Hi guys, Although I suppose to be a banker, I am stilll working with on-line DAQ applications based on ROOT :) So here are a few words about ROOT on-line performance. Well, my statement is: at present ROOT is not optimized for on-line data taking/displaying, although works reasonably well in such applications in general. An example: What we typically do in DAQ/on-line applications: - collect N events - update a set of 256x256 2D COLZ histograms in a canvas Easy. Practically it means several H->Fill(x, y, 1); calls before a graphical update. The point is that on the average you increase only several bins out of millions. Most likely these bins do not go above max of a histogram, so that COLZ scheme stays the same and only those several bins change colors (or not). So.. instead of updating only a few point in a canvas, ROOT recalculates all COLZ levels and redraws all histograms. Consequently instead of a few milliseconds you spend tens of seconds for one screen update. I understand that a smart on-line update requires a number of improvements, but this is somehow a must if you want to promote ROOT in the on-line world. The rule of on-line displays is : update only what has been changed, leave the rest untouched, don't waste time. Regards, Anton http://www.smartquant.com
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:50:36 MET