Floating point types in MCL...

[psz@lcs.mit.edu]

The abysmal performance of virtually all current Lisp implementations
on floating point is indeed awful.  As a hack, there IS Erann Gat's
"Floating Point Compiler for Macintosh Common Lisp, version 1.2a1"
which I got from the MACL contributed code archives at
cambridge.apple.com as fpc.lisp-v1.2a1.  It generates inline floating
point ops for composed arithmetic within a single functional
expression, but of course won't stop consing when you actually have to
store a value in a structure or return it as the value of some
procedure.  Its portability to non-680x0 machines is also nil, of
course, but it does seem to do what it claims.  My experience with it
is, so far, very limited, though.