Rtl Model Consistency

Prerequisites

• refc/build/ref_model built and executable.
• bfm/build/bfm_smoke built and executable.
• RTL files present under rtl/ with simulation infrastructure (iverilog or cocotb).
• Test vectors at sim/consistency/test_vectors.bin (or will be generated — minimum 10 vectors).

If any model binary is missing: WARNING — run that model's build skill first; proceed with 2-way comparison and note the absent model in the report.

Apply steps 1-5 to every vector set requested — do not stop after the first run.

Task(subagent_type="rtl-agent-team:bfm-dev", prompt="Run bfm/build/bfm_smoke on sim/consistency/test_vectors.bin via Bash CLI. " "Capture output as line-oriented hex text (one value per line) to " "sim/consistency/bfm_output.hex. Build first if needed: make -C bfm/.")

Task(subagent_type="rtl-agent-team:func-verifier", prompt="Simulate RTL with sim/consistency/test_vectors.bin as input via Bash CLI (iverilog/cocotb). " "RTL ports use i_/o_ prefixes, clocks are {domain}_clk, resets are {domain}_rst_n. " "Capture output as line-oriented hex text (one value per line) to " "sim/consistency/rtl_output.hex.")

</Tool_Usage>

<Examples>
<example index="1">
<scenario>All three models built; 50 shared vectors; post-BFM-update consistency gate.</scenario>
<expected_output>ref == BFM on 50/50; RTL diverges on vector 23; consistency_report.md diagnoses RTL as outlier; first divergence at index 142 with val_a/val_b values; diagnosis: CABAC encoder rounding difference.</expected_output>
</example>

<example index="2">
<scenario>Ref model updated after spec change; BFM and RTL not yet updated.</scenario>
<expected_output>ref != BFM == RTL; consistency_report.md diagnoses ref model diverged; mismatch details list first-divergence index and values per pair; user informed to update BFM and RTL.</expected_output>
</example>

<example index="3">
<scenario>BFM binary missing; only ref and RTL available.</scenario>
<expected_output>WARNING in report noting BFM absent; 2-way ref vs RTL comparison run; consistency_report.md documents reduced scope; user advised to run bfm-develop before full 3-way check.</expected_output>
</example>
</Examples>

<Escalation_And_Stop_Conditions>
- Three-way mismatch (`ref != BFM != RTL`) → report all divergences; cannot auto-diagnose; escalate to user.
- Any model binary not built → note in report; run 2-way comparison with available models.
- Test vector set missing → generate minimal 10-vector set before proceeding; note generation in report.
</Escalation_And_Stop_Conditions>

## Output

- `sim/consistency/consistency_report.md` — 3-way comparison matrix with pairwise MATCH/MISMATCH verdicts, first-divergence details, and diagnosis.

<Final_Checklist>
- [ ] All three models run on identical input vectors.
- [ ] Pairwise comparison completed for all three pairs (ref/BFM, ref/RTL, BFM/RTL).
- [ ] `sim/consistency/consistency_report.md` written with summary table and diagnosis.
- [ ] Diverging model identified where possible.
- [ ] First-divergence index and expected/actual values reported for each mismatching pair.
- [ ] Absent model binaries noted in report with recommendation to build.
</Final_Checklist>