Dominique
Dominique
There may be a few minor differences in the features supported. I haven't checked lately. The goal here is to support the entire MUMPS API in all four arithmetics. It...
I think working on #47 is more promising than joining with the other package. I just have to find some time.
The move to the native C API and JLL binaries is now complete. As far as I can tell, the other MUMPS package hasn't been updated in the past two...
MUMPS_jll doesn't contain binaries for Windows. https://github.com/JuliaBinaryWrappers/MUMPS_jll.jl/releases/tag/MUMPS-v5.5.1%2B1
Yes. There's an initial experiment here: https://github.com/JuliaSmoothOptimizers/MUMPS.jl/pull/107 Edit: I think the blocker for MUMPS_jll was MPI on Windows, but @amontoison will confirm.
Does the same happen if you don't store the matrix out of core?
I'll try to look into the merger with mumps3.jl in the next few days.
Are you using the same MUMPS library with this package and MUMPS3.jl?
You're right. We should be using `similar`.
Should we extract the Krylov mechanism out of Krylov so it can be used here too?