Sparrow Li (LiYuan)

I have submitted the [pr](https://github.com/rust-num/num-traits/pull/195) in num-trait. Could I ask what this operation does in vek ?( just curious.)

@bluss I am contributing to std::arch to make it a stable feature as soon as possible. I would like to undertake the simd-realization of ndarray. I think we can create...

@bluss Then I hope we create such a crate in rust-ndarray ( instead of a personal crate). So do we need a crate similar to universal intrinsics? Or we can...

I tried to use the simd in the operator overloading of multiplication. [here](https://github.com/SparrowLii/ndarray-simd-test/blob/main/ndarray-sprrowlii/src/impl_ops.rs#L331-L347). And put the usage of avx512f instructions in another [crate](https://github.com/SparrowLii/ndarray-simd-test/tree/main/ndarray-simd) Then a simd test was performed on...

@bluss Could you help pointing out which methods in ndarray should use simd in the first place?

Here is my plan 1. Build a more easy-to-use simd crate based on `stdarch` and `stdsimd` which implements automatic detection of hardware characteristics, doesn't distinguish the vector lengths. 2. Help...

It sounds exciting and I would like to help. Are you planning to create a new eisum crate in rust-ndarray or use the existing one?

[Specialization](https://github.com/rust-lang/rust/issues/31844) can solve DimMax's problems: impl DimMax for D1 { default type Output = D1; } But it seems that there are still many problems to be solved before stability.

Also a little curious. I think the reason why Float does not implement FloatCore is to express that FloatCore is the no_std version of Float, so there will be differences...

No doubt it needs to find a suitable threshold to protect small numbers. And It may be a good practice to look at the implementation in GMP. I did an...