Most failed reviews in this course fail on one cycle: the consequent checked a cycle too early or too late, the attempt started on a level instead of an edge, the stability ended one wait too soon, or the window was written where a counter belonged. This topic works each error on a small stage or on the bridge, always as a pair: the wrong-cycle form failing and the right-cycle form passing on the same design.
Lesson | After it you can... |
|---|---|
Check a consequent in the same cycle or the next, and pair each offset with the mutant that exactly one form catches. | |
Start one attempt per edge instead of one per high cycle, and read the attempt table of a failure. | |
Assume stability on inputs, assert it on outputs, and hold data for the whole stall instead of one cycle. | |
Build within-the-last-N checks from history registers and one-to-one checks from counters. |
Suggested order: top to bottom. The offset pair first, because everything else assumes it; then attempts, because windows count attempts; then stability across a stall; then windows and counters, which combine all three. The safety-versus-liveness topic then reuses the window machinery for bounded response.