Igor Baratta

I had a similar problem running hipSYCL with CUDA backend together with Eigen. And I confirm that > #include > #ifdef SYCL_DEVICE_ONLY > #undef SYCL_DEVICE_ONLY > #endif > #include worked...

Using `init_mpi` instead of `init_petsc`, I get the following error message: ``` An error occurred in MPI_Neighbor_alltoall *** reported by process [305790977,0] *** on communicator MPI COMMUNICATOR 4 CREATE FROM...

We've also added a few improvements in different bits of the code, for instance collision detection and tabulation, to make interpolation faster.

To avoid computing the dual graph in serial we can call a custom partitioner, which sets the destination of all cells to process 0 (in serial): ```python from mpi4py import...

> @IgorBaratta could a Python version be added? Yes, I can do that. And once we have the interpolation operator (different meshes) we could also do matrix free geometric mutigrid.

``` |||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||| ||||||||||| (1) (2) (3) ``` Currently (1) and (2) are mixed, but it would be good to group so one could have a range for (2) with...

> Can you add some more explanation? Currently it produces some definitions such as: ```c++ #define tabulate_cell_a tabulate_tensor_integral_0cc1940dcb4d02004a8f4689364c2ebf7bc61339 #define dim_cell_a0 3 #define dim_cell_a1 3 #define tabulate_cell_L tabulate_tensor_integral_7d5e3ad2bad43b64846403d43f3f8ac26aa57e73 #define dim_cell_L0 3...

Yes. I think another flag in `compute_form_data` would do.

Minimal reproducible example: ```python import dolfinx from mpi4py import MPI comm=MPI.COMM_WORLD msh = dolfinx.mesh.create_interval(comm, 10, points=(0.0, 1.0)) V = dolfinx.fem.FunctionSpace(msh, ("Lagrange", 1)) sols = list() for i in range(2048): print(i)...

I think a more verbose alternative could be: ```python V = dolfinx.fem.FunctionSpace(msh, ("Lagrange", 1)) map = V.dofmap.index_map scatterer = dolfinx.common.Scatterer(map, bs) # add this wrapper for i in range(2048): x...