SV Day 15 — Coverage Closure

Why Closure Matters

Coverage isn't a one-time measurement—it's a continuous process. As RTL changes, coverage can regress. You need automated tracking and closure gates.

Regression Monitoring

// Continuous coverage collection
class CoverageTracker;
  real coverage_history[$];
  string dates[$];

  function void record(real cov, string date);
    coverage_history.push_back(cov);
    dates.push_back(date);
    if (cov < 80.0) begin
      $warning("Coverage REGRESSION: %0.1f%%", cov);
    end
  endfunction

  function void report();
    real avg = 0;
    foreach(coverage_history[i]) avg += coverage_history[i];
    avg /= coverage_history.size();
    $display("Avg coverage: %0.1f%%", avg);
  endfunction
endclass

Closure Criteria

Typical Sign-Off Criteria:
• Code coverage: 100% (statements, branches)
• Functional coverage: ≥90%
• No regression from previous build
• All critical scenarios hit
• Directed tests for known gaps

Automation & CI Integration

// run_tests.sh - automated nightly regression
#!/bin/bash
COVERAGE_THRESHOLD=85

# Compile & run tests
vcs -full64 -sverilog tb.sv testbench.sv
./simv -dir work +coverage=functional

# Extract coverage
coverage_result=$(grep "Overall coverage:" cov.log | awk '{print $3}')
coverage_pct=${coverage_result%\%}

# Pass/Fail
if (( $(echo "$coverage_pct >= $COVERAGE_THRESHOLD" | bc -l) )); then
  echo "✓ Coverage PASSED: $coverage_pct%"
  exit 0
else
  echo "✗ Coverage FAILED: $coverage_pct% < $COVERAGE_THRESHOLD%"
  exit 1
fi

Key Takeaways

✅ Monitor coverage across all regressions
✅ Set closure gates (don't ship below threshold)
✅ Automate metric extraction in CI/CD
✅ Merge coverage from all test suites
✅ Coverage closure ≠ bug-free, but essential

Tomorrow (Day 16): Constrained-random + coverage — automated test generation.