Hi Maxim, In the CVS version, I have improved the array reallocation mechanism for TGraph, TGraphErrors, TGraphAsymmErrors and TGraphBentErrors. I hope that you see an improvement. Let me know. Rene Brun Maxim Nikulin wrote: > > Hi, > > I'm writing a kind of slow control system. I use several histograms, two > dimensional plots and TGTextView widgets. They should be "ready to > display". If program receives entries one by one it works fine, but > reading a file recorded early takes significant time. > > Using the callgrind program I found several poor places. The first one > was TGTextView::AddLine() (Posted in > http://root.cern.ch/cgi-bin/print_hit_bold.pl/root/roottalk/roottalk04/2536.html > and fixed). Than my manipulations with TAxis::SetTimeDisplay() became > most important (old post, for curiosity only > http://root.cern.ch/phpBB2/viewtopic.php?t=992 ) and I found more cheap > workaround. > > For two dimensional plots I use the TGraph class. When the program > receives an entry it calls TGraph::SetPoint(). This leads to allocation, > copying and deallocation of x and y arrays. I don't use TGraph::Set() to > allocate large buffer due to (0, 0) points appear on my plots. > > I propose to introduce additional member and split point number and > array size for the TGraph suite. Are there any arguments against the > solution? I checked it with restricted descendant of TGraph, but in > general case it is more complicated than the TGTextView::AddLine() > issue, that's why I decide to discuss it firstly. > > -- > Max
This archive was generated by hypermail 2b29 : Sun Jan 02 2005 - 05:50:09 MET