CORAL
Core Rust Architecture for Linear Algebra. This is a pure-Rust implementation of BLAS routines
that is optimized and uses NEON AArch64 intrinsics for speed and has no dependencies. However, it is scarily unsafe, as it works
with pointers and unsafe simd.
While it is faster and more comparable with OpenBLAS and faer, it is not
that far off from the fully-safe and portable-simd coral
implementation. For these reasons, unless you currently need double precision or complex routines, or
need an extra 10GFLOP/s on SGEMM, I highly suggest using the safe implementation.
If you really need speed, on AArch64, and do not mind very unsafe code, only then I suggest using
coral-aarch64.
Here are some benchmarks.