On Mon, Feb 16, 2015 at 9:31 AM, Jonathan S. Shapiro <[email protected]> wrote: > On Sun, Feb 15, 2015 at 9:52 PM, Matt Oliveri <[email protected]> wrote: >> >> I understand what you're thinking with that, but it seems too >> Haskell-ish for BitC. To think that a function application that >> doesn't yield an instance of F is not really a function call probably >> would seem like complete bull crap to systems programmers. > > That's because it *is* complete bull crap. [Sorry - couldn't resist.] > > You either called a function or you didn't. Whether that function has > side-effects or not is completely orthogonal to whether you called it. > Either way I needed to know the arity to perform the call.
What I figured Keean was getting at was that function applications that do not involve deconstructing an F would have to be optimized out, statically converting them to the arguments for the actual function call when an F is deconstructed. This is why, in particular, applications that don't get all the way up to deconstructing before they escape, don't work. I'm not sure that's what he meant anymore. It might work, but it seems rather strange. _______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
