Functional Coverage for a Bus

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

WhyWhat IfProve ItBreak ItCompareScale It

1. Functional Coverage for a Bus
2. WhyWhere Coverage Is SampledL2 Mid-Level

   Why sample coverage in the monitor at completion rather than in the sequence or the driver?

   Reveal the Model Answer

   Coverage should record what the design actually did. The monitor sees only accepted transactions, with their real response and timing. It also works the same when the agent is passive (only watching) in a larger environment. Sampling at completion lets a bin record the response and the stall pattern. The driver cannot know these when it sends the item.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Coverage should record what the design actually did. Key point 2 The monitor sees only accepted transactions, with their real response and timing. Key point 3 It also works the same when the agent is passive (only watching) in a larger environment. Key point 4 Sampling at completion lets a bin record the response and the stall pattern. Key point 5 The driver cannot know these when it sends the item.

   What Interviewers Listen For (Mid-Level)

   A good mid-level answer gives two reasons. The monitor sees what was really accepted, not what was intended, and it still works when the agent is passive.

   • Lesson P2.17An AXI Agent, Monitor and Scoreboard
   • Lesson U15.1Coverage Architecture and Planning
3. What IfCoverage Sampled in the DriverL2 Mid-Level

   What if coverage is sampled in the driver instead? Show a bin it fills falsely.

   Reveal the Model Answer

   Say a write is driven, but reset aborts it before the handshake. The driver has already counted a write of that kind. The bin closes, yet the design never completed it. Driver coverage measures what the stimulus asked for, not what the design did.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Say a write is driven, but reset aborts it before the handshake. Key point 2 The driver has already counted a write of that kind. Key point 3 The bin closes, yet the design never completed it. Key point 4 Driver coverage measures what the stimulus asked for, not what the design did.

   Half-Truth

   "Driver coverage equals monitor coverage when there are no errors."

   Reset, a dropped item or a stalled channel can stop a driven item from completing. The driver cannot see that.

   Source Lesson U15.1 key points

   • Lesson U15.1Coverage Architecture and Planning
   • Lesson P2.17An AXI Agent, Monitor and Scoreboard
4. CompareCode Coverage Versus Functional Coverage on a BusL1 JuniorL2 Mid-Level

   For a bus subordinate, what does code coverage tell you that functional coverage does not, and the other way round?

   Reveal the Model Answer

   Code coverage finds RTL that no test reached, such as an error path or an unused state. Nobody has to think of these in advance. Functional coverage checks that the scenarios the spec and plan care about happened. That includes combinations and sequences that no single line of code represents. Code coverage cannot see a missing feature and functional coverage cannot see dead code, so you need both.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Code coverage finds RTL that no test reached, such as an error path or an unused state. Key point 2 Nobody has to think of these in advance. Key point 3 Functional coverage checks that the scenarios the spec and plan care about happened. Key point 4 That includes combinations and sequences that no single line of code represents. Key point 5 Code coverage cannot see a missing feature and functional coverage cannot see dead code, so you need both.

   What Interviewers Listen For (Junior)

   A good junior answer says code coverage is automatic and follows the RTL structure. Functional coverage is written by hand from the spec.

   • Lesson C1What Coverage Measures
5. Scale ItSizing a Cross Before Writing ItL3 Senior

   How many bins does a burst x length x size cross have on a 32-bit AXI interface? How would you cut it down?

   Reveal the Model Answer

   Start from the legal set, not the raw one. On a 32-bit bus Size is 1, 2 or 4 bytes, and INCR allows 256 lengths, FIXED 16 and WRAP 4. That gives 3 x (256 + 16 + 4) = 828 legal combinations, against 6,144 raw. Then bucket the lengths where behaviour does not change: 1, 2 to 15, 16, 17 to 255, and 256. That leaves a few dozen bins, and each one means something.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Start from the legal set, not the raw one. Key point 2 On a 32-bit bus Size is 1, 2 or 4 bytes, and INCR allows 256 lengths, FIXED 16 and WRAP 4. Key point 3 That gives 3 x (256 + 16 + 4) = 828 legal combinations, against 6,144 raw. Key point 4 Then bucket the lengths where behaviour does not change: 1, 2 to 15, 16, 17 to 255, and 256. Key point 5 That leaves a few dozen bins, and each one means something.

