Re: (INCF Variable <float>) creates garbage!?

[kab (Kim Barrett)]

> I' have big loops incrementing floats. These loops create quite some
> garbage since incrementing a variable with a float apparently creates
> garbage (in MCL2.0p2):
> 
> (defvar V 0)
> 
> (time (incf V 1.5))
> 
> ...8 bytes of memory allocated.
> 
> Is there no way arround this? Intuitively I'd think INCF should be
> able to increment floats without producing any garbage (e.g., better
> boxing?).

I'd say 8 bytes is pretty good, since that's the size of the double float object 
it has to cons up as the new value.  incf is *not* a destructive operation, and 
does not bash the old value (who knows who else might be pointing at that 0.0, 
expecting it to remain zerop).