C++ Logo


Advanced search

Re: [SG14] [EXTERNAL] Re: Linear algebra library proposal

From: Hoemmen, Mark <mhoemme_at_[hidden]>
Date: Tue, 18 Jun 2019 21:07:31 +0000
Hi Arthur!

On 6/18/19, 1:08 PM, "Arthur O'Dwyer" <arthur.j.odwyer_at_[hidden]<mailto:arthur.j.odwyer_at_[hidden]>> wrote:
> That is, rather than let the software architecture be dictated by political concerns ("what design is LWG most likely to accept"), I think it would be better to drive it based on technical concerns ("what design is most adoptable for actual programmers and does it satisfy their performance needs"). You can't really do that without an implementation.

The “LWG most likely to accept” does quite get at the right point – sorry for wording it that way. We do think “adding a lot of customization points” is a significant _technical_ concern. Subcommittees would be right to bring that up.

The idea is to “standardize existing practice.” This is a C++ binding for the BLAS. The BLAS has been around since the 1970’s. It has a Standard that has been around since 2002. We have an existing library (kokkos-kernels) a few years old that wraps the BLAS with essentially this design, and other libraries (e.g., Teuchos::BLAS) with a less advanced design but the same functionality that we’ve been using in practice since around 2004. Coauthors have designed and implemented similar libraries.


Received on 2019-06-18 16:09:22