Re: [Boost-users] vector vs list speed test

thanks for the deque tip. i however need to remove from any index (eg remove the 3d element). i just used the remove head example in the code to get the worst case senario. yes, it is interesting that vector performs well when datatypes are small and there are few items (even when size() is 256). i will be using boost::shared_ptr<T>. so i'll have to do some tests with this datatype. so vector may be the best bet for the following requirements: - only insert to end - remove from any index - average size of 64 items - frequent iteration - datatype: boost::shared_ptr<T> i dont like the fact that vector resizes often when small. to get around this i may set the capacity to 64. memory shouldn't be an issue these days...agreed? i ran some tests on the impact of the default capacity. i however didnt find much difference in performance. On 7/13/06, Brian Budge wrote:

Another factor, which affects the constant Mikhail is talking about, is if the vector can fit into cache. This will make a huge difference. If you could allocate your nodes to be close together in memory (ie. use Cory's suggestion of boost::pool), you should be able to reap more cache coherency benefit in your list too.

On 7/12/06, Mikhail Kubzin wrote:

...

There is a simple rule for choosing a container. If the number of elements in it is not huge then use vector.

The reason is that although the complexity of for instance deleting an element from a vector is greater than one from a list, but O-differences (here linear-time vs. logarithmic-time) only matter if the containers are big enough to swamp the constant factor, which for containers of small objects like doubles frequently doesn't happen until after container sizes exceed several thousand elements.

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users