Hi

I have known for a while, that the new IST geometry may have some design issues with the many volumes that holds the structures.
Finally I got around to make some actual timing measurements.
I made a couple of simulations throwing 100 pions in eta (-1,1) and (-2,2) in 3 configurations
Y2013, y2014 and y2014 removing IST.
The results show that the IST setup must be revisited. It can be used for small simulation production, but for major production this is too much
increase in computing time. It is essentailly cause by navigating between IBAM and IBMO volumes. The times are cpu seconds measured by kumac $CPTIME.

Time for 100 pions with momenta between 0.2 and 2 GeV. As seen from the screen shot below the steps becomes very tiny, and in fact this event was aborted because of more than 10,000 steps.

I tried one kind of solution to remove the IBMO mothervolume for IST, such that IBAM would be placed diretly in IDSM, but that made the searching even worse jumping between IDSM and IBAM in 100 micron steps increasing the time from 23.8 seconds to 51 seconds.

The test were done with the macro's in ~videbaks/test  setup.kumac
>exec setup -> y2013
>exec setup#14  y2014
>exec setup#pion ; setup phasespace
>exec setup#ntrig 100   (loop over 100 vents and report cpu time

I do not have any simple obvious solution. The IBAM volume hold the ladders, associated boards, and support structures in a BOX volume. Due to the mounting structures, boards the BOX is fairly large causing large overlap between the IBAM for neighbouring ladders. This is where the significant amount of time is spend in the tracking just going 100 micron at a time in air.. The figure below outlines the borders of the IBAM volumes.

It should be pointed out that if one could use the features of an assembly, as present in TGeo, there might not be this kinds of issues.

cut of same near the supports.

Summary: Feb 18
Jason followed up on this; the behaviour is indeed due to the content knowledge that Geant has.
Jason point out that the slow simuator in particulr for TPC adds another factor of 10 over the praticle genaration. This additional
time form IST is probably not enough of a concern to warrent highpriority work