Hole Analysis Workflow: Triage Before Targeting

Not every hole needs a new test

Hole analysis classifies each uncovered bin before anyone writes stimulus. Categories: stimulus gap (reachable, not exercised), constraint conflict (random constraints exclude the cell), checker/coverage bug (sampling wrong stream), and genuinely unreachable (spec + implementation make the bin impossible).

[CLOSE][SENIOR][UVM] hole triage decision tree

  UNCOVERED BIN
       │
       ├─ sampled on monitor stream? ──NO──► fix subscriber / reset gate
       │
       ├─ constraints exclude cell? ──YES──► relax constraint or directed seq
       │
       ├─ spec allows scenario? ──NO──► waiver candidate (document)
       │
       └─ stimulus gap ──► closure test / targeted vseq
                │
                └─ still open after 3 targeted runs? ──► escalate design review

Hole ticket template

HOLE TICKET
  Plan ID:     FEAT-AXI-030
  Bin:         axi_cov.x_wr_large.wr_large
  Category:    stimulus gap | constraint | bug | unreachable
  Owner:       verification engineer name
  Root cause:  one-line hypothesis
  Action:      directed seq / constraint fix / waiver request
  Status:      open | targeted | closed | waived

• Sort holes by plan priority — mandatory features before nice-to-have.

• Check if any seed came close (partial cross hits) before writing new stimulus.

• Involve design when reachability is unclear — do not guess waivers.

Key takeaways

• Classify every hole before writing closure stimulus.

• Hole tickets make closure meetings factual, not opinion debates.

• Unreachable bins need spec review — not silent waivers.

Common pitfalls

• Writing directed tests for bins that are illegal per spec.

• Assuming random will eventually hit a cross with 1-in-10M probability.

• Closing tickets when one seed hit once — confirm merge stability.

Root-cause signals

Common signatures that point to each category:

[CLOSE][SENIOR] signature → category

  scalar covered, cross never hit        → constraint conflict or ordering
  bin hit in block TB, hole at chip level  → passive agent / wrong hierarchy
  bin hit before ECO, hole after           → RTL change or sampling regression
  related bins all hit, one cell missing   → targeted directed seq candidate

Closure meeting agenda (30 min)

1. Review merged hole diff vs last milestone.

2. Present top 10 priority holes with category + owner.

3. Approve waiver requests with design sign-off.

4. Assign closure sequences for stimulus-gap holes only.

Key takeaways

• Cross holes with covered scalars often mean constraint conflict.

• Chip-level holes may be hierarchy or passive-agent issues.

Common pitfalls

• Skipping category classification — teams write useless directed tests.

• Waiving without design attendance in the closure meeting.