8 comments

[ 3.1 ms ] story [ 31.3 ms ] thread
More precisely, it's an MLP library.

My wishlist for Rust in scientific programming:

- A really nice ndarray library interfacing with BLAS/LAPACK, like NumPy - A nice CUDA library for wrapping the C API - A Jupyter/IPython kernel - and of course, a whole bunch of people using Rust for scientific programming.

In sum, everything you already have in Python.
Everything except execution times that are remotely acceptable.
I've been working slowly on getting solid BLAS/LAPACK abstractions.
Have you already looked into Julia? It has a lot of these things (if not all of them; I'm not sure about the ndarray bit).
Would rust be much faster than Julia for these kinds of things?

And are compile languages even suited for notebook style programming?

I think BLAS/LAPACK are getting old, I know of cases where they got their lunch eaten by JIT'd VMs.

Much of stats and linear alg is embarrassingly vectorizable. I would love to see modern SIMD implementations (preferably written in RUST) :)

Aw, you beat me to finishing! Nice work, I was trying to learn rust at the same time as writting it.

github.com/bfrog/tin