An unbounded proof returns more than a verdict: it returns the invariant it found, and that invariant is evidence to review, not decoration. A grown invariant that nobody reads can smuggle in exactly the over-constraint a hand-written assumption would, while a well-read one teaches the team the design's own structure. This lesson decodes the progress report and runs the comparison exercise on the FIFO, where the human helper is known and the machine alternative is described as concept.
A progress report has the same fields whatever the engine: which engine ran, the bound or frame depth reached, the status of each obligation, the invariant found, and the time and memory spent. The table below says what each field demands from you. The exercise compares the human helper, the tracked-slot invariant that the FIFO proof asserts, with what a learning engine grows: often a larger, less readable conjunction that implies the same closure. The human form is reviewable and reusable; the machine form is complete and ugly. Signoff wants the human form wherever the machine form cannot be explained, which is why the runnable proof below asserts the helper a person would write.
Report field | What it means | What you do with it |
|---|---|---|
Engine | Which algorithm answered | Check it matches the plan for this property family |
Depth or frames | How far the search reached | Compare with the justified bound; frames must stabilise |
Base and step status | Each obligation separately | Never accept one half as the verdict |
Invariant found | The grown over-approximation | Review it like a helper; keep it if it is readable |
Blocked cubes per frame | How much the engine had to learn | Counts that keep growing with no stable frame mean a stall |
Time and memory | Cost of this run | Memory growth without frame progress means stall, not work |
The FIFO data-integrity proof with its tracked-slot helper asserted closes at depth 4: the helper a person writes, memory holds the tracked value, proven alongside the target. Without it the run at depth 12 is inconclusive, the step failing on data integrity. An engine that learns clauses might or might not find a comparable invariant over the memory and pointer bits (wide memory is the usual stall case, and no engine is run on this question here); the exercise is to compare the two invariants for readability and reuse, keeping the human form in the proof. The runnable blocks below are the human half of that comparison, with the failing partner that shows the helper is load-bearing: with write data corrupted, the helper itself fails.
// formal: prove 4 expect pass slang top sync_fifo
`define FORMAL
// sync_fifo: 4-entry, 8-bit synchronous FIFO. Running example of the Formal course.
// Synchronous active-high reset. count is an output so properties can see it.
// BUG selects a planted bug: 0 none, 1 push accepted when full, 2 pop accepted when
// empty, 3 full flag one entry early, 4 simultaneous push+pop miscounts, 5 write
// data bit 7 stuck at 0.
module sync_fifo #(parameter BUG = 0, parameter HELPER = 1) (
input wire clk,
input wire rst,
input wire push,
input wire [7:0] wdata,
input wire pop,
output wire [7:0] rdata,
output wire full,
output wire empty,
output reg [2:0] count
);
reg [7:0] mem [0:3];
reg [1:0] wptr, rptr;
assign full = (BUG == 3) ? (count >= 3'd3) : (count == 3'd4);
assign empty = (count == 3'd0);
assign rdata = mem[rptr];
wire do_push = push && (!full || BUG == 1);
wire do_pop = pop && (!empty || BUG == 2);
always @(posedge clk) begin
if (rst) begin
