Q048FreeDesign Verification

Verify matrix zero propagation

Difficulty: Medium
Topic: Reference Models
Language: SV

Interview prompt

Question

Implement a two-pass predictor and two-entry tagged scoreboard for a matrix sanitizer that zeroes every row and column containing an original zero without cascading from newly written zeroes. R, C, and W are positive constants; input matrix elements are known unsigned W-bit values. Observed response elements retain four-state logic so unknown output corruption is rejected. Call the supplied scoreboard methods only for accepted transactions using serialized callbacks. For same-edge ID retirement and reuse, publish the response first. Reset flushes pending IDs; a delayed old response using an already reused ID requires an external generation or quarantine rule to distinguish it.

Matrix example showing how a zero marks its row and column and produces the transformed matrix — A zero propagates across its complete row and column while unmarked cells retain their original values.

Candidate starting point

Implementation scaffold

class matrix_predictor #(int R = 4, C = 4, W = 8);
  typedef logic [W-1:0] matrix_t[R][C];

  static function void predict(
    input  matrix_t original,
    output matrix_t expected,
    output bit row_zero[R],
    output bit col_zero[C]
  );
  // TODO: Implement this predictor or scoreboard operation.
endfunction
endclass

class matrix_scoreboard #(int R = 4, C = 4, W = 8);
  typedef logic [W-1:0] matrix_t[R][C];
  typedef struct {
    matrix_t original;
    matrix_t expected;
    bit row_zero[R];
    bit col_zero[C];
  } expectation_t;
  expectation_t pending[int unsigned];

  function new();
    if (R < 1 || C < 1 || W < 1)
      $fatal(1, "matrix dimensions and element width must be positive");
  endfunction

  function void accept_request(int unsigned id, matrix_t original);
  // TODO: Implement this predictor or scoreboard operation.
endfunction

  function void reset_epoch();
  // TODO: Implement this predictor or scoreboard operation.
endfunction

  function void check_response(int unsigned id, matrix_t actual);
  // TODO: Implement this predictor or scoreboard operation.
endfunction
endclass

// TODO: Explain the no-zero, single/intersecting/edge-zero, full-row/full-column, all-zero and non-square test cases.

Reviewed example

Trace one case

Input

matrix=[[1,2,0],[4,5,6]]

Expected output

sanitized=[[0,0,0],[4,5,0]]

Only original row 0 and original column 2 are marked; zeroes written during transformation do not cascade into row 1.

What to cover

Requirements

Deep-copy each accepted matrix and derive row and column masks exclusively from the untouched original snapshot.
Match up to two outstanding responses by ID, compare every coordinate, and diagnose the first mismatch with the original matrix and both masks.
Reject duplicate or unknown responses and flush accepted work on reset.
Explain why no-zero, single-zero, intersecting, edge, full-row, full-column, all-zero, and non-square cases are useful follow-up tests.

Continue practicing

Q103 · Reference ModelsVerify unique zero-sum tripletsMedium Q147 · Reference ModelsScoreboard a maximum-interval acceleratorHard Q197 · Reference ModelsMonitor an independently stalled binary adderHard

class matrix_predictor #(int R = 4, C = 4, W = 8); typedef logic [W-1:0] matrix_t[R][C]; static function void predict( input matrix_t original, output matrix_t expected, output bit row_zero[R], output bit col_zero[C] ); // TODO: Implement this predictor or scoreboard operation. endfunction endclass class matrix_scoreboard #(int R = 4, C = 4, W = 8); typedef logic [W-1:0] matrix_t[R][C]; typedef struct { matrix_t original; matrix_t expected; bit row_zero[R]; bit col_zero[C]; } expectation_t; expectation_t pending[int unsigned]; function new(); if (R < 1 || C < 1 || W < 1) $fatal(1, "matrix dimensions and element width must be positive"); endfunction function void accept_request(int unsigned id, matrix_t original); // TODO: Implement this predictor or scoreboard operation. endfunction function void reset_epoch(); // TODO: Implement this predictor or scoreboard operation. endfunction function void check_response(int unsigned id, matrix_t actual); // TODO: Implement this predictor or scoreboard operation. endfunction endclass // TODO: Explain the no-zero, single/intersecting/edge-zero, full-row/full-column, all-zero and non-square test cases.

Question

Implementation scaffold

Trace one case

Requirements

Related questions

Question

Implementation scaffold

Trace one case

Requirements

Related questions