A Covergroup That Counted Idle Cycles

Where the interviewer goes next. Open a node to practise that follow-up.

WhyWhat IfProve ItCompareScale It

1. A Covergroup That Counted Idle Cycles
2. CompareClocked Covergroups vs Calling SampleL1 Junior

   Compare a clocked covergroup with one you sample by calling sample(). When is a clocking event the right choice?

   Reveal the Model Answer

   A clocking event fits values that matter every cycle, such as an FSM state or a FIFO level. Calling sample() fits events, such as transfers, packets and register writes. A class-based bench usually calls sample() from the code that sees each transaction. A clocked covergroup on event data counts idle cycles as if they were events.

   Both covergroups read the same kind. Only the one sampled on transfers reports what the link carried.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 A clocking event fits values that matter every cycle, such as an FSM state or a FIFO level. Key point 2 Calling sample() fits events, such as transfers, packets and register writes. Key point 3 A class-based bench usually calls sample() from the code that sees each transaction. Key point 4 A clocked covergroup on event data counts idle cycles as if they were events.

   Half-Truth

   "Clocked covergroups are always wrong."

   For state that lives every cycle, a clocked covergroup is the natural model. The error is using it for things that happen only on some cycles.

   Source Spear & Tumbush, SystemVerilog for Verification (3rd ed.), §9.6 (PDF p372)

   • Lesson C2Covergroups and Coverpoints
3. What IfWhat a Clocked Covergroup Sees at the EdgeL2 Mid-Level

   Asked if they keep a clocked covergroup

   What if the covergroup is clocked on the same edge that updates the flop it covers? What does this print?

   module tb;
     logic clk = 0;
     always #5 clk = ~clk;
     logic [1:0] q = 0;
     always_ff @(posedge clk) q <= q + 1;     // q is 0 only before the first edge
     covergroup cg @(posedge clk);
       cp: coverpoint q { bins zero = {0}; }
     endgroup
     cg c = new();
     initial begin
       repeat (3) @(posedge clk);
       #1 $display("q=%0d zero-bin coverage=%0.2f", q, c.get_coverage());
       $finish;
     end
   endmodule
   

   Reveal the Model Answer

   It prints zero-bin coverage=100.00. The covergroup samples when the edge event fires. The flop's nonblocking update lands later in that time step, so the sample sees the value from before the edge. That is the same value the flop held during the previous cycle. Plan bins with that one-cycle view in mind.

   One column per rising edge. At 5 ns the sample sees 0, the value from before the edge, so the zero bin is covered.

   RUNVerilator (coverage build) · simulator service

   Edit the code, then run it

   OutputWaves

   RunReset codeExpected: q=3 zero-bin coverage=100.00

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 It prints zero-bin coverage=100.00. Key point 2 The covergroup samples when the edge event fires. Key point 3 The flop's nonblocking update lands later in that time step, so the sample sees the value from before the edge. Key point 4 That is the same value the flop held during the previous cycle. Key point 5 Plan bins with that one-cycle view in mind.

   Misconception

   "It samples the new value, because the flop updates on that edge."

   Nonblocking updates happen in the NBA region, after the edge event wakes the covergroup. The run covers the bin for 0, a value held only before the first edge.

   Source IEEE 1800-2017 §4.4.2.4

   • Lesson C2Covergroups and Coverpoints
   • Lesson 6.2Scheduling Regions and Nonblocking Updates
4. Prove ItSolve 6.1.1: Count the Real TransfersSimulator-graded

   Show it, don't just say it: this Silicon Lab problem is graded by the simulator.

   Open Problem 6.1.1

5. Scale ItCoverage That Slowed the RegressionL2 Mid-LevelL3 Senior

   Your regression got 30% slower after the coverage model was added. What do you look at first?

   Reveal the Model Answer

   Look first for covergroups that sample every clock but only need events. Then look for large crosses and for coverage inside objects that are created by the thousand. Sample once per transaction, guard coverpoints with iff, and cut crosses down to the cells in the plan. Keep a switch that turns coverage off in soak tests that are not measuring it.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Look first for covergroups that sample every clock but only need events. Key point 2 Then look for large crosses and for coverage inside objects that are created by the thousand. Key point 3 Sample once per transaction, guard coverpoints with iff, and cut crosses down to the cells in the plan. Key point 4 Keep a switch that turns coverage off in soak tests that are not measuring it.

   What Interviewers Listen For (Senior)

   A good senior answer measures before changing anything, for example the run time with coverage off. It says which tests still need coverage on for sign-off.

   Wrong Altitude

   "Turn coverage off in the regression and turn it on before tape-out."

   Coverage closure needs the data all along. Holes found at the end leave no time to write the missing tests.

   Source Spear & Tumbush, SystemVerilog for Verification (3rd ed.), §9.1 (PDF p361)

   • Lesson C6Coverage in a Class
   • Lesson U15.3Performance and Scale
6. WhyA Covergroup Inside the Transaction ClassL2 Mid-LevelL3 Senior

   A teammate embeds the covergroup in the transaction class, and the test creates a million transactions. What happens?

   Reveal the Model Answer

   Each transaction object builds its own covergroup instance in its constructor. That is a million covergroup instances, each sampled once or twice, and memory and run time grow with them. Put covergroups in components that live for the whole run, such as a coverage collector, and pass each transaction to them.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Each transaction object builds its own covergroup instance in its constructor. Key point 2 That is a million covergroup instances, each sampled once or twice, and memory and run time grow with them. Key point 3 Put covergroups in components that live for the whole run, such as a coverage collector, and pass each transaction to them.

   What Interviewers Listen For (Senior)

   A good senior answer turns this into a review rule. Covergroups live in components that last the whole run, never in transactions or sequences.

   Misconception

   "The covergroup is shared by all objects of the class, like a static member."

   An embedded covergroup is an instance member. Each object that runs its constructor creates its own instance.

   Source IEEE 1800-2017 §19.4

   • Lesson C6Coverage in a Class
   • Lesson U15.3Performance and Scale