Unify treatment of sector search for UT-based pattern reco

• Put all code that is used to go from track-position -> sector numbers in one location.
• Clean up not used code
• Adapt pattern recognition algorithms (Rec!2404 (merged))

Addresses Rec#147

@peilian

changed the description

added 1 commit

• 9662671c - some clean up, simplifications and protection

Compare with previous version

added 1 commit

• 2b37e7c4 - add text to assert

Compare with previous version

mentioned in merge request Rec!2404 (merged)

changed the description

added 1 commit

• 890a534f - split namespace

Compare with previous version

added 1 commit

• 1b885092 - Added 0 padding to the sectors such that algorithms don't need to deal with it

Compare with previous version

added 1 commit

• 7ea1d62a - formatting

Compare with previous version

• [2021-04-20 21:08] Validation started with lhcb-master-mr#2225
• [2021-04-21 23:03] Validation started with lhcb-master-mr#2232
• [2021-04-22 12:16] Validation started with lhcb-master-mr#2234
• [2021-04-22 23:13] Validation started with lhcb-master-mr#2237
• [2021-04-23 11:46] Validation started with lhcb-master-mr#2239
• [2021-04-26 09:56] Validation started with lhcb-master-mr#2244
• [2021-04-26 19:54] Automatic merge failed in [lhcb-master-mr#2250](https://lhcb-nightlies.web.cern.ch/nightly/lhcb-master-mr/2250/LHCb/checkout
• [2021-04-27 10:05] Validation started with lhcb-master-mr#2251
• [2021-04-28 14:42] Validation started with lhcb-master-mr#2259
• [2021-04-29 14:56] Validation started with lhcb-master-mr#2264

added 1 commit

• a1b2f030 - add forgotten file: adds space for padding element

Compare with previous version

added 1 commit

• e46db455 - fix wrong comparison in assert

Compare with previous version

added 1 commit

• 81c5a6a6 - fix assert and make it actually meaningful

Compare with previous version

added 1 commit

• 4e2063dd - format da file

Compare with previous version

marked the checklist item Clean up not used code as completed

marked the checklist item Adapt pattern recognition algorithms (Rec!2404 (merged)) as completed

added 1 commit

• a053c676 - make clang happy

Compare with previous version

/ci-test --merge Rec!2404 (merged)

added ci-test-triggered label

added hlt1-throughput-increased label

added 34 commits

• 335c872c - Get zero suppresison parameters from DB
• 5db8f034 - Fixed formatting
• b11663ab - Added useParamsFromDB property and keep possibility of configuring zSup threshold
• 9ece8874 - Fixed formatting
• 4139736e - add element wise division in MatVec
• e3363d49 - Adding new matrix operations
• 59f051ce - subtraction between matrices
• 5482a71c - add similarity for Mat
• 3be56169 - converting matsym to mat using cast_to_mat
• a75144fc - added tests for similarity and elementwise division
• f216104c - Fixed formatting
• 013341a8 - adapt to range-v3 0.11
• 511513fd - unify UT decoders + strong enum UTDAQ::version
• 1fc422c6 - streamline range and iterator implementations
• e1916878 - Add integer XOR, NOT, elementwise popcount and to_array to SIMDWrapper
• 23029b20 - Add readme file, with explanations to SOACollection
• 894e4056 - prefer addBank over createBank/adoptBank
• a6f7e02d - Configurable ANNSvc and expose PR descriptor names
• 5c40b6af - remove unsafe Warning(...) counter to avoid race condition in BankKiller
• 09bf6fe8 - Extend SIMDWrappers
• 702dd29d - pass SIMD register types by value instead of const&
• 912995b8 - prefer forwarding constructors which funnel into one constructor
• 08a5e2f3 - Include inputs of held tools.
• 832c227c - Cache expensive, data-flow-traversing functions.
• e8f06bea - Small fixes for DD4hep support of alignment
• 2a59d508 - Workaround dependency issues in DD4hep
• 67f8cc80 - Revert "Do not trigger updateManagerSvc when setting DeltaMatirx"
• 74cfc883 - add constraint to template arguments of {stable,}partitionPosition
• 1e52cdc4 - Add a Generic Track SOA Container
• a5b61f32 - Randomise default File Catalog name
• d4e7062a - Update CONTRIBUTING.md to update supported platforms
• 96c7e9f9 - Add [[nodiscard]] constexpr
• 965ae924 - reduce redundancy, factorize common code
• a3b35508 - remove default type for RawBank::range

