This is the 'MIT' format of the arithmetic support package. 'rt.a68'
contains the basic integer routines. The other modules belong to the
floating point library. The floating point library has been tested
but not certified by any means. Also, since it is 'hand-written' code,
some versions of the MIT assembler will assemble it incorrectly. If
you need to use it, you should generate a small test program first
which exercises all operations.

In particular, some versions of the MIT assembler misassemble
addx and subx instructions, by reversing their operands if they
are memory based.

The normal MIT library can be used in place of this; however the
compiler expects 'fcmp' to leave its arguments on the stack. One
version of the UniSoft distribution fcmp pops them.

This is a licensed package; you may distribute binaries freely,
but not the source.
