Re: Technical Excellence?

From: Artyom Beilis (artyom.beilis_at_[hidden])
Date: 2024-03-28 08:54:32

>
>
>
> One of the big pushes for C++26 is BLAS. I'll bet if you look hard enough
> you can find a box of FORTRAN punch cards with a reference implementation
> on it. Because it's not new is not valuable? There is still quite a bit of
> research going on in the problem space: https://arxiv.org/abs/2210.10173.
>
>
But why?

There is a standard interface cblas that has several highly efficient
implementations like open source OpenBLAS or proprietary MKL.
The Boost.BLAS is actually quite useless library for real computational
context because it is by order of magnitude slower than stuff like OpenBLAS
(at least when I tested it)

Why should this be a part of C++ standard when there is highly established
infrastructure that does it for you? If so just wrap cblas with nice C++
containers and you are done.

Artyom