Optimize hashtable allocator

[Paolo Carlini]

Hi,

On 10/30/2012 10:26 PM, Jonathan Wakely wrote:
>> According latest
>> messages I understood it was still time to break ABI for unordered
>> containers, at least I hope so.
> Paolo?
Yes, as you remarked, lately we fiddled quite a bit with these 
containers. We had too. Luckily, since they don't exist in C++98 mode, 
we don't the kind of annoying issue we have with std::list vs size and I 
don't think a quality implementation can renounce to the EBO for the 
allocator, since it benefits the vast majority of users most of the 
times. When we globally break the ABI we have to do the same for the 
associative containers too.

Thanks!
Paolo.