Drop -mrecip completely.

[ktf]

-mrecip optimizes x/y to x * 1/y which can be convenient on some CPUs at the cost of slightly different results. This was introduced a few weeks ago by @VinInn due to alleged performance improvements that turned out to be not as large when running larger scale tests. Given it is not supported on clang we (@VinInn and myself) agreed to drop it for the moment. Notice also that probably -freciprocal-math provides the same level of optimization and it is also available on clang.

[cmsbuild]

A new Pull Request was created by @ktf (Giulio Eulisse) for CMSSW_7_2_X.

Drop -mrecip completely.

It involves the following packages:

RecoEgamma/EgammaPhotonAlgos
RecoVertex/PrimaryVertexProducer
TrackingTools/GsfTools
TrackingTools/GsfTracking
TrackingTools/KalmanUpdators
TrackingTools/TrackFitters
TrackingTools/TrajectoryState

@nclopezo, @cmsbuild, @Degano, @StoyanStoynev, @slava77 can you please review it and eventually sign? Thanks.
@battibass, @abbiendi, @GiacomoSguazzoni, @rovere, @VinInn, @bellan, @gpetruc, @cerati, @lgray, @trocino, @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.

[slava77]

Now I'm lost.
@VinInn Vincenzo, wouldn't we loose in CPU-s/evt here?

[ktf]

See my comment in #5103. Vincenzo agreed in a private discussion.

[StoyanStoynev]

@ktf Giulio please put some relevant description in the PR.

[cmsbuild]

+1
Tested at: e1e62cc
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-5107/735/summary.html

[ktf]

Done.

[cmsbuild]

Comparison is ready
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-5107/735/summary.html

[slava77]

+1

for #5107 e1e62cc
compared to CMSSW_7_2_X_2014-08-28-1400 as a baseline (test area sign419)
All changes are after all a removal of -mrecip, which is supposedly for the better, more precise.
The effect is visible in physics outputs at numerical precision level and doesn't seem to have any problematic change.

No visible effect on CPU.

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next CMSSW_7_2_X IBs unless changes (tests are also fine).