-
Notifications
You must be signed in to change notification settings - Fork 25
Decoder speedups #784
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Decoder speedups #784
Conversation
933faac
to
e1734b3
Compare
getting consistently ~40% faster on ifarm/cache, github actions, and lappi |
common-tools/clas-detector/src/main/java/org/jlab/detector/swaps/SwapTable.java
Outdated
Show resolved
Hide resolved
See #815, the difference comes from a CCDB conflict for run 809, with LTCC and helicity sharing crate/slot 19/19. |
As it turns out, LTCC winning over helicity is correct for 809 (now, with first match in the arbitrarily ordered list winning, instead of last as previously), and the helicity FADC reading in the KPP test were always bogus. Correct? |
And the extra event disappears after only the CCDB changes (CI jobs won't show it until they pickup the weekly CCDB sqlite snapshot). |
* add access by hash * use hash-based access in decoding * remove clas6 detectors * add docs * only if necessary * break loop upon first c/s/c-detector match * use table-name-only initializer * use DetectorType instead of String, preload tables once * remove clas6 detectors * change hash accessor names (workaround long->int... precedence) * 10k events per decoder test job * add a static for speed * speedup swap table initialization * use default (static) generator * more less hashing * speedup initialization * optimize swap table initialization * cleanup * more comments * check CCDB tanslation tables for crate/slot/channel conflicts * add more performant accessors and use them in decoding * remove cp, use cached
Adding a dedicated CI test first ...