine/cosine and tangent fixed 2D arrays to get to a .1 degree accuracy on trancendental math in order to increase the speed of calculating angles.