-
I found very little documentation about this.
https://nlopt.readthedocs.io/en/stable/NLopt_Reference/#preconditioning-with-approximate-hessians
But that only explains the C interface, and says t…
-
- [ ] merge first matrix-based implementation, #5753
- [ ] add tests/benchmarks
- [ ] Run a large scaling test
- [ ] Add AMG for Schur complement block
- [ ] Experiment with skipping inner S solv…
tjhei updated
1 month ago
-
I saw in simple testing a ~12% increase in performance just by scaling
```julia
function scale_ls!(A)
s = ones(size(A, 2))
for j = 1 : size(A, 2)
i = A.colptr[j]
k = A.colp…
-
problem:
1) `PreconditionerAMG` uses the system matrix for the `solve()` in `MGCoarseKrylov`.
2) currently, not following #692 leads to divergence in some of my tests on Fritz.
Here some details:…
-
I am aware that EUCLID is no longer supported by hypre team, but I find it useful to use it as a preconditioner and in particular with IJ interface. I have realised that EUCLID does not support matric…
-
This is to track a subset of #15979 namely the direct solvers from Amesos2, the iterative solvers from Belos and the preconditioners from Ifpack2 (and maybe other packages)
All these packages provi…
-
@johannes-resch for setting up a diagonal preconditioner, the following code lines should help:
https://github.com/MeltPoolDG/MeltPoolDG-dev/blob/846a3c6086a558674afd4bd2f9c11f9231f433e0/source/lin…
-
Hi,
Could you add information about the Preconditioner format.
If one use the Wikipeida convention ([The Preconditioned Conjugate Gradient Method](https://en.wikipedia.org/wiki/Conjugate_gradien…
-
One of the method I coded in Rust is the "Conjugate gradient solver for positive-definite matrices." I see that you already coded this method. I didn't compare them yet because there's a small differe…
-
All Krylov solvers are designed to take a preconditioner. If no preconditioner is present, an explicit copy of the Krylov vector is invoked. We see this can introduce significant overhead. E.g. for th…