Compare with previous version

added 63 commits

• a3b35508...aa73ae9c - 62 commits from branch master
• 1e4d365c - fix merge conflicts

Compare with previous version

added 1 commit

• d4a46ef2 - trying to manually revert a mess

Compare with previous version

added 1 commit

• 8d40328f - Fix uniquifying function. slight increase in efficiency

Compare with previous version

added 1 commit

• a8899a71 - change the way the sector numbers and the padding element is treated, add more asserts

Compare with previous version

marked this merge request as ready

resolved all threads

assigned to @rmatev

added Tracking label

added UT bug fix labels

let's give this another spin tonight

added lhcb-head label

mentioned in merge request !3052 (closed)

added 1 commit

• 5c8bd20c - add some more global UT numbers to UTInfo

Compare with previous version

• [2021-05-01 00:22] Validation started with lhcb-head#2932

assigned to @jonrob and unassigned @rmatev

• [2021-05-03 00:07] Validation started with lhcb-head#2933
• [2021-05-04 00:07] Validation started with lhcb-head#2934
• [2021-05-04 00:47] Validation started with lhcb-head#2934
• [2021-05-05 00:09] Validation started with lhcb-head#2935
• [2021-05-05 01:15] Validation started with lhcb-head#2935

@graven @decianm Can you please check the Moore/Brunel failures at

https://lhcb-nightlies.web.cern.ch/nightly/lhcb-head/2935/

thanks ?

closed

reopened

removed lhcb-head label

resolved all threads

merged

mentioned in commit 791814d3

unassigned @jonrob

is there a MR in rec that is supposed to be merged with this? if not, there is a discrepancy. https://gitlab.cern.ch/lhcb/Rec/-/blob/master/Pr/PrAlgorithms/src/PrLongLivedTracking.cpp#L392 still refers to LHCb::UTDAQ::findSectorsFullChanIdx while this function has been renamed with this merge. @jonrob @decianm

Ah, I may have just merged the wrong LHCb MR.... I merged this instead of !3050 (merged). mea culpa...

So I believe this MR requires Rec!2404 (merged) and Rec!2427 (merged), and I have started a build with these in https://lhcb-nightlies.web.cern.ch/nightly/lhcb-head/2938/

Thanks.

However, looking at Rec!2427 (merged), I only see one file updated, and not the references that will actually change with Rec!2404 (merged). Is this correct?

Hi @decianm, sorry maybe I did a mistake in the ref update. Let me have a look

@ascarabo Please look at the diffs in the lhcb-head build above.

@jonrob I've just updated my setup to current master of everything and Rec does not compile anymore. After investigation, it's because this MR is needed ! So something has been merged that was depending on this, while this one is still pending. Can you fix that ?

For the details, the problem is the rename of findSectorsFullChanIdx into findSectorsFullID

Yes, this MR and Rec!2404 (merged) belong together. Without Rec!2404 (merged), you'll get a compilation error in PrLongLivedTracking, as you mentioned.

Ok, but that's not my problem. My problem is commit 9df96b06. It was merged into LHCb via MR !2990 (merged) and drops findSectorsFullChanIdx although it's still used in Rec. This MR in which I'm writing fixes the problem, but is not yet merged.

Not that the commit in question has message "fixed formatting" but looks more like the content of the present MR. So I think it has never been tested actually and was merged by mistake !

To add to the situation, this MR can no more be merged/rebased on master as it conflicts, exactly with commit 9df96b06. I think that one is really wrong and the review process/testing did not manage to find it out. Strange. I suspect there were short cuts in the tests here.

The message is 'fixed formatting' because the author of this MR (or someone else) selected the squash option, and I did not notice when I merged and hence gitlab picked the last inappropriate comment....

and yes I accidentally merged a wrong LHCB MR yesterday, see my comment above.

ok thanks for the explanation. Due to the rebase conflict, I'm working with a hand patched version for now.

Totally normal that these things can happen! But I'd like to plead with the maintainer team (@jonrob) that maybe in the future the first step would be a revert of such MRs, to get the stack back into working order quickly.

Fair enough request. I was hoping to avoid having a revert commit in our git history by getting the refs quickly, but that did not happen yesterday.

mentioned in issue Moore#273 (closed)