#30 Timing problem using V01-09-02

[A.Calcaterra]

Hi Wouter,

the macro attached executes in ~25 s under ROOT3 and RooFit
V00-01-02, compared to ~550 s under ROOT4 and RooFit
V01-09-02,
that already includes a time-execution addressing fix.
See:

http://babar-hn.slac.stanford.edu:5090/HyperNews/get/roothelp/907.html

This is keeping us (Inclusive D*D* people in the IHBD
AWG) from
moving on to ROOT4. The whole fit machinery is enormously
complicated, but I have finally been able to isolate
the problem
in a short macro to be posted here.

I hope you'll be able to see where the bug is, either
in RFC/RFM,
or in the way we use it. My test directories are:

~acal/ana23 for ROOT3
~acal/ana24 for ROOT4

Thank you for any help/hints! Sandro

[A.Calcaterra]

[Wouter Verkerke]

Logged In: YES
user_id=427309

Hi Sandro,

This sounds like a relatively straightforward problem. I'll look
into it.

Wouter

PS: I am away till the 23rd, so it will take another week...

[A.Calcaterra]

Logged In: YES
user_id=1225653

Hi Wouter,

Although I can not understand why, it appears that the
problem was
(at least part of it) in this mistake:

virtual Int_t getCoefAnalyticalIntegral(RooArgSet& allVars,
RooArgSet& analVars, const char* rangeName=0) const ;

the last argument (added inadvertently during the transition
from
RooConvolutedPdf to RooAbsAnaConv) had as a side effect the
blowing up of exec times.

SLAC is down for the count, later on I will test better and
post again,
but it seems that this problem was cured. Thanks anyhow! Sandro