wptr <= 2'd0;
rptr <= 2'd0;
count <= 3'd0;
end else begin
if (do_push) begin
mem[wptr] <= (BUG == 5) ? {1'b0, wdata[6:0]} : wdata;
wptr <= wptr + 2'd1;
end
if (do_pop) rptr <= rptr + 2'd1;
if (BUG == 4 && do_push && do_pop) count <= count + 3'd1; // forgets the pop
else count <= count + {2'b0, do_push} - {2'b0, do_pop};
end
end
`ifdef FORMAL
// ---- harness: first cycle is reset, state before it is arbitrary ----
reg init = 1'b1;
always @(posedge clk) init <= 1'b0;
always @(*) if (init) assume (rst);
// ---- safety properties (after the reset cycle) ----
always @(*) if (!init) begin
a_range: assert (count <= 3'd4); // no overflow, no underflow wrap
a_flags: assert (full == (count == 3'd4) && empty == (count == 3'd0));
a_not_both: assert (!(full && empty));
// occupancy equals pointer distance (mod 4)
a_ptr: assert (count[1:0] == (wptr - rptr));
end
// ---- transition properties (use history: skip the cycle after reset) ----
always @(posedge clk) if (!init && !$past(rst)) begin
// push alone when not full adds one; pop alone when not empty removes one
a_push_inc: assert (!($past(push) && !$past(pop) && !$past(full)) || count == $past(count) + 3'd1);
a_pop_dec: assert (!($past(pop) && !$past(push) && !$past(empty)) || count == $past(count) - 3'd1);
// push into a full FIFO (and no pop) is ignored: occupancy unchanged
a_ovf_ign: assert (!($past(push) && !$past(pop) && $past(count) == 3'd4) || count == $past(count));
// simultaneous push and pop (neither empty nor full) leaves occupancy unchanged
a_both: assert (!($past(push) && $past(pop) && !$past(full) && !$past(empty)) || count == $past(count));
end
// ---- data integrity by tracked-slot abstraction (one free slot, one shadow value) ----
// free constant: (* anyconst *) alone is not honoured by the front end, so the
// value is also assumed equal to its one-cycle-old copy (works in both front ends;
// a clocked assume ($stable) lags one step and leaks at the last step: do not use it)
(* anyconst *) wire [1:0] tslot;
reg [1:0] tslot_q;
always @(posedge clk) tslot_q <= tslot;
always @(*) if (!init) assume (tslot == tslot_q);
reg tv; // the tracked slot holds a value we are watching
reg [7:0] td; // the value that should be there
always @(posedge clk) begin
if (rst) tv <= 1'b0;
else begin
if (do_push && wptr == tslot) begin tv <= 1'b1; td <= wdata; end
else if (do_pop && rptr == tslot) tv <= 1'b0;
end
end
always @(*) if (!init && tv && !empty && rptr == tslot) a_data: assert (rdata == td);
// helper invariant needed for the induction step of a_data
generate if (HELPER) begin : g_helper
always @(*) if (!init && tv) a_tracked: assert (mem[tslot] == td);
end endgenerate
// ---- covers: the properties are not vacuous ----
always @(*) if (!init) begin
c_full: cover (count == 3'd4);
c_empty_after_use: cover (empty && tv == 1'b0 && wptr == 2'd2);
end
always @(posedge clk) if (!init) c_both: cover (push && pop && !full && !empty);
`endif
endmoduleHuman helper asserted and proven: the tracked-slot invariant holds, data integrity is inductive under it, and a reviewer can read why. Keep this form in the proof.
// formal: prove 12 expect unproven slang top sync_fifo define HELPER=0
`define FORMAL
// sync_fifo: 4-entry, 8-bit synchronous FIFO. Running example of the Formal course.
// Synchronous active-high reset. count is an output so properties can see it.
// BUG selects a planted bug: 0 none, 1 push accepted when full, 2 pop accepted when
// empty, 3 full flag one entry early, 4 simultaneous push+pop miscounts, 5 write
// data bit 7 stuck at 0.
