Misprediction Detection
Detect when speculative fetch disagrees with the resolved branch outcome or target.
Task
Assert mispredict if predicted taken differs from actual taken, or if both are taken but predicted target differs from actual target.
Examples
- Complete the required behavior and preserve it under the stated control conditions.
Constraints
- Use synthesizable SystemVerilog unless this is a verification-only stage.
What this problem is testing
A strong answer translates the requirement into synthesizable state and combinational logic, states reset and update semantics precisely, and handles boundary conditions rather than only the nominal case.
Interview habit: before coding, state reset behavior, the state that must be stored, and the conditions that update each output.
Verification plan
- Check reset or initialization before normal operation.
- Exercise the nominal sequence cycle by cycle.
- Test boundaries such as zero, maximum count, wraparound, full/empty, or simultaneous controls when they apply.
- Prove the design holds state when an enable or qualifying condition is absent.
Common mistakes
- Incomplete combinational assignments that infer a latch.
- Blocking/non-blocking assignments that do not match combinational versus sequential intent.
- Parameter-width and boundary errors.
- Passing the happy path while missing reset, overflow, simultaneous operations, or illegal requests.
Interview follow-up questions
- How would you parameterize this design without introducing width or boundary bugs?
- What changes if reset timing or polarity changes?
Concepts to review
branch predictor, recovery, control
Open Misprediction Detection in the HDLForge interactive editor.
Related problems
- 2:1 Multiplexer — Easy
- Parameterized Counter — Easy
- Priority Encoder — Medium
- Synchronous FIFO — Hard
- Shift Register — Easy
Explore the RTL Design Practice & Interview Preparation hub.