10. Splitting Vector

• Concept: Designs are built as a hierarchy of modules. A higher-level module instantiates lower-level ones (like PCB → IC → flip-flop).
• Each module definition is standalone, not nested.
• Hierarchy enables modularity, reuse, and scalability.
• Hierarchical names use dot . notation: top.u1.u2.signal.

• Syntax:

  module <name> #(parameter ... ) (port list);
      // declarations
      // functionality
  endmodule

• Modules describe behavior, structure, or both.
• Can contain:
  • Ports (input, output, inout)
  • Nets / regs
  • Continuous assignments
  • Procedural blocks (initial/always)
• macromodule is an alias for module (rarely used).

• Declaration: always starts with module … endmodule.

• Instantiation:

  mod_name inst_name (.port1(sig1), .port2(sig2)); // named mapping
  mod_name inst_name (sig1, sig2);                 // ordered mapping

• Multiple instances allowed in the same module.

1. By order: Ports matched in the order declared.

   adder u1 (a, b, sum);

   ⚠️ Risky: Errors occur if order changes.

2. By name: Ports explicitly named.

   adder u1 (.x(a), .y(b), .z(sum));

   ✅ Recommended: Safer against future changes.

• input – read-only inside the module.
• output – driven inside, observed outside.
• inout – bidirectional (used in buses like I²C, data buses).
• Defaults & caveats:
  • If type not specified, ports default to net type wire.
  • Must explicitly declare reg if procedural assignment needed.

  • Example:

    module m(input clk, input [7:0] data, output reg q, inout bus);

Four states:

• 0 → logic zero
• 1 → logic one
• z/Z → high impedance (tri-stated)
• x/X → unknown (simulation only!)

Hardware reality: Only 0, 1, Z exist. X indicates indeterminate logic (conflict, uninitialized, timing issue) in simulation.

NOT (~)

• ~0 = 1, ~1 = 0
• ~x = x, ~z = x

AND (&)

• If any input is 0 → result = 0
• If all inputs are 1 → result = 1
• Otherwise → x

OR (|)

• If any input is 1 → result = 1
• If all inputs are 0 → result = 0
• Otherwise → x

XOR (^)

• If both inputs known → normal XOR (1 if odd # of 1’s)
• If any input is x/z → result = x

Derived

• NAND = NOT(AND)
• NOR = NOT(OR)
• XNOR = NOT(XOR)

💡 Quick Memory Aid

• AND is “pessimistic” → 0 dominates.
• OR is “optimistic” → 1 dominates.
• XOR is “suspicious” → any x/z poisons result.
• NOT just flips 0/1, turns unknowns into x.

Declaring Vectors

• Syntax:

  <type> [msb:lsb] <name>;

  • msb = most significant bit index
  • lsb = least significant bit index

• Example:

  wire [7:0] data_bus;   // 8-bit wide bus (data_bus[7]..data_bus[0])
  reg  [15:0] acc;       // 16-bit register
  

• Indexing direction:
  • [7:0] → descending (bit 7 = MSB, bit 0 = LSB) — most common.
  • [0:7] → ascending (bit 0 = MSB, bit 7 = LSB) — allowed, but less used.
  • Both are legal in IEEE 1364-2005.

⚠️ Gotcha: A vector’s indexing direction matters in part-selects and concatenations.

Bit Selection

• Access individual bits:

  data_bus[0]   // LSB
  acc[15]       // MSB

• Index must be constant in synthesis (simulation allows variable indices, but not synthesizable in many tools).

Part Selection

• Extract a subset of contiguous bits.

• Syntax:

  vector[msb:lsb]   // fixed part-select
  vector[base +: width]   // indexed part-select (2001+)
  vector[base -: width]

• Examples:

  data_bus[7:4]   // upper nibble of 8-bit bus
  acc[11:8]       // 4 bits from acc
  
  // Indexed part-select (IEEE 1364-2001 & 2005)
  acc[3 +: 4]     // acc[6:3]  (4 bits starting at index 3, ascending)
  acc[7 -: 4]     // acc[7:4]  (4 bits descending from 7)

✅ Indexed part-select avoids confusion when ranges flip.

Concatenation {}

• Combines multiple signals or vectors into a larger one.

• Syntax:

  {expr1, expr2, expr3}

• Examples:

  wire [3:0] a = 4'b1010;
  wire [3:0] b = 4'b1100;
  wire [7:0] c;
  
  assign c = {a, b};  // c = 1010_1100

• Replication operator:
  • {N{expr}} repeats an expression N times.

  • Example:

    {4{1'b1}}    // 1111 (4-bit all ones)
    {2{a}}       // {a,a} → duplicates a vector

Mixed Concatenation

• Scalars and vectors can mix:

  {a[3:0], 2'b11, b[1]}   // concatenates 4+2+1 = 7 bits

• Concatenation result width = sum of operand widths.

Assignments & Truncation

• If RHS wider than LHS → truncated (MSBs dropped).

• If RHS narrower → zero-filled (if unsigned) or sign-extended (if signed).

  reg [3:0] x;
  x = 8'hAB;   // x = 4'hB  (truncated)

Vector Operations

• Bitwise operators apply bit-by-bit:
  • & | ^ ~ ^~ ~^

  • Example:

    assign out = a & b;   // AND each bit

• Reduction operators collapse vector → single bit:
  • &a, |a, ^a, ~&a, ~|a, ~^a

  • Example:

    wire parity = ^data_bus;  // XOR all bits → parity

• Shift operators:
  • a << n, a >> n (logical shift, fill with zeros).
  • >>> (arithmetic right shift, preserves sign bit for signed values).

  • Example:

    reg signed [7:0] s = -8'd8;  // 1111_1000
    $display("%b", s >>> 2);     // 1111_1110 = -2

Corner Cases & Pitfalls

• Out-of-range index: Returns x (unknown).

  wire [3:0] a = 4'b1010;
  $display("%b", a[10]);  // x

• Empty concatenation: Illegal in Verilog-2005.
• Mix signed/unsigned: Width/sign extension rules can surprise. Always cast explicitly if needed.

• Format:
  • <size>: number of bits
  • <base>: b (binary), o (octal), d (decimal), h (hex)
  • <value>: digits

• Examples:

  4'b1010   // 4-bit binary
  8'hFF     // 8-bit hex
  'd25      // 32-bit decimal (unsized default)

Sized vs Unsized

• Sized: Width is explicit → 4'b1010
• Unsized: Defaults to 32 bits → 'hF = 32-bit hex F.

Defaults

• Base: Decimal if unspecified.
• Case: Base (d/h/o/b) and digits are case-insensitive.
• Readability: _ allowed → 16'b1010_1100_1111_0000.

• Preferred syntax:

  -8'd6    // 8-bit signed decimal -6
  4'shF    // signed 4-bit hex = -1

• Note: 8'd-6 is illegal.
• Signed constants ('sd) are sign-extended when assigned.

• Decimal: 12.34, 0.1
• Scientific: 1.2e3, 4.5E-2
• Must have digit before and after decimal point.
• _ can be used for readability: 236.123_763_e-12
• Follows IEEE 754 double precision.