README.md

---
license: mit
tags:
- pytorch
- safetensors
- threshold-logic
- neuromorphic
- arithmetic
- alu
---

# threshold-alu4bit

4-bit Arithmetic Logic Unit as threshold circuit. Performs arithmetic and bitwise operations on two 4-bit operands.

## Circuit

```
A[3:0]  ──┐
B[3:0]  ──┼──► ALU ──┬──► R[3:0]  (result)
op[1:0] ──┘         ├──► Cout    (carry out)
                    └──► Zero    (zero flag)
```

## Operations

| op | Operation | Description |
|----|-----------|-------------|
| 00 | ADD | R = A + B |
| 01 | SUB | R = A - B (2's complement) |
| 10 | AND | R = A & B |
| 11 | XOR | R = A ^ B |

## Flags

- **Cout**: Carry out from addition, borrow (inverted) from subtraction
- **Zero**: Set when result is 0

## Architecture

| Component | Count | Neurons |
|-----------|-------|---------|
| Full Adders | 4 | 28 |
| B XOR (for SUB) | 4 | 12 |
| Bitwise AND | 4 | 4 |
| Bitwise XOR | 4 | 12 |
| Output MUX4 | 4 | 20 |
| Zero detect | 1 | 1 |
| Cout MUX | 1 | 5 |

**Total: 82 neurons, 282 parameters, 5 layers**

## How Subtraction Works

```
A - B = A + (~B) + 1  (2's complement)

When op=01:
- B bits are XORed with 1 (inverted)
- Carry-in is set to 1
- Result is A - B
```

## Usage

```python
from safetensors.torch import load_file

w = load_file('model.safetensors')

# Examples:
# op=0: 5 + 3 = 8, cout=0, zero=0
# op=1: 5 - 3 = 2, cout=1, zero=0
# op=2: 5 & 3 = 1, cout=0, zero=0
# op=3: 5 ^ 3 = 6, cout=0, zero=0
```

## Files

```
threshold-alu4bit/
├── model.safetensors
├── create_safetensors.py
├── config.json
└── README.md
```

## License

MIT