Re: [Boost-users] Boost or Standard when there is the choice?

On 16 February 2014 19:34, Joaquin M Lopez Munoz wrote:

See http://tinyurl.com/okou28x for the actual implementation. You might want to try someting similar in Boost.Unordered.

I saw this in your code a little while ago and tried it, and it is faster than the old implementation, but it's still slower than the newer version, because of the way I find the bucket for a node. I cache the mixed hash function in a node, rather than the original hash function, so that the bucket for a node can be found with a single 'and' operation. This can be done several times when searching for a node, so even an optimised modulus is notably slower. Of course this doesn't apply to MultiIndex as you use a different data structure.