MatterAI Documentation | AI Super Intelligence for Engineering

Verilog Anti-Patterns Overview

Blocking Assignments in Sequential Logic

// Anti-pattern: Using blocking assignments in sequential logic
module counter_bad(input clk, input reset, output reg [3:0] count);
  always @(posedge clk or posedge reset) begin
    if (reset) begin
      count = 0;  // Blocking assignment
    end else begin
      count = count + 1;  // Blocking assignment
    end
  end
endmodule

// Better approach: Use non-blocking assignments for sequential logic
module counter_good(input clk, input reset, output reg [3:0] count);
  always @(posedge clk or posedge reset) begin
    if (reset) begin
      count <= 0;  // Non-blocking assignment
    end else begin
      count <= count + 1;  // Non-blocking assignment
    end
  end
endmodule

Always use non-blocking assignments (<=) for sequential logic within an always @(posedge clk) block. Using blocking assignments (=) in sequential logic can lead to race conditions and simulation-synthesis mismatches.

Non-Blocking Assignments in Combinational Logic

// Anti-pattern: Using non-blocking assignments in combinational logic
module mux_bad(input a, input b, input sel, output reg y);
  always @(*) begin
    if (sel) begin
      y <= a;  // Non-blocking assignment
    end else begin
      y <= b;  // Non-blocking assignment
    end
  end
endmodule

// Better approach: Use blocking assignments for combinational logic
module mux_good(input a, input b, input sel, output reg y);
  always @(*) begin
    if (sel) begin
      y = a;  // Blocking assignment
    end else begin
      y = b;  // Blocking assignment
    end
  end
endmodule

Use blocking assignments (=) for combinational logic within an always @(*) block. Using non-blocking assignments (<=) in combinational logic can lead to simulation-synthesis mismatches and incorrect behavior.

Incomplete Sensitivity Lists

// Anti-pattern: Incomplete sensitivity list
module alu_bad(input [7:0] a, input [7:0] b, input [1:0] op, output reg [7:0] result);
  always @(a or b) begin  // Missing op in sensitivity list
    case (op)
      2'b00: result = a + b;
      2'b01: result = a - b;
      2'b10: result = a & b;
      2'b11: result = a | b;
    endcase
  end
endmodule

// Better approach: Use complete sensitivity list or @(*)
module alu_good(input [7:0] a, input [7:0] b, input [1:0] op, output reg [7:0] result);
  always @(a or b or op) begin  // Complete sensitivity list
    case (op)
      2'b00: result = a + b;
      2'b01: result = a - b;
      2'b10: result = a & b;
      2'b11: result = a | b;
    endcase
  end
endmodule

// Even better: Use @(*) for automatic sensitivity list
module alu_best(input [7:0] a, input [7:0] b, input [1:0] op, output reg [7:0] result);
  always @(*) begin  // Automatic sensitivity list
    case (op)
      2'b00: result = a + b;
      2'b01: result = a - b;
      2'b10: result = a & b;
      2'b11: result = a | b;
    endcase
  end
endmodule

Always use complete sensitivity lists or the automatic sensitivity list @(*) for combinational logic. Incomplete sensitivity lists can lead to simulation-synthesis mismatches and incorrect behavior.

Latches Due to Incomplete Assignments

// Anti-pattern: Incomplete assignments creating latches
module fsm_bad(input clk, input reset, input in, output reg out);
  reg [1:0] state, next_state;
  
  always @(*) begin
    case (state)
      2'b00: begin
        if (in) next_state = 2'b01;
        // Missing else clause creates a latch for next_state
      end
      2'b01: begin
        next_state = 2'b10;
        out = 1'b1;  // Missing assignment in other states creates a latch for out
      end
      2'b10: begin
        next_state = 2'b00;
      end
      // Missing default case
    endcase
  end
  
  always @(posedge clk or posedge reset) begin
    if (reset) state <= 2'b00;
    else state <= next_state;
  end
endmodule

// Better approach: Complete assignments to avoid latches
module fsm_good(input clk, input reset, input in, output reg out);
  reg [1:0] state, next_state;
  
  always @(*) begin
    // Default assignments to avoid latches
    next_state = state;
    out = 1'b0;
    
    case (state)
      2'b00: begin
        if (in) next_state = 2'b01;
        // else clause not needed due to default assignment
      end
      2'b01: begin
        next_state = 2'b10;
        out = 1'b1;
      end
      2'b10: begin
        next_state = 2'b00;
      end
      default: begin
        next_state = 2'b00;
      end
    endcase
  end
  
  always @(posedge clk or posedge reset) begin
    if (reset) state <= 2'b00;
    else state <= next_state;
  end
endmodule

Always provide default assignments to all variables at the beginning of combinational always blocks, and include a default case in case statements. Incomplete assignments can create unintended latches in your design.

