Efficient Lisp functions -- using results of symbolic calculations numerically

Robert Dodier wrote:
> On 9/7/07, Raymond Toy (RT/EUS) <raymond.toy at ericsson.com> wrote:
> 
>>> #<FUNCTION $H ($X) (DECLARE (IN-DEFUN $H) (SPECIAL $X) (FLONUM $X))
>>>   (BLOCK $H
>>>    (PROGN NIL (+ (* 2.0 (COS (EXPT$ $X 2))) (- (* (EXPT$ $X -2) (SIN
>>> (EXPT$ $X 2)))))))>
> 
>> It would probably help a bit more if Lisp actually knew what a flonum
>> type was. :-)  When I compile this with cmucl, it complains that flonum
>> is an unknown type.  We probably need to add (deftype flonum ()
>> 'double-float) somewhere.
> 
> Heh. Who knows, we could even change the translator to
> emit (DECLARE (DOUBLE-FLOAT ...)) or whatever ...

Thought about that, but I decided to add the deftype and emit (declare 
(type double-float ...)).

Changes committed, and no warnings from cmucl when this example is compiled.

Is there a way to tell the translator to insert some optimization 
settings?  Just curious.

Ray