Scoreboard Env Wiring: Connect-Phase Contracts

Wiring contracts

Keep scoreboard wiring centralized in env connect_phase so all data paths are reviewable and deterministic.

[ENV][UVM][CHECK] scoreboard wiring contract

required links:
  ctrl_agt.mon.ap -> sb.ctrl_imp
  req_agt.mon.ap  -> sb.req_imp
  rsp_agt.mon.ap  -> sb.rsp_imp
  err_agt.mon.ap  -> sb.err_imp

reject:
  conditional runtime connects
  hidden helper-side reconnects

function void connect_phase(uvm_phase phase);
  super.connect_phase(phase);
  ctrl_agt.mon.ap.connect(sb.ctrl_imp);
  req_agt.mon.ap.connect(sb.req_imp);
  rsp_agt.mon.ap.connect(sb.rsp_imp);
  err_agt.mon.ap.connect(sb.err_imp);
endfunction

• Every scoreboard input should map to exactly one monitor stream.

• Connect code should be short, explicit, and discoverable.

• Missing links should fail fast in smoke tests.

Key takeaways

• Explicit wiring contracts prevent silent checker bypasses.

• Connect-phase centralization improves code review effectiveness.

Common pitfalls

• Using convenience wrappers that hide critical connects.

• Deferring connectivity decisions to run_phase.

Applied Patterns

Pair connect statements with expected-link assertions so graph regressions are caught immediately.

[CHECK] link assertion pattern

expected_links = 4
actual_links   = count_connects()

if actual != expected:
  fatal with missing endpoint names

optional:
  print full graph table in UVM_LOW

function void assert_sb_links();
  if (sb_link_count != 4)
    `uvm_fatal("SB_WIRE",
      $sformatf("scoreboard links expected=4 got=%0d", sb_link_count))
endfunction

• Assertion failures should name missing endpoints explicitly.

• Link-count checks are cheap and highly effective.

• Run assertion before first sequence item is issued.

Integration Drilldown

Intentionally remove one connect and confirm the assertion pinpoints the exact missing edge.

[CHECK] missing-edge test

inject:
  comment rsp_agt.mon -> sb.rsp_imp connect

expect:
  startup fatal naming rsp endpoint
  no deep-run false confidence

Key takeaways

• Edge-removal tests validate checker wiring diagnostics.

Common pitfalls

• Trusting full regressions without startup connectivity guards.