When you say MPLAB, do you mean assembly or C18? MPLAB is a design environment pkg, not a language, compiler, or assembler itself.
C18 has a floating point lib already available. As noted, fp is extremely cycle-intensive and rarely necessary. If it's essential, fine, but it rarely is. Even the fuzzy logic task you mention would seem better served by understand the task well enough to fit the numbers into 32 bit integers. These will be hundreds of times faster than fp.
If you're doing assembly, making your own 64 bit integer data type is possible if you modify your math routines. That'll still be much faster.