History of VecGeom validation in gcc10

[makortel]

As requested in the ORP meeting today, this issue is to collect the history of VecGeom validation in gcc10.

[makortel]

assign core, simulation, geometry, operations

[cmsbuild]

New categories assigned: geometry,core,operations,simulation

@fabiocos,@mdhildreth,@mdhildreth,@qliphy,@davidlange6,@rappoccio,@ianna,@Dr15Jones,@Dr15Jones,@smuzaffar,@perrotta,@makortel,@makortel,@bsunanda,@civanch,@civanch you have been requested to review this Pull request/Issue and eventually sign? Thanks

[cmsbuild]

A new Issue was created by @makortel Matti Kortelainen.

@Dr15Jones, @perrotta, @dpiparo, @rappoccio, @makortel, @smuzaffar, @qliphy can you please review it and eventually sign/assign? Thanks.

cms-bot commands are listed here

[makortel]

type documentation

[makortel]

The gcc10 vs gcc9 validation was done in 12_0_0_pre4. The issue about the observed GEM DQM differences is #35030.

VecGeom version in 12_0_0_pre4 was 1.1.16
https://github.com/cms-sw/cmsdist/blob/b81759451981b68eb03f3e425138ba649007ad32/vecgeom.spec#L1
The default optimization level was used at the time
https://github.com/cms-sw/cmsdist/blob/b81759451981b68eb03f3e425138ba649007ad32/vecgeom.spec#L20-L37
I suppose that means -O3 (since in cms-sw/cmsdist#7992 -O2 is requested explicitly).

VecGeom was updated to 1.1.17 in 12_1_0_pre4 with cms-sw/cmsdist#7352.

The production architecture was switched to gcc10 in 12_3_0_pre2.

[civanch]

PR #7992 has been merged. It is planned to perform private validation comparing CMSSW_12_5_X_2022-07-19-2300 (VecGeom -O2) versus CMSSW_12_5_X_2022-07-19-1100 (VecGeom -O3).

Backport to 12_4 #7997 is submitted.

[qliphy]

The above links should be: cms-sw/cmsdist#7992 and cms-sw/cmsdist#7997

[perrotta]

PR #7992 has been merged. It is planned to perform private validation comparing CMSSW_12_5_X_2022-07-19-2300 (VecGeom -O2) versus CMSSW_12_5_X_2022-07-19-1100 (VecGeom -O3).

See also cms-sw/cmsdist#7992 (comment)

[bsunanda]

I created big xml files using the O2 and O3 options. I ran then 1000 ttbar events using either of these options in CMSSW versions with O2 and O3 option. The conclusion is generation of the big xml file is not affected by the compiler otimization option but the results from the SIM results do. I attach a text file with mean differences of the ratio from 1 in 3 cases:
(1) geometry generated using options O2 or O3 and the SIM step done using O3
(2) geometry generated using options O2 or O3 and the SIM step done using O2
(3) geometry generation and SIM step done using options O2 or O3

Only third case the ratio differs from 1 in all the observables. So there is no need to recreate GT's