cospectrum / microgemm

General matrix multiplication with custom configuration in Rust. Supports no_std and no_alloc environments.
Apache License 2.0
6 stars 1 forks source link

Add accelerated f64 kernels #9

Open ratchief opened 11 months ago

ratchief commented 11 months ago

Implement Wasm/Neon kernels for f64

cospectrum commented 10 months ago

I don't see much difference in performance between Generic kernels and WasmSimd128Kernel, so it will probably be deprecated. But NeonKernel<f64> can be implemented, for sure.