Loom: A Scalable Analytical Neural Computer Architecture

TL;DR

Loom is a scalable neural computer architecture that executes C-compiled programs within a transformer-based system, enabling fixed-cost, program-independent execution of diverse programs.

Contribution

It introduces Loom, a neural architecture with fixed weights that can execute any compiled program efficiently within a transformer framework.

Findings

Loom can execute a Sudoku solver with 284 instructions.

The architecture uses a fixed-size tensor for full machine state.

Source code is publicly available at the provided GitHub URL.

Abstract

We present Loom, a computer architecture that executes programs compiled from C inside a looped transformer whose weights are derived analytically. The architecture implements a 22-opcode instruction set in 8 transformer layers. Each forward pass executes one instruction; the model is applied iteratively until the program counter reaches zero. The full machine state resides in a single tensor $X \in \mathbb{R}^{d \times n}$ of fixed size, and every step has fixed cost for fixed $d$ and $n$, independent of program length or execution history. The default configuration uses $d = 155$ and $n = 1024$, yielding 4.7 million parameters and 928 instruction slots. A compact configuration at $d = 146$ and $n = 512$ suffices for a 9$\times$9 Sudoku solver (284 instructions). The weights are program-independent: programs live in the state tensor, and the same fixed-weight model executes any compiled program. We make Loom source code publicly available at https://github.com/mkturkcan/Loom.