Vincent Barrielle

Latest micro-optimizations in `sprs-ldl` (see #207) have put its performance on-par with the original C implementation of LDL (see https://github.com/PTNobel/sprs-performance-test/pull/2). I guess the way to go now would be to...

The binding of SuiteSparse's CAMD has enabled some more performance gains, and, as mentionned in https://github.com/PTNobel/sprs-performance-test/pull/2#issuecomment-706390315, I don't think there's much more improvements to be hoped for using LDL. A...

I'll need to investigate this issue. I've definitely encountered the type resolution recursion issue (see https://github.com/rust-lang/rust/issues/82779) while developping `d2f0da7`. This error happens because I've tried to minimize the number of...

Hello, sorry for the silence, things are a bit complicated here and I don't have time to work on this right now (and probably for the next month as well)....

> Proposed change: > > * modify the interface of `mul_acc_mat_vec_{csr,csc}(mat, in_vec, res_vec)` such that `in_vec` and `res_vec` are `ArrayBase` instead of slices; > > * add an `impl Mul...

Now that you mentionned it, I recall I've written a trait to express that a datatype is a vector of a given dimension: https://docs.rs/sprs/0.6.2/sprs/vec/trait.VecDim.html It's probably a good idea to...

Hi @toastronics, sorry for answering so late. > Would changing the `impl`'s of `VecDim` to be like `DenseVector`, with explicit `impl`'s for > > * `&[N]`, > * `Vec`, >...

Hello @adinapoli-mndc, thanks for the kind words, I'm really glad this library suits you. You should be able to use https://docs.rs/sprs/0.6.5/sprs/binop/fn.mul_mat_same_storage.html to compute the Hadamard product, though I must admit...

There's a caveat though: I implemented this function a long time ago, and I'm not sure it's as performant as it could be.

For your use case you want a diagonal iterator I guess, you'll be more efficient since you will avoid allocating. But the PR you made is already a good start!