   What Interviewers Listen For (Senior)

   A good senior answer counts the legal combinations first. It places each length bucket where the design's behaviour really changes.

   • Lesson C5Cross Coverage
6. Break ItThe Backpressure Nobody CoveredL2 Mid-LevelL3 Senior

   All coverage is green, yet a customer hits a hang when the response channel is stalled. How did that escape?

   Illustrative regression/customer comparison
   Regression: response traffic covered, all reported coverage goals met
   Customer: RVALID=1, RREADY=0 for an extended interval
   RREADY returns to 1; transfer never completes

   Reveal the Model Answer

   Most likely the bench kept BREADY and RREADY high, so the subordinate's response-side stall path never ran. Every stall rule then passed vacuously, because no stall ever happened. The coverage model also had no stall-length bins on B or R. Randomise ready with long low runs on every channel the testbench drives, and cover stall length per channel. Also count how often each checker antecedent matched.

   attempt startsvacuous success: the antecedent did not match

   1. Edge 2 · t=15

      Beat one is taken at 15 with ready already high. valid && !ready is false, so the attempt ends as it starts.

   2. Edge 5 · t=45

      Five edges, five open circles. The hold rule was never exercised, and the report says pass.

   Read the lanes as text

   • edge @5, from edge 1 (t=5): succeeds vacuously at edge 1 (t=5).
   • edge @15, from edge 2 (t=15): succeeds vacuously at edge 2 (t=15).
   • edge @25, from edge 3 (t=25): succeeds vacuously at edge 3 (t=25).
   • edge @35, from edge 4 (t=35): succeeds vacuously at edge 4 (t=35).
   • edge @45, from edge 5 (t=45): succeeds vacuously at edge 5 (t=45).

   Sampled values, edge 1 to 6: valid 0 1 1 1 0 0; ready 1 1 1 1 1 1.

   The program behind the lanes

   The lanes come from running this program on Verilator 5.052. Each lane is a separate copy of hold_valid that can start only at that lane's edge, so every verdict and its edge are read from that copy's own output.

   module tb;
     logic clk = 0;
     always #5 clk = ~clk;                 // rising edges at 5, 15, 25, ...
     logic valid = 0, ready = 1;           // the sink is ready at every edge
   
     hold_valid: assert property (@(posedge clk) valid && !ready |=> valid);
   
     initial begin
       @(negedge clk); valid = 1;          // three beats, taken at 15, 25 and 35
       @(negedge clk);
       @(negedge clk);
       @(negedge clk); valid = 0;
       @(negedge clk);
       @(negedge clk);
       $finish;
     end
   endmodule
   

   A sink that is always ready never stalls the link. The hold rule passes five times without ever being tested.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Most likely the bench kept BREADY and RREADY high, so the subordinate's response-side stall path never ran. Key point 2 Every stall rule then passed vacuously, because no stall ever happened. Key point 3 The coverage model also had no stall-length bins on B or R. Key point 4 Randomise ready with long low runs on every channel the testbench drives, and cover stall length per channel. Key point 5 Also count how often each checker antecedent matched.

   What Interviewers Listen For (Mid-Level)

   A good mid-level answer points at ready held high on the response channels, so no stall ever happened.

   • Lesson P1.10From the Spec to a Check
   • Lesson P2.16The AXI Protocol Checker
   • Problem P2.16.1Six Properties for the Link
7. Prove ItSolve P2.17.1: The AXI MonitorSimulator-graded

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

   Open Problem P2.17.1