threshold-carrylookahead4bit

4-bit carry lookahead adder. Computes all carries in parallel using generate (G) and propagate (P) signals, avoiding ripple delay.

Circuit

Inputs: A[3:0], B[3:0], Cin (9 inputs)
Outputs: S[3:0], Cout (5 outputs)

For each bit i:
  P_i = A_i XOR B_i   (propagate)
  G_i = A_i AND B_i   (generate)

Carries computed in parallel:
  C1 = G0 + P0·Cin
  C2 = G1 + P1·G0 + P1·P0·Cin
  C3 = G2 + P2·G1 + P2·P1·G0 + P2·P1·P0·Cin
  Cout = G3 + P3·C3

Sum bits:
  S_i = P_i XOR C_i

Truth Table (Examples)

A	B	Cin	S	Cout
0000	0000	0	0000	0
0001	0001	0	0010	0
1111	0001	0	0000	1
1111	1111	1	1111	1

Binary: A + B + Cin = (Cout << 4) | S

Architecture

Component	Neurons
P/G generation	8
Carry lookahead	8
Sum XORs	8

Total: 24 neurons, 108 parameters, 4 layers

Advantage Over Ripple Carry

Ripple carry: O(n) delay as each carry waits for previous Carry lookahead: O(1) delay for carry computation (parallel)

For 4 bits: CLA computes all carries simultaneously.

Usage

from safetensors.torch import load_file

w = load_file('model.safetensors')

# Verify: all 512 input combinations (16 x 16 x 2) produce correct sums

Files

threshold-carrylookahead4bit/
├── model.safetensors
├── create_safetensors.py
├── config.json
└── README.md

License

MIT