module sync_fifo #(parameter BUG = 0, parameter HELPER = 1) (
input wire clk,
input wire rst,
input wire push,
input wire [7:0] wdata,
input wire pop,
output wire [7:0] rdata,
output wire full,
output wire empty,
output reg [2:0] count
);
reg [7:0] mem [0:3];
reg [1:0] wptr, rptr;
assign full = (BUG == 3) ? (count >= 3'd3) : (count == 3'd4);
assign empty = (count == 3'd0);
assign rdata = mem[rptr];
wire do_push = push && (!full || BUG == 1);
wire do_pop = pop && (!empty || BUG == 2);
always @(posedge clk) begin
if (rst) begin
wptr <= 2'd0;
rptr <= 2'd0;
count <= 3'd0;
end else begin
if (do_push) begin
mem[wptr] <= (BUG == 5) ? {1'b0, wdata[6:0]} : wdata;
wptr <= wptr + 2'd1;
end
if (do_pop) rptr <= rptr + 2'd1;
if (BUG == 4 && do_push && do_pop) count <= count + 3'd1; // forgets the pop
else count <= count + {2'b0, do_push} - {2'b0, do_pop};
end
end
`ifdef FORMAL
// ---- harness: first cycle is reset, state before it is arbitrary ----
reg init = 1'b1;
always @(posedge clk) init <= 1'b0;
always @(*) if (init) assume (rst);
// ---- safety properties (after the reset cycle) ----
always @(*) if (!init) begin
a_range: assert (count <= 3'd4); // no overflow, no underflow wrap
a_flags: assert (full == (count == 3'd4) && empty == (count == 3'd0));
a_not_both: assert (!(full && empty));
// occupancy equals pointer distance (mod 4)
a_ptr: assert (count[1:0] == (wptr - rptr));
end
// ---- transition properties (use history: skip the cycle after reset) ----
always @(posedge clk) if (!init && !$past(rst)) begin
// push alone when not full adds one; pop alone when not empty removes one
a_push_inc: assert (!($past(push) && !$past(pop) && !$past(full)) || count == $past(count) + 3'd1);
a_pop_dec: assert (!($past(pop) && !$past(push) && !$past(empty)) || count == $past(count) - 3'd1);
// push into a full FIFO (and no pop) is ignored: occupancy unchanged
a_ovf_ign: assert (!($past(push) && !$past(pop) && $past(count) == 3'd4) || count == $past(count));
// simultaneous push and pop (neither empty nor full) leaves occupancy unchanged
a_both: assert (!($past(push) && $past(pop) && !$past(full) && !$past(empty)) || count == $past(count));
end
// ---- data integrity by tracked-slot abstraction (one free slot, one shadow value) ----
// free constant: (* anyconst *) alone is not honoured by the front end, so the
// value is also assumed equal to its one-cycle-old copy (works in both front ends;
// a clocked assume ($stable) lags one step and leaks at the last step: do not use it)
(* anyconst *) wire [1:0] tslot;
reg [1:0] tslot_q;
always @(posedge clk) tslot_q <= tslot;
always @(*) if (!init) assume (tslot == tslot_q);
reg tv; // the tracked slot holds a value we are watching
reg [7:0] td; // the value that should be there
always @(posedge clk) begin
if (rst) tv <= 1'b0;
else begin
if (do_push && wptr == tslot) begin tv <= 1'b1; td <= wdata; end
else if (do_pop && rptr == tslot) tv <= 1'b0;
end
end
always @(*) if (!init && tv && !empty && rptr == tslot) a_data: assert (rdata == td);
// helper invariant needed for the induction step of a_data
generate if (HELPER) begin : g_helper
always @(*) if (!init && tv) a_tracked: assert (mem[tslot] == td);
end endgenerate
// ---- covers: the properties are not vacuous ----
always @(*) if (!init) begin
c_full: cover (count == 3'd4);
c_empty_after_use: cover (empty && tv == 1'b0 && wptr == 2'd2);
end
always @(posedge clk) if (!init) c_both: cover (push && pop && !full && !empty);
`endif
endmoduleHelper removed: base passes, step fails on data integrity for every depth tried up to 12. Whatever invariant the machine would grow, it must exclude the untracked-memory windows this run exhibits.
// formal: bmc 3 expect fail slang top sync_fifo define BUG=5
`define FORMAL
// sync_fifo: 4-entry, 8-bit synchronous FIFO. Running example of the Formal course.
// Synchronous active-high reset. count is an output so properties can see it.
// BUG selects a planted bug: 0 none, 1 push accepted when full, 2 pop accepted when
// empty, 3 full flag one entry early, 4 simultaneous push+pop miscounts, 5 write
// data bit 7 stuck at 0.
