On Feb 11, 2013, at 1:34 PM, Camillo Bruni <camillobruni@gmail.com> wrote:
I can dig up all my benchmarks from my master ;) but I remember that we were faster on most operations or the same speed in the worst case. I have benchmarks for every possible case :P (that's a bit exaggerated), but there are 42 different micro benchmarks which stress most dictionary aspects.
For operations like iterating over the keys/values of a dictionary we are 5 times faster (less overhead due to fixed structures)
#includes on Set was also considerably faster in our implementation.
and of course degeneration due to the bad hashing should be much better handled⦠the current implementation, when it starts to have collision, it will make more collisions even more likely by using a free slot. And in case of collision, the buckets should be much faster to find something as with linear seating the entry can be much further away then just the number of colliding for one entry. Marcus