Prepare the Patatrack branch for merging into CMSSW

[fwyzard]

Track here any issues related to merging the Patatrack branch with the upstream CMSSW master branch.

[fwyzard]

Work to be integrated before the final upstream submission:

• revert forcing use of HLT conditions ? see RecoLocalTracker/SiPixelClusterizer/python/SiPixelClusterizer_cfi.py
• merge with CMSSW_10_5_0_pre1
  • merge Renumber GPU workflows #259 to fix the conflict in the runTheMatrix.py workflows
  • review and merge Next prototype of the framework integration #100 (closes Next prototype of the framework integration #119)
• review the host to device memory transfers
  • merge Use unified memory for conditions #157
  • merge Make transfer of BeamSpot asynchronous #245, superseded by Move BeamSpot transfer to GPU to its own producer #318
• review [Do not merge] Customize offline pixelTracks to the online config #160 and decide if it should be used
• restore lost changes to RiemannFit.h after the migration to 10.4.x (done by faster track fit and faster clustering #216)
• merge Add infrastructure around cub CachingDeviceAllocator, and use it in SiPixelRawToCluster #172
• review and merge faster track fit and faster clustering #216 (supersedes reduce duplicates, improve vertexing #189, Re-enable reconstruction of pixelVertices on the GPU #192 and Riemann fit patatrack #19 7, closes re-enable vertex reconstruction on the GPU #191)
• merge Fix MTV validation of initialStepPreSplitting tracks and add B-hadron MTV variation to pixel track validation sequence #199 (includes Add B-hadron MTV variation to pixel track validation sequence #193 and Fix MTV validation of initialStepPreSplitting tracks #198)
• review and merge Automatize the reuse of the CUDA stream of an input product #305
• migrate all HeterogeneousEDProducers to the new SwitchProducer approach (see Port producers to SwitchProducer  #229)
• settle on a good SoA abstraction (see Implement/Translate CPU version of the GPU-only algorithms  #221)

Do we need to include these files ?

• 27kB / 1.8k lines: EventFilter/SiPixelRawToDigi/test/RawId_moduleId.txt
• 3.7MB / 53k lines: EventFilter/SiPixelRawToDigi/test/Pixel_Phase1_Raw2Digi_GPU_Cabling_Map.dat
• 2.1MB / 6.3k lines: RecoPixelVertexing/PixelTriplets/test/pixHits.ipynb

Fix conflicts with CMSSW master branch:

• Fix conflict with Migrate tracker local reconstruction and pixel tracking to Tasks cms-sw/cmssw#25163 (done via Migrate tracker local reconstruction and pixel tracking to Tasks (backport #25163) #202)
• Synchronise with CMSSW_10_4_0_pre3
• Synchronise with CMSSW_10_4_0_pre4
• Synchronise with CMSSW_10_4_0
• Synchronise with CMSSW_10_5_0_pre1
• Synchronise with CMSSW_10_5_0_pre2
• Synchronise with CMSSW_10_5_0
• Synchronise with CMSSW_10_6_0_pre2
• Synchronise with CMSSW_10_6_0

RFC PR for master: cms-sw#25647 .

[fwyzard]

Remove obsolete files under EventFilter/SiPixelRawToDigi/test/ with #201 .

[makortel]

• what about Next prototype of the framework integration #100 and Next prototype of the framework integration #119 ?

I'd like to see #100 included and all modules migrated to it before merging to CMSSW master (after completing the steps in #100 (comment)). We can of course revisit this after all other steps have been finished if my plan takes long to execute.

[makortel]

@fwyzard I noticed cms-sw#25353 uses CMSSW_10_4_X_Patatrack as the source branch. Do we want the CMSSW PR to get updated every time we merge anything in the patatrack repository?

[fwyzard]

Do we want the CMSSW PR to get updated every time we merge anything in the patatrack repository?

For the moment I'd say we do.
Once we fix the conflicts, get some feedback, and the checklist above stabilises a bit, we can branch off something for integration.