Mark Gates

`copy` is an odd name and description for an in-place operation. From Intel's docs, "A transposition operation can be a _normal matrix copy_, a transposition, a conjugate transposition, or just...

@rileyjmurray Respectfully, I disagree. If Reference BLAS / LAPACK is a (de facto) standard, it should be reasonably self-consistent. `imatcopy` and `omatcopy` don't follow the BLAS naming scheme in any...

It doesn't appear that those `ge_trans` and `ge_copy` were ever implemented. They aren't in the reference BLAS and LAPACK. However, LAPACKE does implement an out-of-place [`LAPACKE_zge_trans`](https://github.com/Reference-LAPACK/lapack/blob/master/LAPACKE/utils/lapacke_zge_trans.c), among other [transpose functions](https://github.com/Reference-LAPACK/lapack/tree/master/LAPACKE/utils)....

An easy fix should be `-DSCALAPACK_LIBRARIES=none`. The testers won't be able to compare to a reference solution, but should otherwise work. We don't actually need `-lscalapack` on Frontier, but I'm...

Did it work with a different ROCm version, when you closed the issue? What version of SLATE (`git log --oneline -n 1`)? I presume you've done `git submodule update` so...

What version of SLATE, e.g., `git log --oneline -n 1`? What version of HIP / ROCm, e.g., `hipcc --version`?

This appears to be an issue with an older version of ROCm, 5.2 (circa July 2022). I don't have ready access to ROCm 5.2, and was unable to reproduce it...

To be honest, this PR is too large to review. It needs to be broken into smaller, manageable chunks. Even then, with such a large change, it should be coordinated...

`skewsymm` and `skewhemm` make a lot of sense to me — very clear (assuming you already know `symm` and `hemm`), whereas `kymm` and `kemm` are pretty cryptic (s**K**ew-s**Y**mmetric, s**K**ew-h**E**rmitian).

Regarding skew-symmetric vs. skew-Hermitian, `ky` and `ke` could work for skew-symmetric and skew-Hermitian, but `kb`, `kp`, `kt` would conflict between skew-symmetric and skew-Hermitian: `sp` and `hp` => `kp` `sb` and...