Hello:
I think that following today's discussion (I've attached slides.ps.gz)
it is clear we want to have an answer for runs with poorer S/N (which
requires "fixing" FF) and also get newton back in.
Following a discussion w/ Chi/Chris I wanted to suggest the following :
__ We need to log when cuts/checkpoints have been used during crunching.
In other words who does the job of reducing the number of bad tracks/segments
and which cuts are critical in terms of code efficiency. Right now crunching is
still resembling a little bit a black box: I'd not know how important is to have
fake stubs relative to single hits. I'd not know where we loose efficiency if we
change the cuts a little bit.
-- We only have hardwired cuts and they only abort events which are too
long to process. Would be nice to see more cuts f.i. on FF vertex, f.i.
throw away stubs which at 90 deg with a lazy track. Likewise would be nice
to collect these parameters in a resource file (for instance managed via TOpt)
-- This may allow to keep a rock solid version of recon (e.g. for online
mode), understand critical parameters and allow more people to understand
what is going on and get involved in benchmarking the code
-- keep in touch.
________________________________________________________________________________
Tancredi Botto, phone: +1-617-253-9204 mobile: +1-978-490-4124
research scientist MIT/Bates, 21 Manning Av Middleton MA, 01949
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:29 EST