Revision of the PSimHit type storage, revert trackId to its basic meaning#49969
Conversation
|
cms-bot internal usage |
|
-code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-49969/47752 Code check has found code style and quality issues which could be resolved by applying following patch(s)
|
cmsbuild
added
code-checks-rejected
code-checks-pending
and removed
code-checks-pending
code-checks-rejected
labels
|
+code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-49969/47755 |
|
A new Pull Request was created by @fabiocos for master. It involves the following packages:
@Moanwar, @civanch, @cmsbuild, @ctarricone, @gabrielmscampos, @jfernan2, @kpedro88, @mandrenguyen, @mdhildreth, @nothingface0, @rseidita, @srimanob can you please review it and eventually sign? Thanks. cms-bot commands are listed here |
|
please test |
|
-1 Failed Tests: UnitTests RelVals RelVals-INPUT Failed Unit TestsI found 1 errors in the following unit tests: ---> test testPhase2PixelNtuple had ERRORS Failed RelVals
Expand to see more relval errors ...Failed RelVals-INPUT
|
|
There si apparently a problem in MixingModule, likely reading old PSimHit. The backward compatibility mechanism needs to be verified. |
|
The unit test failure seems to occasionally appear also in the regular IB, and it looks independent on this PR. Here the only differences seen are related to this PR, and they are understood. So, all the other differences appearing in the previous tests with the alternative ROOT6 build are due to the different version of that, without or with the fix by P. Canal. @cms-sw/orp-l2 @makortel @civanch @kpedro88 please suggest how to move forward with the integration. |
|
unhold |
|
-1 Failed Tests: UnitTests RelVals RelVals-INPUT
Failed Unit TestsI found 1 errors in the following unit tests: ---> test testPhase2PixelNtuple had ERRORS Failed RelVals
Failed RelVals-INPUT
|
|
Is the conclusion that with cms-sw/root#231 the read rule would work? (sorry, I lost track) |
|
I would say yes, the Randomly browsing plots here and there, differences are hardly visible, but when you get tens thousand differences it is quite difficult to make a solid statement. Let me relaunch the test in the "golden" working scenario |
|
please test with cms-sw/root#231 using full cmssw for CMSSW_17_0_ROOT6_X/el8_amd64_gcc13 |
|
-1 Failed Tests: UnitTests The following merge commits were also included on top of IB + this PR after doing git cms-merge-topic:
You can see more details here: Failed Unit TestsI found 2 errors in the following unit tests: ---> test TestFWCoreReflectionClassVersionUpdate had ERRORS ---> test TestIOPoolInputNoParentDictionary had ERRORS Comparison SummarySummary:
|
|
Would we then be ready for root-project/root#21669 to be merged in ROOT master (after their review completes) and a backport to 6.36?
That volume of "hopefully numerical differences" occur pretty much every time between ROOT major versions. |
|
@makortel on my side I believe that, with the fix, this PR seems to be doing what we want. The unit test failures should be addressed in some way, but they seem unrelated to this specific PR, in my understanding. |
Correct, those failures are related on "updating ROOT". Those tests by construction lead to such failures (and would be hard to improve). In cmsdist PRs that update ROOT we just ignore those failures. @pcanal @vepadulano In order to make progress here we'd need root-project/root#21669 merged in ROOT master and backported to 6.36. |
|
hold given that the ROOT fix needed to get this PR working (thanks a lot to @pcanal for his effort) is not going to be integrated into ROOT 6.36, I agreed with @makortel to hold its integration until CMSSW_20_0_X. At that point most likely I will remove the |
|
Pull request has been put on hold by @fabiocos |
|
@fabiocos , I am sorry, I had this concern in past but we my (wrong9 impression was that it is possible to converge in 17_0. If we postpone this PR, I would propose to make an intermediate step: PSimHits without fine bit structure but use only integer numbers. We need this in order to continue GPU development. It would be good to have some variant of hits, which is free from track ID overflow. |
…lasses_def.xml" This reverts commit a9fdfc5.
|
+code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-49969/49140
|
|
Pull request #49969 was updated. @Moanwar, @civanch, @cmsbuild, @ctarricone, @gabrielmscampos, @jfernan2, @kpedro88, @mandrenguyen, @mdhildreth, @nothingface0, @rseidita, @srimanob can you please check and sign again. |
|
please test |
|
+1 Size: This PR adds an extra 20KB to repository Comparison SummarySummary:
|
|
+1 |
1 similar comment
|
+1 |
|
unhold |
|
I do not see how the Tracker plots failures in wf 34434.0 could be related to this PR, I let other expert to comment in case. |
L1 track trigger has a non-reproducibility #47071 |
|
@makortel ok, thanks for confirming |
|
@cms-sw/dqm-l2 any comment? |
Sorry, I missed this one. lgtm |
|
+dqm |
|
This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @ftenchini, @sextonkennedy, @mandrenguyen (and backports should be raised in the release meeting by the corresponding L2) |
|
+1 |
11 participants
PR description:
Following #49732 and the discussion in the Simulation meeting https://indico.cern.ch/event/1634520/contributions/6878108/attachments/3199713/5699141/SIM_20260116.pdf , this PR proposes a revision of the mechanism used to store and propagate the hit type classification, so far used only by MTD for Phase2, but in principle applicable to any interested sub-detector.
trackIdis reverted to the pure Geant4 id, without any offset and limitation on the maximum available (beyond the uint32_t capability), so as not to interfere with exceptionally populated events, and software developments for GPU.The hit type is moved as a 7-bits subfield in the
processTypemember ofPSimHit, where the maximum code is at present 403, and this is a used-defined integer, not a pseudo-random variable. The interface is adapted, the assignment of hit type in MTD is adjusted accordingly, and all the dependencies in the code (truth accumulator, validation) are adapted.PR validation:
Tests on 100 single pions (wf 34506.0) are successful, the detailed Geant4 debug printout shows the desired behaviour, when activated, and the usual DQM histogram test shows perfect comparison between histograms.