property list hack, profiling, compiler optimization, looking at code

An additional observation regarding speed and  zerop1, which is defined 
this way:

  (defun zerop1 (x)
   (or (and (integerp x) (= 0 x))
       (and (floatp x) (= 0.0 x))
       (and ($bfloatp x) (= 0 (second x)))))

It could be "optimized" by adding a declaration this way:

  (defun zerop1 (x) (declare (optimize (speed 3)(safety 0)))
   (or (and (integerp x) (= 0 x))
       (and (floatp x) (= 0.0 x))
       (and ($bfloatp x) (= 0 (second x)))))

or the declaration could be placed on the whole file (zeropi is defined 
in source code file simp.lisp).

What changes if optimization (speed=3) is set?

1. The code is 221 bytes rather than 258 bytes long
2. The code no longer check for wrong number of arguments, stack 
overflow, or pending interrupts.
3.  (I suspect) the typical use, of testing an integer to see if it is 
zero or not, is reduced from
about 15 instructions to 10 instructions.  Your compiler/assembler/CPU/ 
may differ.

I don't know if the optimization settings of the different lisps is 
equivalent when people
publish "times to execute the test suite".  What is fair? Perhaps to set 
to the highest
optimization for which the test suite runs?   Perhaps to set to some 
optimization level
for which keyboard interrupts still work? Clearly a lisp that saves 
extra debugging info
is at a timing disadvantage in timing tests.

RJF