@crtrott Per our discussion this afternoon, I've fixed the scaled(alpha, A) discussion in the batched BLAS paper. It explains why scaled cannot take an array or mdspan of scaling factors, and therefore that the multiple-scaling-factors case needs overloads that take the scaling factors explicitly.
I've also removed the stub sections and added some links to existing vendor libraries. The paper is ready for submission, though it could still use more work.
@crtrott Per our discussion this afternoon, I've fixed the
scaled(alpha, A)
discussion in the batched BLAS paper. It explains whyscaled
cannot take an array or mdspan of scaling factors, and therefore that the multiple-scaling-factors case needs overloads that take the scaling factors explicitly.I've also removed the stub sections and added some links to existing vendor libraries. The paper is ready for submission, though it could still use more work.