I would like to make one observation that I think is a bit striking:
If I can beleive David's results then the situation for the MINI torus is
rather peculiar:
You save 1/2 !!! the amount of running time turning this volume off all
together (NOGEOM.)
You "only" save 1/4 of the time not making SEC's in this volume.
Does this mean that GSIM spends 1/4 of it's time figuring out particle
tracking NOT RELATED TO SECONDARIES in the MINI ?
That seems very expensive and needs further investigation.
1) Could it be because NOGEOM had no MINI field ?
You can have a MINI torus field without MINI torus geometry.
The code should give a warning message about that. So this
effect is not due to the lack of a field in one vs. the other
test ( I hope.)
2) The geometry definition of the MINI was never cleaned up by Dale (who
implemented it) and was left with a rather high number of named volumes. He
should have used the GSPOSP feature. Doing this may improve the 'NOSEC'
result. The only way to prove that would be to do it. (Any volunteers ?)
This seems a cheap place to get some extra speed. The other place where the
change to the use of GSPOS from the exessive use of GSPOS could help us
would be in the FOIL geometry.
-- Maurik
>
> Hello Everyone,
>
> I have rerun a bunch of the speed tests using the first 100 events of
> Will's 2.4gev_p_tgt_1.evt data file, and his 2250_6000_1.ffread file. Some
> differences from past results are to be expected, since there are fewer
> showering particles in this data set, and the cutoffs for stopping to track
> particles have gone from a combination of 1 and 10 MeV to 100 keV. I don't
> think any conclusions significantly change, however.
>
> Default - everything on: 9.99s
> EC1 turned off: 9.24s
> EC turned off: 7.24s
> EC+EC1 turned off: 6.49s
> FOIL turned off: 9.00s
> MINI torus turned off: 5.05s
> DC turned off: 8.55s
> TORUs turned off: 10.22s
> CC+SC turned off: 7.83s
> ALL turned off: 0.11s
>
> NOSEC EC: 7.72s
> NOSEC EC1: 9.37s
> NOSEC DC+FOIL+CC+SC+TORU+MINI: 7.53s
> NOSEC ALL: 0.98s
>
> So no one component seems to dominate the time. The first thick object a
> particle encounters seems to start the shower process, but if that is removed,
> it starts in the next object. The time is clearly dominated by the tracking of
> secondaries (`conventional wisdom'), just not in a particular component.
>
> -David