rtl-conformance-test

Verify that RTL-encoded bitstreams are bitexact matches against JM (H.264) or HM (H.265) reference decoder output. This is the final standards compliance gate before tape-out sign-off.

<Use_When>

• RTL passes functional verification and standards compliance must be confirmed
• Codec standard version has changed and re-conformance is required
• A specific compliance test suite (e.g., ITU-T JVT conformance) must be run </Use_When>

<Do_Not_Use_When>

• Reference model conformance_report.json does not exist (run ref-model first)
• Non-codec RTL design (not applicable) </Do_Not_Use_When>

<Why_This_Exists> Standards compliance cannot be inferred from internal verification. JM/HM produce the normative reference bitstreams. Bitexact match is the only accepted evidence of conformance for codec IP. </Why_This_Exists>

<Coding_Convention_Requirements> Conformance testbenches and simulation wrappers MUST follow project conventions (CLAUDE.md):

• Port connections: i_ prefix for inputs, o_ prefix for outputs (e.g., i_pixel_data, o_bitstream)
• Clock: clk (single domain) or {domain}_clk (multiple domains, e.g., sys_clk) — NOT clk_i
• Reset: rst_n (single domain) or {domain}_rst_n (multiple domains, e.g., sys_rst_n) — NOT rst_ni
• DUT instance: u_dut or u_ prefix
• Use skills/rtl-conformance-test/scripts/conformance_compare.py for bitexact comparison
• Use skills/rtl-conformance-test/templates/golden-metadata.json for stream metadata
• Use logic only (NOT reg/wire) </Coding_Convention_Requirements>

<Execution_Policy>

• vcodec-syntax-entropy-expert selects the applicable conformance test suite (bitstream-level compliance)
• video-processing-expert validates test vector applicability
• eda-runner runs RTL simulation via Bash CLI and captures output bitstream
• Comparison done via Bash CLI: cmp -l rtl_output.bin jm_output.bin
• Any mismatch is a hard FAIL (no tolerance) </Execution_Policy>

1. vcodec-syntax-entropy-expert selects conformance test vectors (JVT suite or project-specific) 2. video-processing-expert validates vector coverage against requirements 3. eda-runner runs RTL simulation on each vector via Bash CLI, captures encoded bitstream - Simulation uses correct `i_`/`o_` port naming and `sys_clk`/`sys_rst_n` 4. Compare RTL bitstream vs JM/HM output via Bash CLI: `cmp -l rtl_output.bin jm_output.bin` 5. Record PASS/FAIL per vector in sim/conformance/results.json 6. Any FAIL: capture divergence byte offset, attach to result entry

<Tool_Usage>

Task(subagent_type="rtl-agent-team:vcodec-syntax-entropy-expert",
     prompt="Select H.264 conformance test vectors applicable to CABAC encoder. List vector files and JM version to use.")

Task(subagent_type="rtl-agent-team:eda-runner",
     prompt="Run RTL conformance simulation via Bash CLI for each vector in sim/conformance/vectors/*.yuv. Compile and run: scripts/run_sim.sh --sim iverilog --top tb_cabac_conformance --outdir sim/conformance --trace rtl/cabac_encoder/cabac_encoder.sv sim/top/tb_cabac_conformance.sv. Compare output bitstreams with JM 19.0 reference: cmp -l sim/conformance/rtl_output.bin sim/conformance/ref/jm_output.bin. Report PASS/FAIL per vector.")

Task(subagent_type="rtl-agent-team:func-verifier",
     prompt="Validate sim/conformance/results.json: verify all vectors ran, all have status, any FAIL has byte offset and divergence details.")

</Tool_Usage>

500 conformance vectors; RTL simulation uses `sys_clk` and `i_pixel_data`/`o_bitstream` correctly; 498 PASS; 2 FAIL at byte offset 1024 (CABAC flush sequence); vcodec-syntax-entropy-expert identifies spec section 9.3.4.6; RTL fix applied; re-run all 500 PASS. Accepting 99% bitexact match — codec standards require 100% bitexact; partial match means non-conformant. Using `data_i` (suffix convention) in conformance testbench — violates project conventions (use `i_data` prefix). Bare `clk` is valid for single-domain designs.

<Escalation_And_Stop_Conditions>

• JM/HM not available → halt, instruct user to install (provide download URL)

• 10 vectors fail → likely systemic issue; escalate to ref-model for model review before RTL debug

• Single vector fails repeatedly after fix attempts → report to relevant sub-domain expert (vcodec-syntax-entropy, vcodec-prediction, vcodec-transform-quant, or vcodec-filter-recon) for spec interpretation
• Testbench naming convention violation → must fix before running conformance </Escalation_And_Stop_Conditions>

<Final_Checklist>

• All conformance testbenches use correct naming (i_/o_ prefix, {domain}_clk/{domain}_rst_n)
• All ITU-T conformance vectors run
• 100% bitexact match achieved
• sim/conformance/results.json written with per-vector status
• JM/HM version recorded in results </Final_Checklist>

Run both encoder conformance (RTL encodes, JM decodes) and decoder conformance (JM encodes, RTL decodes) if design includes both paths. Conformance vectors can be parallelized: each vector runs in independent simulation via Bash CLI.

Block-level conformance for decoders: When the design is a video codec decoder, read domain-packages/video-codec/knowledge/block-level-conformance.md for the cross-phase verification chain. In addition to end-to-end bitexact comparison, verify that each RTL processing block (CABAC, inverse TQ, prediction, reconstruction, deblocking, SAO) produces output matching the C reference model at block boundaries. This catches bugs localized to specific pipeline stages that end-to-end comparison alone may mask through error cancellation.