ponder-lab / Hybridize-Functions-Refactoring / #2272

80%

main: 80%

Build Type

Pull #712

github

Committed by khatchad

Commit Message

Add an alwaysCheckTensorComputation option, mirroring the recursion flag

Completes the tensor-op check's parity with the recursion and side-effect
checks. `computeTensorComputation` is gated on a tensor-parameter candidate
(the only case it can affect a transformation decision); the new
`alwaysCheckTensorComputation` option overrides the gate to run it on every
function, for measuring barren-ness corpus-wide and hedging against a missed
tensor parameter, exactly as `alwaysCheckRecursion` does for recursion.

The forced value is now observable: `functions.csv` gains a per-function
`tensor computation` column alongside `recursive` and `side-effects`, and
`results.csv` records the flag among the experimental settings. The option is
plumbed as `EvaluationOption.ALWAYS_CHECK_TENSOR_COMPUTATION` (system property
`edu.cuny.hunter.hybridize.eval.alwaysCheckTensorComputation`) and the runner's
`CHECK_TENSOR_COMPUTATION` environment knob, set on the processor via a setter
like `targetedCfaDepth`. Off by default, so the precondition matrix is
unchanged.

Refs #709.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FuhM8SsTRB6BraLt3Ph7ZC

Pull Request Pull Request #712: Add benefit preconditions for barren functions

Coverage Stats

119 of 127 new or added lines in 5 files covered. (93.7%)

29 existing lines in 2 files now uncovered.

1754 of 2197 relevant lines covered (79.84%)

0.8 hits per line