>>>>> "Harvey" == Harvey J Stein <[EMAIL PROTECTED]> writes:
Harvey> For example, why were macros used to convert things like:
Harvey> (* a b)
Harvey> to
Harvey> (the 'double-float (* (the 'double-float a) (the 'double-float
b)))?
Hmm, even though I participated in that, I don't recall why this was
done. It's certainly not necessary in cmucl if a and b have been
declared to be double-float's.
[snip]
Harvey> I was surprised that the compiler doesn't figure out that s is a
fixnum
Harvey> as well, given that I start off by assigning it to zero, and only
add
Harvey> fixnums to it, but in retrospect, to do be completely correct, the
Harvey> compiler would have to sum the series to be able to safely deduce
Harvey> this, so of course it won't.
I think it could if the compiler saw the dotimes, but I think by the
time the type-inferencer sees it, it's been macroexpanded to something
else. I think cmucl does know that s is an unsigned-byte, which
doesn't help too much.
[snip]
Harvey> Why do I get no notes on the fixnum version, but get a cost 13 float
Harvey> to pointer coercion in the double-float version?
Because fixnum's don't ever need boxing, but the result of the
function needs to be boxed when returning a double-float.
Ray
