← all exploratory attempts · paired baseline · show in source

breadth-09: Switch schedule and port types

Agent report: complete Review pending

Session: completed

docs/spec-fabric-switch.md, lines 29–39

## Schedule predicate and port types The schedule predicate selects the port type kind of every input and output of the op. The two cases are mutually exclusive. | Schedule | Port type | Uniformity | |-------------|----------------------------|-----------------------------------------| | `spatial` | `!fabric.bits<W>` | All ports must share the same `W` (>= 0). | | `temporal` | `!fabric.bits_tag<W, T>` | All ports must share the same `(W, T)` (`W` >= 0, `T` >= 1). | Spatial ports may not use `bits_tag`; temporal ports may not use `bits`.

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 author explanation recorded.

Reported numeric verdicts

not measured

Tested obligations

• O1 spatial schedule selects !fabric.bits<W> for every input and output port
• O1 temporal schedule selects !fabric.bits_tag<W, T> for every input and output port
• O2 mutual exclusion: a single spatial port spelled bits_tag, or a single temporal port spelled bits, is rejected (mutation wrong_kind, 20/100 seeds)
• O3 spatial uniformity of W across all ports (mutation nonuniform_W, spatial, 8/100 seeds); W = 0 accepted as conforming (>= 0)
• O4 temporal uniformity of W and of T across all ports (mutations nonuniform_W/nonuniform_T temporal, 14/100 seeds)
• O4 T >= 1 (mutation tag_zero, 4/100 seeds; rejected by the fabric.bits_tag type constraint 'requires tagWidth > 0')
• Both op forms covered: named function_type template (generic MLIR) and anonymous SSA form (custom syntax inside fabric.module)

Untested obligations

• Anonymous 'source-type to destination-port-type' clauses: the destination type (not the source type) being the type subject to the uniformity/schedule checks was not exercised (no 'to' clause generated). This belongs to a later paragraph but interacts with this passage.
• Very large or unusual W/T values outside the sampled sets (W > 255, T > 8); no source upper bound exists, so this is a sampling limit only.
• Multi-port simultaneous mutations (only a single port is mutated per non-tag_zero case).
• Named-vs-anonymous port count sourcing, K*L bounds, connectivity_table, route tables, grant policies - other paragraphs, deliberately held legal as preconditions.

Format

Reproduction

Measured subject code coverage

measured · Measured handwritten Loom subject code only; external LLVM/MLIR and generated code excluded. Includes main executions, including rejected inputs; deliberate controls excluded. No matching native-suite baseline for this executable has been imported, so no coverage gain is claimed. Semantic fidelity remains a separate audit.

Execution: {'subject_exit_codes': {'0': 54, '1': 46}, 'subject_invocations': 100}
Baseline: No matching baseline measured
Minimized cases: not measured

Published evidence