← Design Verification Interviews
Interview Question · Design
Closing the Last Eight Percent
The Question
The interviewer asks
Functional coverage has sat at 92% for two weeks. Tape-out is a month away. How do you close it?
You have two and a half minutes. Speak or type, then press Get Feedback under your answer. No credits needed.
Speech-to-text runs in your browser. Chrome and Edge may send audio to their own speech service.
You can edit your answer before you send it. Feedback uses AI and checks your answer only against our written answer. No credits are used and your answer is not saved.
Reveal the Model Answer
First list the holes and sort them by cause and by risk, not by count. Fix bench limits, such as wrong constraints or missing knobs, because one fix often clears many holes. Bias the constraints toward rare cases, and write directed tests only for cases random stimulus cannot reach in time. Prove unreachable holes with formal or a designer review, and exclude them with the reason. Report what remains as named risks.
- Cycle 0
After 32 random items, 6 of 9 bins are hit. The holes are level 4, write when full and reset with data.
- Cycle 1
32 more random items reach level 4. Write when full is still a hole, and no seed can produce the reset.
- Cycle 2
Six directed items: the fifth write meets a full FIFO, then a reset finds data stored. All 9 bins are hit.
Measured on a FIFO bench. More random items could not reach the bins the constraints forbid. Six directed items closed them.
Self-Check
Tick each key point you covered in your answer.
What Interviewers Listen For (Mid-Level)
A good mid-level answer groups the holes by cause. It starts with the one fix that clears the most holes.
What Interviewers Listen For (Senior)
A good senior answer ranks the holes by how much damage a bug there would do. It agrees with the project which remaining risks are accepted.
Cross-Questions
Where the interviewer goes next. Open a node to practise that follow-up.
WhyWhat IfCompareScale It
- Closing the Last Eight Percent
-
WhyWhy the Last Few Percent Take the LongestL1 JuniorL2 Mid-Level
Why does coverage rise fast at first and then crawl?
2:30You have two and a half minutes. Speak or type, then press Get Feedback under your answer. No credits needed.
Speech-to-text runs in your browser. Chrome and Edge may send audio to their own speech service.
You can edit your answer before you send it. Feedback uses AI and checks your answer only against our written answer. No credits are used and your answer is not saved.
Reveal the Model Answer
Early samples land on empty bins, so almost every sample adds coverage. Later, most samples land on bins already hit, and only the rare ones add anything. Bins with low probability under the constraints may take thousands of samples each. Holes that cannot be reached never close, so they look like a crawl that never ends.
Self-Check
Tick each key point you covered in your answer.
Half-Truth
"Slow progress means the tests are bad."
Some slowdown is normal arithmetic: the last empty bins are the hardest to hit by chance. A plateau, where progress stops, is the real warning.
Source Spear & Tumbush, SystemVerilog for Verification (3rd ed.), §9.1 (PDF p361)
-
Scale ItReporting Closure ProgressL3 Senior
Your lead wants a weekly closure report for a subsystem. What goes in it besides one percentage?
2:30You have two and a half minutes. Speak or type, then press Get Feedback under your answer. No credits needed.
Speech-to-text runs in your browser. Chrome and Edge may send audio to their own speech service.
You can edit your answer before you send it. Feedback uses AI and checks your answer only against our written answer. No credits are used and your answer is not saved.
Reveal the Model Answer
Show coverage per plan feature, not one total, with its trend over the weeks. Show how many new bins each regression adds, because a flat line means you need new work, not more runs. Add the open holes grouped by cause and owner, and the bug rate. A falling bug rate with flat coverage is the pattern you want before sign-off.
Self-Check
Tick each key point you covered in your answer.
Wrong Altitude
"One merged percentage is enough for management."
One number hides which feature is behind and why. A 95% total can mean one critical feature at 40%.
Source Mehta, SystemVerilog Assertions and Functional Coverage (3rd ed.), §25.5 (PDF p446)
-
CompareBiasing Random Stimulus vs a Directed TestL2 Mid-Level
Compare biasing the constraints toward a hole with writing a directed test for it.
2:30You have two and a half minutes. Speak or type, then press Get Feedback under your answer. No credits needed.
Speech-to-text runs in your browser. Chrome and Edge may send audio to their own speech service.
You can edit your answer before you send it. Feedback uses AI and checks your answer only against our written answer. No credits are used and your answer is not saved.
Reveal the Model Answer
Biasing, for example with
distweights or a test-level constraint, keeps the other fields random. So the rare case also meets many neighbours. A directed test hits the case once, fast and repeatably. Bias first when the case is reachable but rare. Go directed when it needs a long exact sequence, such as an error during a full FIFO after reset.Find the Holes, Then Aim at ThemA loop finds the empty cells from its own hit counts. It then asks the solver for one packet per hole with an inline constraint, and the other fields stay random. Read the diagram as text
- initial block: round 1, then round 2
- p.randomize(): class constraints only
- p.randomize() with: the kind and class of one hole
- record(p): cov.sample(p); hits[k][c]++
- cov: pkt_cross: 7 legal cells
- hits[kind][class]: the same cells, counted
- initial block → p.randomize(): randomize
- initial block → p.randomize() with: aimed randomize
- initial block → record(p): record(p)
- record(p) → cov: sample(p)
- record(p) → hits[kind][class]: count
- hits[kind][class] → initial block: holes()
The hole list comes from the counts, so each aimed packet goes exactly where coverage is missing, and nowhere else.
Self-Check
Tick each key point you covered in your answer.
Wrong Altitude
"Directed tests are old-fashioned; always bias."
Some holes need a precise sequence of events that random stimulus almost never builds. A directed test is the cheapest way there.
Source Spear & Tumbush, SystemVerilog for Verification (3rd ed.), §9.1 (PDF p362)
-
What IfTape-Out with Coverage Holes LeftL3 SeniorL4 Staff
What if tape-out is in a week and 3% is still open? What do you do?
2:30You have two and a half minutes. Speak or type, then press Get Feedback under your answer. No credits needed.
Speech-to-text runs in your browser. Chrome and Edge may send audio to their own speech service.
You can edit your answer before you send it. Feedback uses AI and checks your answer only against our written answer. No credits are used and your answer is not saved.
Reveal the Model Answer
Go through each remaining hole with the designers and the project lead. For each, state what feature it covers, how likely the case is in real use, and what a bug there would cost. Close the high-risk ones with targeted tests or formal. For the rest, record an accepted risk with an owner, plus a plan for post-silicon tests or a software workaround.
Self-Check
Tick each key point you covered in your answer.
What Interviewers Listen For (Staff)
A good staff answer turns the hole list into a risk register that the lead signs. Each accepted item gets a post-silicon test plan.
Delivery
"Waive everything open; the date is fixed."
A blanket waiver hides the one hole that matters. Each item needs a decision, even if most of them are accepted.
What People Get Wrong
Common wrong answers and why they fail.
Half-Truth
"Double the seeds in the nightly run."
That only helps holes that are rare but reachable. A plateau usually means the generator cannot reach them at all.
Source Spear & Tumbush, SystemVerilog for Verification (3rd ed.), §9.1 (PDF p362)
Misconception
"Lower option.goal to 90 and we are done."
The goal only changes when reports call the group finished. The holes, and any bugs in them, are still there.
Source Spear & Tumbush, SystemVerilog for Verification (3rd ed.), §9.10.5 (PDF p394)
Delivery
Starts proposing fixes before asking what the holes are.
The right action depends on why each hole is empty. A good answer begins by reading the holes.