Hi Walter,
You are right about the TEventList increment mechanism.
I have made the trivial mod in the Enter function such that one gets
a log behaviour and not a linear behaviour.
Rene Brun
Walter F.J. Mueller wrote:
>
> Dear Rooters,
>
> I was wondering why building an TEventList with a call like
>
> ntuple->Draw(">>el2id", "cp==2&&a2>0","",10000000,0);
>
> tends to take progressively more time when the size of the final
> eventlist gets larger. A short inspection of the code showed that
> the list is increased by a fixed number, given by fDelta and
> setable by SetDelta, with a rather small default of 100 !
>
> Unfortunately, such a constant increment resize policy leads always
> to a quadratic increase of the CPU time as a function of final size,
> no matter what the value of fDelta is !
>
> A much better strategy is to increase the size by a constant factor.
> This leads to an N ln(N) rather than a N^2 behaviour. The factor is
> not critical, just doubling is fine. Using the doubling strategy one
> gets after all on average a 75% filling, so the memory efficiency is
> adequate.
>
> I propose to change TEventList along those lines. If there are other
> flexible containers in ROOT with the same algorithmic problem I propose
> the same cure for them too.
>
> With best regards, Walter
>
> --
> Walter F.J. Mueller Mail: W.F.J.Mueller@gsi.de
> GSI, Abteilung KP3 Phone: +49-6159-71-2766
> D-64291 Darmstadt FAX: +49-6159-71-2989
> WWW: http://www-kp3.gsi.de/www/kp3/people/mueller.html
This archive was generated by hypermail 2b29 : Tue Jan 02 2001 - 11:50:31 MET