Re: maximum number of fills in TH3 ?

By default TBrowser uses the default option when drawing histograms. You can change the default option (see box at the top level of the browser). The problem that you reported with your special histogram has been identified and fixed in the SVN trunk. If you download the source of the trunk you will have no problems in drawing your histogram with the default option.

Rene Brun

Alessio Mangiarotti wrote:
> Dear Rene,
>
> first let me thank you for you suggestion of using the "box" option.
> That solved my problem. I can now plot both the root files from the
> simple macro as I sent in the previous message, as well as the files
> from my Monte Carlo. So we can exclude any problem with the filling
> and writing of the histograms.
>
> I have been testing it a lot again and I still have the problem if I
> try to plot the histograms without any option. If I try to plot the
> histogram from my macro, even just with the TBrowser, root keeps
> running for hours without any output. In the case of the output from
> my Monte Carlo, just by opening the root file with the TBrowser and
> clicking on the histogram with the biggest number of entries, root
> runs for some minutes and then starts to use up all the memory in the
> machine (7-8 GB). Then I have to kill it. Since you can not reproduce
> it, it may be peculiar of my installation (compiler or flags, I use
> Scientific Linux 4.8 and compile with g++4). I am unsure on how to
> proceed. I am happy because I can work with the "box" option. What do
> you suggest?
>
> Thank you,
> Alessio.
>
>
> On Tue, 19 Jan 2010, Rene Brun wrote:
>
>> I cannot reproduce your problem with the memory.
>> Note that when drawing your histogram, you should not set the log
>> scale for x and y axis because xmin and ymin are negative.
>> Also do not draw with the default option (scatter-plot) as it may
>> take a long time with this very special case.
>> Simply replace your graphics lines by
>> TCanvas *c1 = new TCanvas("c1","hist",200,10,900,900);
>> h1->Draw("box");
>>
>>
>> Rene Brun
>>
>> Alessio Mangiarotti wrote:
>>> Dear rooters,
>>>
>>> I am experiencing a problem with the TH3 class. I have written a
>>> simple Monte-Carlo to investigate the PID capabilities of our
>>> detector. It takes few hours to process the simulated events, but
>>> because I do not need the full information, I prefer to generate a
>>> TH3 and not a tree. I am experiencing a memory leak with root when I
>>> try to open the file containing the generated TH3 histograms and
>>> plot them (essentially root tries to use all the available memory on
>>> the machine up to 7-8 GB).
>>>
>>> In the beginning I thought it was connected to the number of
>>> channels and so I reduced the dimensions to 100*100*100 channels.
>>> The problem remained.
>>>
>>> I noticed that everything works well if I use a reduced statistics
>>> (1/10). So I started to experiment with the TH3 class and I believe
>>> I found a problem when a channel is filled more than 2**31 times. I
>>> tried to use a TH1 and it works without problems. I attach a small
>>> simple macro I used to do my checks. I have used root 5.22 and 5.26,
>>> both have the problem. (Note that the file is written, root only
>>> stalls after).
>>>
>>> Is this a known limitation ? Can it be removed or avoided somehow?
>>> If not, the limit is in the number of fills for a single channel or
>>> for all the histogram (then I could add a check in my program)?
>>>
>>> Thank you for any assistance,
>>> Alessio Mangiarotti.
>>
>>
>
Received on Thu Jan 21 2010 - 08:21:58 CET