Skip to content

SimpleFlashAlgo: allocate the per-bin PE spectrum only where there are hits - #987

Open
jzennamo wants to merge 1 commit into
developfrom
feature/jaz_flashfinder_sparse_pespec
Open

jzennamo wants to merge 1 commit into
developfrom
feature/jaz_flashfinder_sparse_pespec

Conversation

@jzennamo

Copy link
Copy Markdown
Contributor

SimpleFlashAlgo kept a static (time bins) x (optical detectors) table of doubles, 142 MB in the reco1 job, and zeroed all of it on every call. Rows are now allocated when the first OpHit lands in a time bin and released at the next call, and the integration loop skips empty rows.

Measured effect: reco1 peak memory down 0.15 GB.

Validation: Flashes unchanged.

Context: one item of a menu of time, memory, and disk savings for the Gen3 MC production. Each change is its own PR with one commit on develop, so they can be taken (or not) independently. Profiling and validation are in the slides "Gen3 MC Production: Where the Time and Disk Go" (Nov 2026). The validation is a fixed-seed comparison of the full gen-to-CAF chain against v10_26_00 (124 events, release run twice to measure its own run-to-run variation), plus each change applied alone on the release's own inputs and compared in the flat CAF (2895 branches).

🤖 Generated with Claude Code

…e hits

The static pespec_v table was (time bins) x (optical detectors) doubles, 142 MB in the reco1 job, and was zeroed in full for every flash finder call. Rows are now created when the first OpHit lands in a time bin and released at the next call; the integration loop skips empty rows. Flashes are unchanged.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant