Description
Six bug reports from one project. For each, name the formal app that finds this kind of bug with the least work.
Six bug reports each come with a small model, the check that exposes the bug and its recorded run. For each report, choose the formal app that finds this kind of bug with the least work. Then select the line the report is about.
Pinned interface
| Port | Direction | Bits |
|---|---|---|
| clk | input | 1 |
| rst | input | 1 |
| sel | input | 1 |
| a | input | 1 |
| b | input | 1 |
| wr | input | 1 |
| wdata | input | 1 |
| ev | input | 1 |
| go | input | 1 |
| abort | input | 1 |
| x | input | 2 |
| y | input | 2 |
| start | input | 1 |
| pad | output | 1 |
| flag | output | 1 |
| en | output | 1 |
| ph | output | 2 |
| busy | output | 1 |
| sum_ref | output | 3 |
| sum_new | output | 3 |
| done | output | 1 |
Full specification
Each card is a bug report with a small model of the block. It shows the check that exposes the bug and its recorded run. Each check came from a different source. The sources are a table, a register map, the RTL itself, a coverage report, the specification and a second design.
For every card, choose the app. Then select the line of the block that the report is about.
| Recorded case | Local contract |
|---|---|
| Report 1: pad 12 carries the wrong signal in one mode | The check is row 12 of the connection table. |
| Report 2: firmware clears the status flag by writing 0 | The check is the write-one-to-clear policy from the register map. |
| Report 3: the block is enabled before firmware arms it | The check is "not enabled in the first cycle after reset". No register has a declared value. |
| Report 4: the coverage report has an empty bin, phase 3 | The run is a cover of the bin at depth 20. |
| Report 5: an abort does not stop a running command | The check was written by a person from the specification: an abort ends the command in the next cycle. |
| Report 6: results changed after a timing fix | The check compares the changed adder with the approved one. |
| Execution rule | Meaning |
|---|---|
| Interface | Use the pinned port names, directions and bit widths shown above. Observed state and tracker outputs are not environment inputs. |
| Sampling | One positive edge of clk; binary values. Non-overlapping implication observes the next sample. Reset and initialization follow the source and contract above. |
| Submission | Select a cause and a numbered line for every recorded case. Checking these choices runs no engine. |
| How this is graded | The cause and responsible line must agree with each recorded local contract. A timeout or induction candidate must not be diagnosed as a reachable RTL failure. |
Report 1: Pad 12 Carries the Wrong Signal in One Mode
The check is row 12 of the connection table.
FAIL · bmc · depth 4 · 1105 ms
Samples from the measured run. Free input lanes are shaded; the terminal VCD record is omitted.
Numbered source and task
1 module top(input clk, input rst, sel, a, b, output pad);
2 // connection table, row 12: pad = sel ? a : b
3 assign pad = sel ? b : a;
4 always @(posedge clk) begin if ($initstate) begin assume(rst); end else begin assume(!rst); end end
5 endmodule
6 // props.sv
7 assert property (@(posedge clk) disable iff (rst) pad == ((sel && a) || (!sel && b)));Report 2: Firmware Clears the Status Flag by Writing 0
The check is the write-one-to-clear policy from the register map.
FAIL · bmc · depth 8 · 1031 ms
Samples from the measured run. Free input lanes are shaded; the terminal VCD record is omitted.
Numbered source and task
1 module top(input clk, input rst, wr, wdata, ev, output reg flag=0);
2 // register map: flag is W1C (write 1 to clear); a hardware event sets it
3 always @(posedge clk)
4 if (rst) flag <= 0;
5 else if (ev) flag <= 1;
6 else if (wr) flag <= 0;
7 always @(posedge clk) begin if ($initstate) begin assume(rst); end else begin assume(!rst); end end
8 endmodule
9 // props.sv
10 assert property (@(posedge clk) disable iff (rst) wr && !wdata && !ev && flag |=> flag);Report 3: The Block Is Enabled Before Firmware Arms It
The check is "not enabled in the first cycle after reset". No register has a declared value.
