RFC: add BLAS Level 1 sdot routine

[splrk]

Checklist

Please ensure the following tasks are completed before submitting a feature request.

• Read and understood the Code of Conduct.
• Searched for existing issues and pull requests.
• The issue name begins with RFC:.

Description

This RFC proposes implementing the dot (or inner) product routine for a single precision floats as defined in BLAS Level 1 routines. Specifically adding @stdlib/blas/base/sdot is proposed.

Related Issues

Does this feature request have any related issues?

No.

Questions

Any questions for reviewers?

No.

Other

Any other information relevant to this feature request? This may include screenshots, references, sample output, and/or implementation notes.

No.

[splrk]

sdsdot and dsdot are sdot and ddot that

Compute the inner product of two vectors with extended precision accumulation and result.

That is a quote from the LAPACK sdsdot documentation documentation.

From the implementation it looks like they will run a little slower, but give you a little more accurate result. Maybe these routines would be a good place to experiment with pairwise summation (no. 328 in develop\TODO) instead of the LAPACK implementation?

(Sorry about adding them at first. I wrote the RFC then realized it would probably be better to write and RC per routine rather than group them together)

[kgryte]

@splrk Thanks for the clarification! Yes, agreed. We should also add sdsdot and dsdot, each under its own RFC.

Regarding pairwise summation, we can probably hold off on this for the time being. Thanks for having a look over the TODOs!