Multiple Drivers for the Same Signal

// Anti-pattern: Multiple drivers for the same signal
module multiple_drivers_bad(input clk, input reset, input en, output reg [7:0] count);
  always @(posedge clk or posedge reset) begin
    if (reset) count <= 8'h00;
    else if (en) count <= count + 1;
  end
  
  always @(posedge clk) begin
    if (count == 8'hFF) count <= 8'h00;  // Second driver for count
  end
endmodule

// Better approach: Single driver for each signal
module multiple_drivers_good(input clk, input reset, input en, output reg [7:0] count);
  always @(posedge clk or posedge reset) begin
    if (reset) count <= 8'h00;
    else if (en) begin
      if (count == 8'hFF) count <= 8'h00;
      else count <= count + 1;
    end
  end
endmodule

Ensure that each signal has only one driver. Multiple drivers for the same signal can lead to simulation-synthesis mismatches and undefined behavior.

Mixing Blocking and Non-Blocking Assignments

// Anti-pattern: Mixing blocking and non-blocking assignments
module mixed_assignments_bad(input clk, input reset, input [7:0] data_in, output reg [7:0] data_out);
  reg [7:0] temp;
  
  always @(posedge clk or posedge reset) begin
    if (reset) begin
      temp <= 8'h00;
      data_out <= 8'h00;
    end else begin
      temp = data_in + 8'h01;  // Blocking assignment
      data_out <= temp;  // Non-blocking assignment
    end
  end
endmodule

// Better approach: Consistent assignment types
module mixed_assignments_good(input clk, input reset, input [7:0] data_in, output reg [7:0] data_out);
  reg [7:0] temp;
  
  always @(posedge clk or posedge reset) begin
    if (reset) begin
      temp <= 8'h00;
      data_out <= 8'h00;
    end else begin
      temp <= data_in + 8'h01;  // Non-blocking assignment
      data_out <= temp;  // Non-blocking assignment
    end
  end
endmodule

Avoid mixing blocking and non-blocking assignments within the same always block. Use non-blocking assignments consistently for sequential logic and blocking assignments consistently for combinational logic.

Asynchronous Reset Release

// Anti-pattern: Asynchronous reset release
module async_reset_release_bad(input clk, input reset, output reg [3:0] count);
  always @(posedge clk or posedge reset) begin
    if (reset) count <= 4'h0;
    else count <= count + 1;
  end
endmodule

// Better approach: Synchronous reset release
module sync_reset_release_good(input clk, input reset, output reg [3:0] count);
  reg reset_sync1, reset_sync2;
  
  // Synchronize reset release
  always @(posedge clk or posedge reset) begin
    if (reset) begin
      reset_sync1 <= 1'b1;
      reset_sync2 <= 1'b1;
    end else begin
      reset_sync1 <= 1'b0;
      reset_sync2 <= reset_sync1;
    end
  end
  
  // Use synchronized reset
  always @(posedge clk or posedge reset) begin
    if (reset) count <= 4'h0;
    else if (!reset_sync2) count <= count + 1;
  end
endmodule

Synchronize asynchronous reset release to avoid metastability issues. While asynchronous reset assertion is often used for reliable system initialization, the release should be synchronized to the clock to prevent timing violations.

Combinational Loops

// Anti-pattern: Combinational loop
module comb_loop_bad(input a, input b, output reg c);
  reg temp;
  
  always @(*) begin
    temp = a & c;  // c depends on temp, and temp depends on c
    c = temp | b;
  end
endmodule

// Better approach: Break the combinational loop
module comb_loop_good(input a, input b, input clk, output reg c);
  reg temp;
  
  // Use a register to break the loop
  always @(posedge clk) begin
    temp <= a & c;
  end
  
  always @(*) begin
    c = temp | b;
  end
endmodule

Avoid creating combinational loops in your design. Combinational loops can lead to oscillations, unstable behavior, and synthesis issues. Break loops using registers or redesign your logic.