FAIL · bmc · depth 4 · 993 ms
Samples from the measured run. Free input lanes are shaded; the terminal VCD record is omitted.
Numbered source and task
1 module top(input clk, input rst, go, output en);
2 reg armed; // set by firmware after reset
3 always @(posedge clk)
4 if (go) armed <= 1;
5 assign en = armed;
6 reg age_ok=0; // high in the first cycle after reset only
7 always @(posedge clk) age_ok <= rst;
8 always @(posedge clk) begin if ($initstate) begin assume(rst); end else begin assume(!rst); end end
9 endmodule
10 // props.sv
11 assert property (@(posedge clk) disable iff (rst) age_ok |-> !en);Report 4: The Coverage Report Has an Empty Bin, Phase 3
The run is a cover of the bin at depth 20.
FAIL · cover · depth 20 · 1051 ms
No witness was produced for this query.
Numbered source and task
1 module top(input clk, input rst, start, output reg [1:0] ph=0, output reg done=0);
2 always @(posedge clk)
3 if (rst) begin ph <= 0; done <= 0; end
4 else begin
5 done <= ph == 2'd2;
6 if (ph == 2'd0) begin if (start) ph <= 2'd1; end
7 else if (ph == 2'd2) ph <= 2'd0;
8 else ph <= ph + 2'd1;
9 end
10 always @(posedge clk) begin if ($initstate) begin assume(rst); end else begin assume(!rst); end end
11 endmodule
12 // props.sv
13 cover property (@(posedge clk) !rst && ph == 3);Report 5: An Abort Does Not Stop a Running Command
The check was written by a person from the specification: an abort ends the command in the next cycle.
FAIL · bmc · depth 8 · 953 ms
Samples from the measured run. Free input lanes are shaded; the terminal VCD record is omitted.
Numbered source and task
1 module top(input clk, input rst, go, abort, output reg busy=0);
2 reg [1:0] n=0;
3 always @(posedge clk)
4 if (rst) begin busy <= 0; n <= 0; end
5 else if (!busy && go) begin busy <= 1; n <= 2'd3; end
6 else if (busy) begin
7 if (n == 0) busy <= 0; else n <= n - 2'd1;
8 end
9 always @(posedge clk) begin if ($initstate) begin assume(rst); end else begin assume(!rst); end end
10 endmodule
11 // props.sv
12 assert property (@(posedge clk) disable iff (rst) busy && abort |=> !busy);Report 6: Results Changed After a Timing Fix
The check compares the changed adder with the approved one.
FAIL · bmc · depth 4 · 969 ms
Samples from the measured run. Free input lanes are shaded; the terminal VCD record is omitted.
Numbered source and task
1 module top(input clk, input rst, input [1:0] x, y, output reg [2:0] sum_ref=0, output reg [2:0] sum_new=0);
2 // sum_ref: the approved RTL. sum_new: the same adder after a timing fix.
3 always @(posedge clk)
4 if (rst) begin sum_ref <= 0; sum_new <= 0; end
5 else begin
6 sum_ref <= x + y;
7 sum_new <= {x[1] & y[1], x[1] ^ y[1], x[0] ^ y[0]};
8 end
9 always @(posedge clk) begin if ($initstate) begin assume(rst); end else begin assume(!rst); end end
10 endmodule
11 // props.sv
12 assert property (@(posedge clk) disable iff (rst) sum_new == sum_ref);Requirements
- Use the published ports and widths. The checker and the DUT variants are owned by the judge.
- Do not add assumptions unless the statement explicitly asks for one. A passing assertion without a reachable trigger is insufficient.
Hints
Hint 1
Ask where the rule came from. A rule that a script can generate from a table belongs to a narrow app.
Hint 2
A missing feature has no table and no second design. Someone has to state what the block should do.
Hint 3
Use the "Which app catches which bug" table of F5.1.
Single clock, binary values, synchronous reset. The judge checks the pinned interface, reachability and the declared proof or bound. UNKNOWN, resource limits and timeouts remain inconclusive.
The recordings are fixed. Checking a diagnosis does not run an engine.
More problems on this topic