`format`, `source_format`, and `destination_format` are exact scalar floating
types. `element_format` is the exact vector element type. Equal representation
width does not make arithmetic formats equivalent. Scalar and fixed-vector sign
manipulation use width because `f16` and `bf16` negate and absolute value are
bit-identical sign-bit transforms. Compare/minmax `numeric_format` is the
closed tagged sum `ExactFormat(canonical_type)` or
`RepresentationWidth(u32)`. It uses `ExactFormat` whenever a reachable
NaN-defined representative requires that exact format; it uses
`RepresentationWidth` only for an uncovered `nnan`-normalized behavior. A
mixed strict/`nnan` image therefore reuses its exact-format modes, while an
all-`nnan` 16-bit image can collapse `f16` and `bf16` to one width mode.
Fixed-vector shape and lane count remain admission facts and do not enter a
floating arithmetic key. Physical filtering must establish at least one
positive-lane actor witness before quotienting.
Independent review
No independent interface audit has been supplied for this attempt.
Execution provenance
The original pilot recorded the compiler hash and checkout revision separately; its binary build revision was not verified. Coverage follow-ups use separately identified rebuilt binaries. Original simulator binary hashes were not retained.
Original standalone-test agent report
This is exploratory evidence, not an accepted production-format PBT.
No reachable interface (CLI, pass, or IR surface) exposes the derived floating-arithmetic behavior quotient key that the passage constrains; the key exists only in private C++ library internals and fixed-case unit tests.
Reported numeric verdicts
not measured
Tested obligations
none recorded
Untested obligations
O1 format/source_format/destination_format are exact scalar floating types; element_format is exact vector element type
O2 equal representation width does not make arithmetic formats equivalent
O3 scalar and fixed-vector sign keys use active_representation_width (f16/bf16 collapse)
O4 compare/minmax numeric_format is the closed tagged sum ExactFormat(canonical_type) | RepresentationWidth(u32)
O5 ExactFormat whenever a reachable NaN-defined representative requires it; RepresentationWidth only for uncovered nnan-normalized behavior (mixed strict/nnan reuse; all-nnan 16-bit collapse)
O6 fixed-vector shape and lane count do not enter a floating arithmetic key
O7 physical filtering must establish at least one positive-lane actor witness before quotienting
Format
none (no PBT executed); ProGRMR+.spct and standalone Python were both ruled out for lack of an observable subject interface
Reproduction
not measured
Measured subject code coverage
blocked · No executable PBT was generated for this passage.
Coverage-preserving minimized cases: not measured
Full structured agent result
artifacts
README.md
INTENT.md
EVIDENCE.md
reproduce.sh
RESULT.json
barriers
Bash tool ssh requires dangerouslyDisableSandbox; a later probe was auto-denied with no approval surface, ending remote inspection for the session.
A C++ adapter against the private fabric::detail API is the only plausible route to a real PBT for this passage; it needs the MLIR/actor-projection build environment and exceeded the session budget.
checker path
not measured
controlled violation
attempted
false
reason
no property harness exists to violate; a synthetic mutation would only test a self-written restatement of the key algebra, which the task forbids as a replacement subject
coverage
attempted
false
reason
budget spent on interface search; no PBT to cover
format chosen
none (no PBT executed); ProGRMR+.spct and standalone Python were both ruled out for lack of an observable subject interface
generator path
not measured
interface evidence
behavior domain consumers
lib/Hardware/RTL/Providers/*.cpp only (C++ internal)
fixed case executable
loom-fabric-behavior-relation-finalization-test (test/fabric/behavior_relation_finalization.test) - hard-coded cases, cannot take generated inputs
no td or mlir occurrence
true
only fabric pass
--loom-elaborate-fabric-instances
private api
fabric::detail::resolveScalarFloatCompareBehaviorDomain / projectScalarFloatCompareBehavior in lib/Fabric/IR/ImplementationFamilyScalarFloatCompareBehavior.h (not under include/)
No reachable interface (CLI, pass, or IR surface) exposes the derived floating-arithmetic behavior quotient key that the passage constrains; the key exists only in private C++ library internals and fixed-case unit tests.