Inferring Clock Gating

// Anti-pattern: Inferring clock gating
module clock_gate_bad(input clk, input enable, input [7:0] data_in, output reg [7:0] data_out);
  wire gated_clk;
  
  assign gated_clk = clk & enable;  // Inferred clock gating
  
  always @(posedge gated_clk) begin
    data_out <= data_in;
  end
endmodule

// Better approach: Use enable within the always block
module clock_gate_good(input clk, input enable, input [7:0] data_in, output reg [7:0] data_out);
  always @(posedge clk) begin
    if (enable) data_out <= data_in;
  end
endmodule

// Or use proper clock gating cells provided by the technology library
module clock_gate_best(input clk, input enable, input [7:0] data_in, output reg [7:0] data_out);
  wire gated_clk;
  
  // Instantiate a proper clock gating cell
  CKLNQD1 clock_gate_cell (.CP(clk), .E(enable), .TE(1'b0), .Q(gated_clk));
  
  always @(posedge gated_clk) begin
    data_out <= data_in;
  end
endmodule

Avoid inferring clock gating using simple logic gates. Use enable signals within sequential blocks or instantiate proper clock gating cells provided by your technology library.

Using Initial Blocks for Hardware

// Anti-pattern: Using initial blocks for hardware initialization
module initial_block_bad(input clk, input [7:0] data_in, output reg [7:0] data_out);
  reg [7:0] memory [0:15];
  
  initial begin
    // Initialize memory
    for (int i = 0; i < 16; i = i + 1) begin
      memory[i] = 8'h00;
    end
    data_out = 8'h00;
  end
  
  always @(posedge clk) begin
    data_out <= memory[data_in[3:0]];
  end
endmodule

// Better approach: Use reset for initialization
module initial_block_good(input clk, input reset, input [7:0] data_in, output reg [7:0] data_out);
  reg [7:0] memory [0:15];
  integer i;
  
  always @(posedge clk or posedge reset) begin
    if (reset) begin
      // Initialize memory
      for (i = 0; i < 16; i = i + 1) begin
        memory[i] <= 8'h00;
      end
      data_out <= 8'h00;
    end else begin
      data_out <= memory[data_in[3:0]];
    end
  end
endmodule

Avoid using initial blocks for hardware initialization, as they are generally not synthesizable for hardware (except for some specific cases like memory initialization). Use reset signals to initialize your design.

Ignoring Clock Domain Crossing

// Anti-pattern: Ignoring clock domain crossing
module cdc_bad(input clk_a, input clk_b, input [7:0] data_a, output reg [7:0] data_b);
  always @(posedge clk_b) begin
    data_b <= data_a;  // Direct crossing between clock domains
  end
endmodule

// Better approach: Use proper synchronization
module cdc_good(input clk_a, input clk_b, input [7:0] data_a, output reg [7:0] data_b);
  reg [7:0] data_sync1, data_sync2;
  
  // Double-flop synchronizer
  always @(posedge clk_b) begin
    data_sync1 <= data_a;
    data_sync2 <= data_sync1;
    data_b <= data_sync2;
  end
endmodule

// Even better for multi-bit signals: Use handshaking or FIFO
module cdc_best(input clk_a, input reset_a, input clk_b, input reset_b,
               input [7:0] data_a, input valid_a, output reg ready_a,
               output reg [7:0] data_b, output reg valid_b, input ready_b);
  // Simplified handshaking CDC (a complete implementation would be more complex)
  reg valid_a_sync1, valid_a_sync2;
  reg valid_b_toggle;
  reg valid_b_toggle_sync1, valid_b_toggle_sync2;
  reg valid_b_toggle_prev;
  reg [7:0] data_capture;
  
  // Capture data in source domain
  always @(posedge clk_a or posedge reset_a) begin
    if (reset_a) begin
      valid_b_toggle <= 1'b0;
      ready_a <= 1'b1;
      data_capture <= 8'h00;
    end else begin
      if (valid_a && ready_a) begin
        data_capture <= data_a;
        valid_b_toggle <= ~valid_b_toggle;
        ready_a <= 1'b0;
      end else if (valid_b_toggle_sync2 != valid_b_toggle) begin
        ready_a <= 1'b1;
      end
    end
  end
  
