[v3] constexpr tuple

[Marc Glisse]

(removed gcc-patches)

On Thu, 8 Sep 2011, Marc Glisse wrote:

> On Thu, 8 Sep 2011, Christopher Jefferson wrote:
>
>> This might be totally insane, but I believe that:
>> 
>> tuple_cat(tuple_cat(A,B), C) always equivalent to tuple_cat(A,B,C);
>
> That's a fine way to find the return type, but for code, doesn't it generate 
> many copies? I think I'd forward_as_tuple and use some magic indices to 
> extract everything at once.

Or as an alternative, we could have f<3>(tuple0,tuple1,tuple2) call (with 
proper forwarding) f<2>(tuple1,tuple2,x01,x02) which calls 
f<1>(tuple2,x01,x02,x10,x11,x12,x13) which calls 
f<0>(x01,x02,x10,x11,x12,x13,x20) which is what the tuple constructor 
needs. This avoids the forward_as_tuple.

-- 
Marc Glisse