The new software implementation described in this previous post has been compared with the code in the CVS repository, i.e.

$CVSROOT/StRoot/StEEmcPool/StEEmcIUPi0/

After the bug fix described in this post was implemented in the CVS version (and a number of bugs in the new software implmentation were fixed), the new software is found to have identical results as the old software when running on the (run 6) file

/star/institutions/anl/sgliske/data/Star/2006/st_physics_7136033_raw_1020001.MuDst.root

Some items that are specifically compared (event by event) include:

• The number of SMD clusters per sector and plane, along with their seed index position and total energy; (nClusters)
• The number of points per event, along with their associated U and V SMD clusters, their total energy, and the particular part of the point making algorithm that created the point; (nPoints)
• The final number of candidate pi0s after all cuts, including their mass and energy. (nFinal)

The cuts and parameters used are exactly those included in the files residing in CVS as of 8 Aug, 2011.  The statistics are

All descrepancies between the two methods are found to be based on the different placement of various cuts in the two implementations.  E.g., the CVS version does not apply the trigger cut until after pointmaking, while the newer implementation applies this cut before even making any clusters.  The final mass plots for both methods are following, with the plot from the CVS on the left and the plot from the new code on the right.

Furthermore, the new code was applied to the whole of run 7136033, with the resulting mass plot following.