#machine-learning #simd #vectorized #function #mathml #ml-model


SIMD vectorized implementations of various math functions used in ML models

9 releases (breaking)

0.10.0 May 25, 2024
0.8.0 Apr 29, 2024
0.6.0 Mar 31, 2024
0.1.0 Dec 31, 2023

#455 in Machine learning

Download history 88/week @ 2024-02-14 100/week @ 2024-02-21 163/week @ 2024-02-28 151/week @ 2024-03-06 135/week @ 2024-03-13 202/week @ 2024-03-20 719/week @ 2024-03-27 359/week @ 2024-04-03 266/week @ 2024-04-10 64/week @ 2024-04-17 278/week @ 2024-04-24 229/week @ 2024-05-01 126/week @ 2024-05-08 518/week @ 2024-05-15 342/week @ 2024-05-22 195/week @ 2024-05-29

1,198 downloads per month
Used in 5 crates (via rten)




This crate provides portable SIMD types that abstract over SIMD intrinsics on different architectures. Unlike std::simd this works on stable Rust. There is also functionality to detect the available instructions at runtime and dispatch to the optimal implementation.

This crate also contains SIMD-vectorized versions of math functions such as exp, erf, tanh, softmax etc. that are performance-critical in machine-learning models.


SIMD-vectorized implementations of various math functions that are commonly used in neural networks.

For each function in this library there are multiple variants, which typically include:

  • A version that operates on scalars
  • A version that reads values from an input slice and writes to the corresponding position in an equal-length output slice. These have a vec_ prefix.
  • A version that reads values from a mutable input slice and writes the computed values back in-place. These have a vec_ prefix and _in_place suffix.

All variants use the same underlying implementation and should have the same accuracy.

See the source code for comments on accuracy.