  // Synchronize to destination domain
  always @(posedge clk_b or posedge reset_b) begin
    if (reset_b) begin
      valid_a_sync1 <= 1'b0;
      valid_a_sync2 <= 1'b0;
      valid_b_toggle_sync1 <= 1'b0;
      valid_b_toggle_sync2 <= 1'b0;
      valid_b_toggle_prev <= 1'b0;
      valid_b <= 1'b0;
      data_b <= 8'h00;
    end else begin
      valid_a_sync1 <= valid_b_toggle;
      valid_a_sync2 <= valid_a_sync1;
      valid_b_toggle_sync1 <= valid_b_toggle;
      valid_b_toggle_sync2 <= valid_b_toggle_sync1;
      
      if (valid_b_toggle_sync2 != valid_b_toggle_prev && !valid_b) begin
        data_b <= data_capture;
        valid_b <= 1'b1;
        valid_b_toggle_prev <= valid_b_toggle_sync2;
      end else if (valid_b && ready_b) begin
        valid_b <= 1'b0;
      end
    end
  end
endmodule

Always use proper synchronization techniques when crossing clock domains. For single-bit signals, use a double-flop synchronizer. For multi-bit signals, use handshaking protocols, gray coding, or asynchronous FIFOs.

Improper Use of Delays

// Anti-pattern: Using delays for hardware timing
module delay_bad(input clk, input [7:0] data_in, output reg [7:0] data_out);
  always @(posedge clk) begin
    #2 data_out <= data_in;  // Delay is not synthesizable
  end
endmodule

// Better approach: Let the synthesis tool handle timing
module delay_good(input clk, input [7:0] data_in, output reg [7:0] data_out);
  always @(posedge clk) begin
    data_out <= data_in;
  end
endmodule

Avoid using delays (#) in synthesizable code. Delays are for simulation only and are ignored during synthesis. Let the synthesis tool handle timing based on your timing constraints.

Ignoring Tool Warnings

// Anti-pattern: Code that generates tool warnings
module warning_generator(input clk, input reset, input [7:0] data, output reg valid);
  always @(posedge clk or posedge reset) begin
    if (reset) valid <= 1'b0;
    else if (data == 8'hFF) valid <= 1'b1;
    // Warning: valid is not assigned in all branches
  end
endmodule

// Better approach: Fix the issues causing warnings
module warning_free(input clk, input reset, input [7:0] data, output reg valid);
  always @(posedge clk or posedge reset) begin
    if (reset) valid <= 1'b0;
    else if (data == 8'hFF) valid <= 1'b1;
    else valid <= valid;  // Explicitly maintain current value
  end
endmodule

Don’t ignore warnings from your synthesis and simulation tools. Warnings often indicate potential issues in your design that could lead to unexpected behavior. Fix the issues causing warnings or document why they can be safely ignored.

Not Using Parameters for Constants

// Anti-pattern: Hardcoded constants
module hardcoded_bad(input clk, input reset, output reg [3:0] count);
  always @(posedge clk or posedge reset) begin
    if (reset) count <= 4'h0;
    else if (count == 4'h9) count <= 4'h0;
    else count <= count + 1;
  end
endmodule

// Better approach: Use parameters for constants
module parameterized_good #(
  parameter WIDTH = 4,
  parameter MAX_COUNT = 9
)(input clk, input reset, output reg [WIDTH-1:0] count);
  always @(posedge clk or posedge reset) begin
    if (reset) count <= {WIDTH{1'b0}};
    else if (count == MAX_COUNT) count <= {WIDTH{1'b0}};
    else count <= count + 1'b1;
  end
endmodule

Use parameters for constants in your design. This makes your code more maintainable, reusable, and easier to modify. Parameters can be overridden during module instantiation, allowing for flexible designs.

Not Using Named Port Connections

// Anti-pattern: Positional port connections
module top_bad();
  wire clk, reset, enable;
  wire [7:0] data_in, data_out;
  
  // Positional port connections are error-prone
  counter counter_inst(clk, reset, enable, data_in, data_out);
endmodule

// Better approach: Use named port connections
module top_good();
  wire clk, reset, enable;
  wire [7:0] data_in, data_out;
  
  // Named port connections are clearer and less error-prone
  counter counter_inst(
    .clk(clk),
    .reset(reset),
    .enable(enable),
    .data_in(data_in),
    .data_out(data_out)
  );
endmodule

Use named port connections when instantiating modules. Named connections are clearer, less error-prone, and allow for port reordering without breaking your design.

