Luc Berger
Luc Berger
Yes, I will review the patch and make it into a PR as well as add unit-test to make sure this does not creep back up.
Hi @oliverchang our most prominent users are [Trilinos](https://trilinos.github.io/) and [PETSc](https://petsc.org/). I am looking into the CLA now, let's see if my organization finds it appropriate.
Yes, that's how the solver is being exercised in this case.
I am having a look at this
@jjellio the approach you are proposing seem to imply that the user should enable ROCm (or HIP whatever we call it) and hope the TPLs are in standard places and...
Yes, this was actually done with PR #11750 just above your comment, we can probably just close this
@jjellio what you say is correct but ignores the fact that NVIDIA is packaging it's libraries differently from AMD which is really at the source of some of these differences....
Actually the saner thing to do would be to stop turning on cublas and cusparse in my opinion and make them work like rocblas and rocsparse. Also note that Kokkos...
@jjellio if you want to have a quick chat about this I'm happy to set up a discussion with you and we can find a reasonable approach. My only requirements...
Yes, I think we can, sorry we took a while to make the final decision on how to move forward. The code changes in GEMM led to issues with CUDA...