Re: [ROOT] Multiple booking

From: Rene Brun (Rene.Brun@cern.ch)
Date: Thu Nov 29 2001 - 19:05:43 MET


Hi Robert,

a TH2F with 50000*15 channels will occupy about 34 MBytes
a TH2S with the same number of channels will occupy only 17 Mbytes.
May be this is enough for you (and may be even a TH2C)

50000 TH1F with 15 channels each will occupy more space (about 40 MBytes),
because each histogram has the overhead of the axis, titles, etc.

Rene Brun

Robert Zitoun wrote:
> 
>   Thank you Peter
> 
> Your idea is very good. I tried it and worked fine AND QUITE INSTANTLY!
> 
>   Still, Rene
> 
> I am running on a slow but "big" D0 machine (d0mino). Booking a 2D with
> 50000*15 channels should take as much page swapping as 50000
> 1D's. Shouldn't it ?
> 
>                          Robert
> 
> On Thu, 29 Nov 2001, Peter Skensved wrote:
> 
> >       Robert,
> >   Have you touught about using a singel TH2F instead and projecting out
> the `channel'
> > in question ?
> >
> >  Something like
> >
> > TH2F *h = new  TH2F( "h", "", 15, 0, 15, 50000, 0 )
> >
> > and
> >
> > h->ProjectionX("", 277,277 )->Draw()   to get `channel' 277
> >
> >                                                        peter
> >
> > ----
> >
> > Peter Skensved                          Email : peter@SNO.Phy.QueensU.CA
> > Dept. of Physics,                       Phone:  (613) 533-2676
> > Queen's University,                     Fax:    (613) 533-6813
> > Kingston, Ontario,
> > Canada
> >
> 
> > On Wed, Nov 28, 2001 at 04:25:10PM -0600, Robert Zitoun wrote:
> > >
> > >   Hi Rooters
> > >
> > >  I am trying to book 50,000 histograms in a macro doing
> > >
> > > TH1F *h[50000];
> > > char text[100];
> > >
> > >   for( int chan=0 ; chan<50000 ; chan++ )
> > >   {
> > >     sprintf(text, "h%d", chan);
> > >     h[chan] = new TH1F( text, "", 15, 0, 15);
> > >   }
> > >
> > > It take me 20 minutes to get histos booked. The booking time for one
> > > histogram increased with its ordinal number (quite fast in the beginnig,
> > > verrry slow for the last ones). I even used a compiled version to get it
> > > done in 20'.
> > >
> > >  There is certainly a more clever (faster) way to do it. Can you tell me ?
> > > Also can you tell why what it did is so long.
> > >
> > >                   Thank you
> > >
> > >                                  Robert Zitoun, Stony Brook and Annecy
> 
> >



This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:51:10 MET