Guardado en:
Detalles Bibliográficos
Autor principal: Turkcan, Mehmet Kerem
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2604.08816
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910118350684160
author Turkcan, Mehmet Kerem
author_facet Turkcan, Mehmet Kerem
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2604_08816
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Loom: A Scalable Analytical Neural Computer Architecture
Turkcan, Mehmet Kerem
Machine Learning
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.
title Loom: A Scalable Analytical Neural Computer Architecture
topic Machine Learning
url https://arxiv.org/abs/2604.08816