Reverse-Engineered Analysis of Apple Neural Engine Architecture Across A11-A18 and M1-M5 Chips
This article presents a reverse-engineered technical analysis of the Apple Neural Engine (ANE), the fixed-function matrix accelerator found in Apple's A-series and M-series chips. Based on direct measurement and static analysis of private runtime, compiler, kernel driver, and firmware, it documents the ANE's datapath, roofline performance bounds, dispatch route below Core ML, compiler and program format, weight-compression scheme, and the kernel driver/firmware/command protocol. Coverage spans A11 through A18 and M1 through M5 families, with claims labeled by evidence type (measured, decompile-derived, or predicted). It notes that while a direct user-space route exists for research and measurement, Core ML remains the only supported path for shipping software.
Key quotes
The Apple Neural Engine (ANE) is the fixed-function matrix accelerator that has shipped in Apple systems-on-chip since the A11-class iPhone and iPad chips and the M1-class Mac chips, exposed to applications only through the Core ML model framework.
This guide reports a reverse-engineered account of the engine, based on direct measurement on Apple silicon and static analysis of the private runtime, compiler, kernel driver, and firmware.
Claims are labeled as measured, decompile-derived, or predicted, and the methodology and open questions are recorded.
The direct route is callable from ordinary user space but remains undocumented, unsupported, and version-fragile; it is intended for measurement, research, and on-device work, not for shipping software, where Core ML remains the supported path.
From the article
You might also wanna read
Apple Announces M5 Chip with Major AI Performance Improvements
Apple has announced its new M5 chip, representing a significant advancement in AI performance and overall chip capabilities. Built using thi
Reverse Engineering Apple's M1 GPU: The Technical Journey to Linux Support on Apple Silicon
The article details the technical journey of reverse-engineering Apple's M1 GPU and developing Linux support for M1 and M2 Macs. It follows
AI-Powered Reverse Engineering of Apple's Rosetta 2 Technology for Linux Systems
This article documents an AI-powered reverse-engineering project called 'attesor' that aims to understand and potentially recreate Apple's R
github.com·4mo ago
Designing a Systolic Array AI Accelerator in Two Weeks for Global Foundries 180nm Tapeout
The article details the author's ambitious project to design a systolic array AI accelerator with in-silicon debug infrastructure from scrat
essenceia.github.io·5mo ago
M5 iPad Pro Neural Accelerator AI Performance Testing Results
The article is a follow-up review of the M5 iPad Pro, specifically testing Apple's claims about 3.5× improvements in local AI processing thr
macstories.net·7mo ago
Running local AI models on an M4 MacBook with 24GB memory: A practical guide
The article details the author's experiments with running local AI language models on an M4 MacBook with 24GB memory. It covers the setup pr
Johanna Larsson·1mo ago
Comments
Sign in to join the conversation.
No comments yet. Be the first.