Not Using Generate Blocks

// Anti-pattern: Duplicated code for similar structures
module ripple_carry_bad(input [3:0] a, input [3:0] b, input cin, output [3:0] sum, output cout);
  wire c1, c2, c3;
  
  full_adder fa0(.a(a[0]), .b(b[0]), .cin(cin), .sum(sum[0]), .cout(c1));
  full_adder fa1(.a(a[1]), .b(b[1]), .cin(c1), .sum(sum[1]), .cout(c2));
  full_adder fa2(.a(a[2]), .b(b[2]), .cin(c2), .sum(sum[2]), .cout(c3));
  full_adder fa3(.a(a[3]), .b(b[3]), .cin(c3), .sum(sum[3]), .cout(cout));
endmodule

// Better approach: Use generate blocks
module ripple_carry_good #(
  parameter WIDTH = 4
)(input [WIDTH-1:0] a, input [WIDTH-1:0] b, input cin, output [WIDTH-1:0] sum, output cout);
  wire [WIDTH:0] c;
  
  assign c[0] = cin;
  assign cout = c[WIDTH];
  
  genvar i;
  generate
    for (i = 0; i < WIDTH; i = i + 1) begin : adder_loop
      full_adder fa(
        .a(a[i]),
        .b(b[i]),
        .cin(c[i]),
        .sum(sum[i]),
        .cout(c[i+1])
      );
    end
  endgenerate
endmodule

Use generate blocks for repetitive structures in your design. Generate blocks allow for parameterized, scalable designs and reduce code duplication.

Not Using Assertions

// Anti-pattern: No assertions for verification
module fsm_no_assertions(input clk, input reset, input in, output reg out);
  reg [1:0] state, next_state;
  
  // State transition logic
  always @(*) begin
    next_state = state;
    case (state)
      2'b00: if (in) next_state = 2'b01;
      2'b01: next_state = 2'b10;
      2'b10: next_state = 2'b00;
      default: next_state = 2'b00;
    endcase
  end
  
  // Output logic
  always @(*) begin
    case (state)
      2'b01: out = 1'b1;
      default: out = 1'b0;
    endcase
  end
  
  // State register
  always @(posedge clk or posedge reset) begin
    if (reset) state <= 2'b00;
    else state <= next_state;
  end
endmodule

// Better approach: Use assertions for verification
module fsm_with_assertions(input clk, input reset, input in, output reg out);
  reg [1:0] state, next_state;
  
  // State transition logic
  always @(*) begin
    next_state = state;
    case (state)
      2'b00: if (in) next_state = 2'b01;
      2'b01: next_state = 2'b10;
      2'b10: next_state = 2'b00;
      default: next_state = 2'b00;
    endcase
  end
  
  // Output logic
  always @(*) begin
    case (state)
      2'b01: out = 1'b1;
      default: out = 1'b0;
    endcase
  end
  
  // State register
  always @(posedge clk or posedge reset) begin
    if (reset) state <= 2'b00;
    else state <= next_state;
  end
  
  // Assertions
  // Check that state is always valid
  property valid_state;
    @(posedge clk) disable iff (reset)
      state inside {2'b00, 2'b01, 2'b10};
  endproperty
  assert property (valid_state) else $error("Invalid state detected");
  
  // Check state transition from state 01
  property state_01_transition;
    @(posedge clk) disable iff (reset)
      (state == 2'b01) |=> (state == 2'b10);
  endproperty
  assert property (state_01_transition) else $error("Invalid transition from state 01");
  
  // Check that output is high only in state 01
  property output_high_only_in_state_01;
    @(posedge clk) disable iff (reset)
      (out == 1'b1) |-> (state == 2'b01);
  endproperty
  assert property (output_high_only_in_state_01) else $error("Output high in wrong state");
endmodule

Use assertions to verify the correctness of your design. Assertions can catch bugs early in the design process and serve as documentation for expected behavior.

Introduction

Getting Started

Cortex Features

Patterns

Integrations

Enterprise

Verilog