[erlang-questions] dict, orddict and gb_tree
Thu Feb 22 00:09:11 CET 2007
Not really, the runs are pretty long and have to run them a number of
times anyway to even out gc. If you do it in a new process everytime
then doing the initial increase of heap will punish the results.
Anyway in this case I was not really interested in absolute timings only
in showing relative indicative times. For real and relevant measurements
you need an application.
Also it provides a bit of incentive to optimise the code.
Bjorn Gustavsson wrote:
> Robert Virding <> writes:
>>If any one can find some mistake in my code which lead to systematic
>>errors please tell me. Soon there will have to be more statistical
>>results as well. :-) Not!
> You must run each individual benchmark in a newly spawned process;
> otherwise garbage collection may punish some benchmarks more than other.
More information about the erlang-questions