
22 Jul
2008
22 Jul
'08
3:31 a.m.
AMDG Joel de Guzman wrote:
David Abrahams wrote:
[snip]
Nifty! I implemented this in fusion. Pretty straightforward.
Um. You changed the result type calculation, but there are still O(n^2) function template instantiations. (at_impl<cons_tag>::apply<Sequence, N>::call<Cons, N2>) (Unfortunately, I have no idea what the relative cost of function template instantiations vs. class template instantiations is.) In Christ, Steven Watanabe