EIC code displayed by LXR master/​epic-lfhcal-tbana/​examples/​yall/​README.md	Source navigation Diff markup Identifier search General search
	Version: master test Revert   Change

Warning, /epic-lfhcal-tbana/examples/yall/README.md is written in an unsupported language. File is not indexed.

 # LFHCal Yall examples
 
 **New BNL user: start with [SETUP.md](SETUP.md).** It installs in your chosen
 software workspace, uses shared PWG raw input and your own scratch work area,
 and walks through small host/container tests, the three-task EIC Condor smoke
 test, and then `scan-set-1`. All interactive commands use your normal tcsh
 login/submit session. FullSet F2 is not the first test.
 
 ## Production workflows
 
 `scan-set-1` covers pedestal/muon pairs `296/298`, `299/300`, `303/304`,
 `307/308`, `309/310`, using `ToAOffsets_TBSPS2026_ParamScan_1.csv`.
 `scan-set-2` covers the later parameter scan, using
 `ToAOffsets_TBSPS2026_ParamScan_2.csv`.
 
 The repository has reproduction examples for all 14 published TB2026 FullSet
 B-G merged-muon calibrations:
 
 | Workflow | Pedestal | Muon runs in Fredi's merge order | Published set |
 | --- | ---: | --- | --- |
 | `fullset-b1-repro` | 071 | 072–084 | `FullSetB_1` |
 | `fullset-b2-repro` | 126 | 127–133 | `FullSetB_2` |
 | `fullset-c1-repro` | 137 | 147,148,149,146,145,144,141,142,143,140,139,138 | `FullSetC_1` |
 | `fullset-c2-repro` | 188 | 189,194,190,191,193,192 | `FullSetC_2` |
 | `fullset-c3-repro` | 278 | 289–293 | `FullSetC_3` |
 | `fullset-d1-repro` | 206 | 208–224 | `FullSetD_1` |
 | `fullset-d2-repro` | 264 | 266–270 | `FullSetD_2` |
 | `fullset-e1-repro` | 372 | 373–378 | `FullSetE_1` |
 | `fullset-e2-repro` | 420 | 421,422,423,425,424 | `FullSetE_2` |
 | `fullset-e3-repro` | 471 | 473,474,477,478,481,482 | `FullSetE_3` |
 | `fullset-f1-repro` | 431 | 426–430 | `FullSetF_1` |
 | `fullset-f2-repro` | 471 | 472,475,476,479,480,483 | `FullSetF_2` |
 | `fullset-g1-repro` | 485 | 484,486–491 | `FullSetG_1` |
 | `fullset-g2-repro` | 529 | 530–537 | `FullSetG_2` |
 
 The nine published HV-scan calibration tables are covered by
 `hvscan-repro`, which shares pedestal 188 across muon runs 194-202 and
 reproduces the 42-46 V scan in 0.5 V steps.
 
 Together with the 14 FullSet references, these workflows cover all 23 calibration text tables currently published under `calibrations/TB2026`.
 
 Each FullSet workflow has one typed `@table runs type run:` table. A single
 `convert-{type}-{run}` family converts every row. `merge-muon` binds
 `type=muon`; the `{type}-{run}` pedestal family binds only
 `type=pedestal` and inherits its run from the compatible parent.
 
 FullSet B-F use the V2 summing-board mapping and shared FullSetA-F bad-channel
 map. FullSet G uses V1 and the FullSetG bad-channel map. The examples preserve
 Fredi's ToA selections, including FullSetC_2 offsets for C3 and FullSetF
 offsets for E3.
 
 The B-E and HV-scan examples include `download_raw.tcsh`, which downloads exactly the
 pedestal and merged-muon inputs from the JLab XRootD source. All published
 FullSet results can be compared with:
 
 ```tcsh
 python3 "$LFHCAL_REPO/examples/yall/compare_fullset.py" --set-name FullSetC_2
 ```
 
 The comparison tool reports refinement-stage differences and assembles plot
 PDFs using `pdfunite`, `qpdf`, or Ghostscript.
 
 The production Yallfiles read `Run<run>.h2g` from `LFHCAL_DATA` and write
 under `LFHCAL_WORK/<example-name>`. Source `env.tcsh` from the example
 directory to load the bootstrapped installation and prepare the work area.
 
 ## Teaching examples
 
 `hgcroc-study` is a one-task local example. `calibration-pair` demonstrates two
 calibration tasks using pre-existing ROOT inputs. `lfhcal-simple` is the bounded
 local smoke test: it converts three pedestal/MIP pairs and extracts the three
 pedestals with a 1000-event limit. It deliberately stops before MIP scaling and
 waveform-summary work because the bounded sample is not intended to populate
 those fits and histograms reliably. The pair table is retained so the smoke test
 still exercises shared conversion and dependency expansion.
 