module sync_fifo #(parameter BUG = 0, parameter HELPER = 1) (
input wire clk,
input wire rst,
input wire push,
input wire [7:0] wdata,
input wire pop,
output wire [7:0] rdata,
output wire full,
output wire empty,
output reg [2:0] count
);
reg [7:0] mem [0:3];
reg [1:0] wptr, rptr;
assign full = (BUG == 3) ? (count >= 3'd3) : (count == 3'd4);
assign empty = (count == 3'd0);
assign rdata = mem[rptr];
wire do_push = push && (!full || BUG == 1);
wire do_pop = pop && (!empty || BUG == 2);
always @(posedge clk) begin
if (rst) begin
wptr <= 2'd0;
rptr <= 2'd0;
count <= 3'd0;
end else begin
if (do_push) begin
mem[wptr] <= (BUG == 5) ? {1'b0, wdata[6:0]} : wdata;
wptr <= wptr + 2'd1;
end
if (do_pop) rptr <= rptr + 2'd1;
if (BUG == 4 && do_push && do_pop) count <= count + 3'd1; // forgets the pop
else count <= count + {2'b0, do_push} - {2'b0, do_pop};
end
end
`ifdef FORMAL
// ---- harness: first cycle is reset, state before it is arbitrary ----
reg init = 1'b1;
always @(posedge clk) init <= 1'b0;
always @(*) if (init) assume (rst);
// ---- safety properties (after the reset cycle) ----
always @(*) if (!init) begin
a_range: assert (count <= 3'd4); // no overflow, no underflow wrap
a_flags: assert (full == (count == 3'd4) && empty == (count == 3'd0));
a_not_both: assert (!(full && empty));
// occupancy equals pointer distance (mod 4)
a_ptr: assert (count[1:0] == (wptr - rptr));
end
// ---- transition properties (use history: skip the cycle after reset) ----
always @(posedge clk) if (!init && !$past(rst)) begin
// push alone when not full adds one; pop alone when not empty removes one
a_push_inc: assert (!($past(push) && !$past(pop) && !$past(full)) || count == $past(count) + 3'd1);
a_pop_dec: assert (!($past(pop) && !$past(push) && !$past(empty)) || count == $past(count) - 3'd1);
// push into a full FIFO (and no pop) is ignored: occupancy unchanged
a_ovf_ign: assert (!($past(push) && !$past(pop) && $past(count) == 3'd4) || count == $past(count));
// simultaneous push and pop (neither empty nor full) leaves occupancy unchanged
a_both: assert (!($past(push) && $past(pop) && !$past(full) && !$past(empty)) || count == $past(count));
end
// ---- data integrity by tracked-slot abstraction (one free slot, one shadow value) ----
// free constant: (* anyconst *) alone is not honoured by the front end, so the
// value is also assumed equal to its one-cycle-old copy (works in both front ends;
// a clocked assume ($stable) lags one step and leaks at the last step: do not use it)
(* anyconst *) wire [1:0] tslot;
reg [1:0] tslot_q;
always @(posedge clk) tslot_q <= tslot;
always @(*) if (!init) assume (tslot == tslot_q);
reg tv; // the tracked slot holds a value we are watching
reg [7:0] td; // the value that should be there
always @(posedge clk) begin
if (rst) tv <= 1'b0;
else begin
if (do_push && wptr == tslot) begin tv <= 1'b1; td <= wdata; end
else if (do_pop && rptr == tslot) tv <= 1'b0;
end
end
always @(*) if (!init && tv && !empty && rptr == tslot) a_data: assert (rdata == td);
// helper invariant needed for the induction step of a_data
generate if (HELPER) begin : g_helper
always @(*) if (!init && tv) a_tracked: assert (mem[tslot] == td);
end endgenerate
// ---- covers: the properties are not vacuous ----
always @(*) if (!init) begin
c_full: cover (count == 3'd4);
c_empty_after_use: cover (empty && tv == 1'b0 && wptr == 2'd2);
end
always @(posedge clk) if (!init) c_both: cover (push && pop && !full && !empty);
`endif
endmoduleThe helper on cue: with bit 7 of the write data cleared, the tracked-slot invariant fails at step 2, so a machine-grown invariant that implied it would fail the same way.
Every unbounded verdict files the invariant found, human or machine, next to the status.
Machine-grown invariants are either explained in the review or replaced by readable helpers that imply them.
Progress fields are quoted with units: frames are counts, memory is bytes, time is seconds on a stated machine.
Recheck an engine-found invariant by asserting it yourself in a fresh run: a genuine invariant proves from reset in the base case and carries the step, while an over-fitted one fails the base on a reachable state. When the machine form is unreadable but load-bearing, keep both: the machine form in the run, the human form asserted alongside as documentation that the tool checks every time.
Not yet: assert a readable helper that implies the property, prove both, and file the readable form. Keep the machine form in the run if it carries something the readable one does not, with that gap stated in the review.