Update CkfTrajectoryBuilder to handle hitless seeds

[cmsbuild]

Making the pull request after discussing with Tracking POG and giving the presentation during the Tracking meeting on the 17th November (as discussed and linked in the code comments).
Automatically ported from CMSSW_7_3_X #6566

[cmsbuild]

A new Pull Request was created by @cmsbuild for CMSSW_7_4_X.

Update CkfTrajectoryBuilder to handle hitless seeds

It involves the following packages:

RecoTracker/CkfPattern

@cmsbuild, @nclopezo, @StoyanStoynev, @slava77 can you please review it and eventually sign? Thanks.
@ghellwig, @makortel, @GiacomoSguazzoni, @rovere, @VinInn, @appeltel, @gpetruc, @cerati, @dgulhan, @venturia this is something you requested to watch as well.
You can sign-off by replying to this message having '+1' in the first line of your reply.
You can reject by replying to this message having '-1' in the first line of your reply.

[nclopezo]

+1
Latest commit marked as "Tests OK" in CMSSW_7_3_X

[cerati]

Does this mean you are closing #6566 (still being discussed)?

[StoyanStoynev]

Not yet supported by Tracking and/or Muon POG. The status is unclear to me.

[davidlange6]

Is this PR obsolete?

[StoyanStoynev]

It is not but action was requested for #6566 (same but in 73X).

[BenjaminRS]

Hi Marco (@rovere),

Sorry for the delay with the tests. I brought the OutsideInMuonSeeder online and tested it compared to the seeder I created. The problem is the timing. Testing on a WMnuNu sample with PU: using the normal configuration (3 hits on 3 layers) we can get 94% efficiency but it takes 4.7 ms for the L3 reconstruction. Using the fastest possible configuration of hits (1 hit on 1 layer) gives 63% efficiency in 2.2 ms. Using the hitless seeds I can get 92% efficiency in 2.1 ms. From our studies we wish to have hitless seeds available.

From our view we don't mind having a new class inherit from CkfTrajectoryBuilder and making the modifications there which would then come under Muon POG control. We just thought the code duplication could be minimised by making the modification directly in CkfTrajectoryBuilder.

Thanks,
Benjamin

[slava77]

-1

@BenjaminRS
From the last message it sounds like this PR can be closed.
#7003 seemed to be the target customer, but it apparently lived without it.
Let me know if I missed something.

[slava77]

ok, it sounds like the situation hasn't changed since about 6 weeks ago (also discussion in #6566).

If we are going to stay with muon-specific seeding (not using tools used in iterative tracking),
then as I understand tracking POG preference, this muon-specific code should be implemented in a muon-specific class/builder.

[VinInn]

Dear all, after discussion in the Tracking-POG it has been decided that this PR represents the less problematic solution to the issue.
Please reopen an re-sync this PR.
We will come back to an optimal solution in the context of a re-design of base tracking pattern-recognition that will include also requirements coming from phase1 and phase2

[slava77]

OK, I'm looking forward for an open/new PR with this feature.

[slava77]

@davidlange6
David, please reopen this.
It appears to still merge and compile fine.

[davidlange6]

Done.

[slava77]

@cmsbuild please test

the old test area is long gone

[cmsbuild]

The tests are being triggered in jenkins.

[cmsbuild]

+1
Tested at: 5cb6e62
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-6623/1995/summary.html

[cmsbuild]

Comparison is ready
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-6623/1995/summary.html

[slava77]

+1

for #6623 5cb6e62
code eventually OK'd
the jenkins tests pass (runs and no differences appear)

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next CMSSW_7_4_X IBs unless changes (tests are also fine). This pull request will be automatically merged.