ARM64 assembly implementation of linear regression using the NEON SIMD features for vectorization. Mainly as exercise in asm programming, for experimentation and learning. The vectorization speedup is measurable, on my M2 Macbook Air the the SIMD implementation is ~2x faster than the naive C implementation.
- Wikipedia: https://en.wikipedia.org/wiki/Simple_linear_regression
- Howard Oakley's ARM Assembly articles, in particular "Lanes and loads in NEON" https://eclecticlight.co/2021/08/23/code-in-arm-assembly-lanes-and-loads-in-neon/
- Alexander von Below's Hello Silicon repo https://github.com/below/HelloSilicon
- Raymond Chen's ARM64's series https://devblogs.microsoft.com/oldnewthing/20220726-00/?p=106898
Feedback Bugs/Suggestions/Comments/Feedback to: Email: LG@THLG.NL