 The `calibration-pair`, `hgcroc-study`, and `hgcroc-study-condor` directories each
 contain a README describing the input file that must be supplied. They do not
 ship example ROOT data. The notes use TB2026 runs `296/298` as an example
 pedestal/MIP pair and run `298` as an example HGCROC study input; those run
 numbers are examples rather than hard requirements.
 
 These older local recipes have their own input assumptions and do not
 automatically acquire a container wrapper or switch to scratch output just by
 sourcing an environment file. Do not run them as heavy work on a login node.
 `hgcroc-study-condor` is a small batch example with a pre-existing converted
 input. For a fresh BNL installation, use the no-data EIC/Condor test in SETUP.md
 instead.
 
 ## Runner version and graph checks
 
 `lfhcal-simple` and the scan workflows require combined named-source `@each`
 support from yall-run PR #26 (commit
 `a690f2551edb21e099aa2adf4b2b077d12c6a787`). The FullSet workflows also
 require partial explicit binding and patterned-parent inheritance from yall-run
 PR #33, merged as commit
 `1081e9dd39418262588248272618130ce0503b8a`. The explicit account-provenance
 policy now raises the minimum to **yall-run 0.12.0a7**, including PR #45
 (merged as `5a39498d023b1c9edcaa08926c80f2bb848efda9`). Use `main` containing
 that change, not just a matching alpha version string. The Condor workflows
 also use the merged payload-wrapper and `%time` features.
 
 ```tcsh
 python3 "$LFHCAL_REPO/examples/yall/check_shared_conversions.py" -v
 ```
 
 This checks graph expansion without ROOT, raw data or scheduler submission.
 Conversions visit the ordered union of pedestal and muon columns once each;
 shared pedestal runs do not produce duplicate output owners. It also checks all 14 published FullSet task counts, typed conversion
 families, muon-only merge fan-in, inherited pedestal runs, and transfer
 dependencies that follow the run table. Reusing the same muon run with different pedestal choices still requires
 separate work areas or explicitly pair-specific output paths.
 
 ## Execution boundary
 
 Yall and Condor run on the host. Production scientific commands use the
 `tools/run-in-eic-shell.sh` adapter, which sends a quoted command through
 `eic-shell` standard input and preserves argument/exit-status behavior.
 The `.sh` suffix identifies its interpreter, not an interactive shell you
 need to enter. Do not submit Condor campaigns from inside the container.
 
 The scan workflows request one CPU and set `ROOT_MAX_THREADS=1` and
 `OMP_NUM_THREADS=1` inside the payload environment. The adapter is archived by
 Yall; its external EIC launcher and container image are not automatically
 archived or pinned. Use the same EIC environment to compile and execute LFHCal.
 A launcher pointing at `nightly` still selects a moving image.
 
 Software, inputs, outputs and campaign records must be visible on batch nodes,
 including through the relevant container binds. Host readability alone does
 not establish container readability; that is why the batch smoke test comes
 before an analysis.
 
 ## Runtime and existing campaigns
 
 `%time 2h` is a per-task request, including container startup, not a campaign
 runtime estimate. Condor renders it as `+MaxRuntime = 7200`. Enforcement depends
 on site policy; it is not the separate `allowed_execute_duration` setting.
 Do not assume BNL enforces CERN's runtime policy.
 
 Pulling code or editing a Yallfile does not update already-created campaigns or
 queued jobs. `resume` reuses frozen requests. Use new campaigns and fresh output
 roots for changed recipes or resource requests, and do not rebuild the shared
 LFHCal checkout while jobs are using it.
 
 Yall records campaign, task, executable, input/output, scheduler and attempt
 provenance separately; it does not inject LFHCal-specific provenance into ROOT
 files. Preserve campaign records together with important final products.
 
 ## Account provenance
 
 The larger examples explicitly set `%account-provenance off`, with a comment
 showing when to select `full`. This directive requires **yall-run 0.12.0a7 or
 newer**. Account recording also defaults to off when the directive is omitted.
 
 Use `full` to preserve creator, submitter, and host-worker account attribution
 when permitted by your site's privacy policy. A creation-time
 `--account-provenance full` or `--account-provenance off` overrides the recipe;
 the resolved choice is frozen for that campaign before any account lookup.
 
 This controls explicit OS-account snapshots only. Paths, command arguments,
 logs, scheduler records, and analysis outputs are not anonymized. Existing
 campaigns and their archived workers are not retroactively changed.

This page was automatically generated by the 2.3.7 LXR engine. The LXR team