[Openmcl-devel] Floating point performance

[Paul Onions]

Hello,

I'm trying to understand how to optimize some floating point code  
under CCL, but I get wildly different timings from what I expect. For  
example:

Case 1: a reasonable outcome

(defun test-fun-1 ()
   (loop repeat 1000000
         for x = 1.0 then (+ x 0.01)
         sum (expt x 2)))

takes about 0.6 second on CMUCL and about 1.2 second on CCL, which is  
okay, but

Case 2: an _unreasonable_ outcome

(defun test-fun-2 ()
   (loop repeat 1000000
         for x = 1.0 then (+ x 0.01)
         sum (sqrt x)))

takes about 0.8 second on CMUCL but about 200 seconds(!) on CCL.

Can anybody enlighten me as to what is going on here?

BTW I'm running 1.5-dev-r13391M-trunk (DarwinPPC32)

Thanks,
Paul