2015-03-15 19:58 GMT+03:00 Cosmin Boaca
Hello,
I have runned benchmarks on the original implementation. It works slightly better even than the implementation based on std::map. Insertion time on benchmarks it's about -0.4 seconds better and I can't figure out why. That implementation updates 4 additional pointers and the current implementation does only one copy for updating the value. Also, node size is smaller. I don't understand how the original implementation can be more cache friendly.
Something really strange is happening. Give me the revision number of the original version and the revision numbers of the updated version. I'll do measures on different compilers, different systems and different optimization flags. -- Best regards, Antony Polukhin