Assorted internet references:
- dougallj/aarch64_amx.py - IDA (disassembler) and Hex-Rays (decompiler) plugin for Apple AMX
- dougallj/amx.h - amx simulator and hardware tests
- amx-rs - Rust bindings and simulator, includes ASM trick and genlut framework
- https://www.realworldtech.com/forum/?threadid=187087&curpostid=187120 - Instruction names
- https://news.ycombinator.com/item?id=24464807#24472004 - Instruction names
- include/mach/arm/_structs.h - Register file
- https://nod.ai/comparing-apple-m1-with-amx2-m1-with-neon/ - "AMX2"
- Google for
AMX_STATE_T_EL1
Patents:
| Number | Title | Expires |
|---|---|---|
| US20180074824A1 | Outer Product Engine | Abandoned |
| US10831488B1 | Computation engine with extract instructions to minimize memory access | 2038-10-31 |
| US20190129719A1 | Matrix Computation Engine (2017) | 2038-01-14 |
| US10592239B2 | Matrix computation engine (2019) | 2037-11-01 (anticipated) |
| US10877754B2 | Matrix computation engine (2020) | 2037-11-01 (anticipated) |
| US10754649B2 | Computation engine that operates in matrix and vector modes (2018) | 2038-10-19 |
| US11042373B2 | Computation engine that operates in matrix and vector modes (2020) | 2038-07-24 (anticipated) |
| US10970078B2 | Computation engine with upsize/interleave and downsize/deinterleave options | 2038-05-25 |
| US10642620B2 | Computation engine with strided dot product (2018) | 2038-07-24 |
| US10990401B2 | Computation engine with strided dot product (2020) | 2038-04-05 (anticipated) |
Academic papers:
- Fast polynomial multiplication using matrix multiplication accelerators with applications to NTRU on Apple M1/M3 SoCs - "Our work repurposes AMX to implement polynomial multiplication and applies it to the NTRU cryptosystem, setting new speed records on the Apple M1 and M3 systems-on-chip (SoCs)."
- PQC-AMX: Accelerating Saber and FrodoKEM on the Apple M1 and M3 SoCs - "In this paper, we leverage AMX to accelerate the post-quantum lattice-based cryptosystems Saber and FrodoKEM, and benchmark their performance on Apple M1 and M3 SoCs."