babyworm/rtl-agent-team/rtl-p5s-integration-test-policy

Tier 4 Integration Test Policy

Coding Convention Requirements

Integration testbenches MUST follow project conventions:

• Port connections: i_ prefix for inputs, o_ prefix for outputs
• Clock: clk (single domain) or {domain}_clk (multiple, e.g., sys_clk) — NOT clk_i
• Reset: rst_n (single domain) or {domain}_rst_n (multiple, e.g., sys_rst_n) — NOT rst_ni
• DUT instance: u_dut for top-level, u_ prefix for sub-modules
• Use logic only (NOT reg/wire)
• cocotb signals: dut.sys_clk, dut.i_*, dut.o_*

Integration Test Ordering (mandatory sequence)

1. Static connectivity checks first (fast, catches wiring bugs)
2. Simple data flow (single transaction end-to-end)
3. Sustained throughput (back-to-back transactions)
4. Backpressure and stall scenarios
5. Error injection and recovery

Result JSON Schema

{
  "top_module": "{top}",
  "tier": 4,
  "connectivity": {
    "reset_propagation": "PASS",
    "clock_connectivity": "PASS",
    "port_width_match": "PASS"
  },
  "data_flow": [
    {"scenario": "basic_forward", "status": "PASS", "req_ids": ["REQ-U-001"]},
    {"scenario": "max_throughput", "status": "PASS", "req_ids": ["REQ-U-007", "REQ-P-003"]}
  ],
  "handshake": [
    {"test": "backpressure_propagation", "status": "PASS", "req_ids": ["REQ-U-012"]},
    {"test": "pipeline_flush", "status": "PASS", "req_ids": ["REQ-U-015"]}
  ],
  "e2e_reference": "PASS",
  "verdict": "PASS"
}

Simulation Commands

SV TB:

scripts/run_sim.sh --sim verilator --top tb_{top}_integration \
  --filelist rtl/filelist_top.f --outdir sim/top --trace \
  sim/top/tb_{top}_integration.sv

cocotb:

make -C sim/top sim SIM=verilator TOPLEVEL={top} MODULE=test_{top}_integration

Fallback (iverilog):

scripts/run_sim.sh --sim iverilog --top tb_{top}_integration \
  --filelist rtl/filelist_top.f --outdir sim/top --trace \
  sim/top/tb_{top}_integration.sv

Multi-Clock Domain Integration

• Verify CDC synchronizers are connected correctly
• Test data transfer across clock domains
• Use rtl-p5s-cdc-verify for detailed CDC analysis

Acceptance Criteria Traceability (Optional at Tier 4)

Integration test results may optionally include ac_ids for AC-level traceability.

When structured acceptance_criteria (with ac_id fields) exist on a requirement in iron-requirements.json:

• Integration tests SHOULD tag ac_ids where feasible alongside existing req_ids
• This is advisory at Tier 4 (not required for gate pass/fail)
• Example: {"scenario": "basic_forward", "status": "PASS", "req_ids": ["REQ-U-001"], "ac_ids": ["REQ-U-001.AC-1"]}

When no structured AC exists (absent or plain string-array format): use req_ids only (existing behavior).

Tier Transition Rules

• Tier 3 PASS or PARTIAL_PASS (rtl-p5s-func-verify) → Tier 4 eligible
• Tier 4 PASS (this skill) → Phase 5 final compliance eligible
• Tier 4 FAIL → isolate to module, fix via rtl-p4s-bugfix, retest Tier 2→3→4

Escalation & Stop Conditions

• Connectivity check fails (width mismatch, missing connection) → rtl-coder must fix before sim
• End-to-end reference mismatch → waveform-analyzer + func-verifier isolate to module → Tier 2 retest
• Handshake protocol violation → protocol-checker for detailed analysis
• Cross-module timing issue → timing-advisor for pipeline analysis
• More than 3 integration bugs found → escalate to rtl-architect for interface review

Final Checklist

• [ ] Connectivity check passed (reset propagation, clock, port widths)
• [ ] Integration TB written (SV or cocotb)
• [ ] End-to-end data flow tests executed
• [ ] Handshake/backpressure tests executed
• [ ] End-to-end reference comparison done (byte-by-byte)
• [ ] sim/top/integration_results.json produced
• [ ] Each data_flow and handshake entry has req_ids populated (REQ traceability)
• [ ] All integration tests PASS
• [ ] run_sim.sh used for SV TB simulation
• [ ] Waveform analysis done for any failures