Umberto Villa

@tzanio The very first implementation of MINRES for MFEM was a direct porting of Saunder’s Matlab code: https://web.stanford.edu/group/SOL/software/minres/ There indeed they used the unpreconditioned norm of the residual as stopping...

Hi @huidong80, Is the number of iterations decreasing as the mesh gets finer? This is definitely an interesting behavior, however it is often the case to observe some mild dependence...

I am using `pygalmesh` to generate meshes for FEniCS 2019.1. Here's my code snippet, where `data` is the 3D numpy array and `h` is the list of voxels sizes in...

Hi there, I know this issue has already been discussed in several threads, but I would like to understand better the rationale behind using `MPI_Comm_dup` when wrapping the MPI comm...

Fixes #208

@garth-wells and @chrisrichardson : Thank you for the discussion. I want to take the lead if no one is already working on this yet.

I tried to write a simple code to reproduce the issue. Unfortunately, I did not observe the MPI Comm duplication issue, but I can see some memory leakage. ``` from...

@jorgensd : Thank you for your prompt reply. I tried using the docker container as you recommended. The increase in memory is much less significant than using conda but still...

@jorgensd Thank you very much for your help with this. I was not aware of that issue with PETSc Vec. This explains a lot of strange things happening in my...

This is in fact what I am proposing. `dolfinx.la.Vector` should expose a property `.vector` implemented as the current implementation of `dolfinx.fem.Function`. Then, to preserve backward compatibility, implement `dolfinx.fem.Function.vector` as ```...