Francesc Verdugo

Hi @GeliezaK I realize that function `constant_prolongator_with_block_size` is not the easiest way of achieving our goal. An alternative is to transform the `node_to_aggregate` to `dof_to_aggregate` (with function below) and pass...

Hi @GeliezaK , yet another edit. I don't think we need a "constant prolongator" with block size > 1. I ralized this after looking at Algorithm 7, page 46, here:...

NB. When qr is mentioned in Algorithm 7, it is really a "thin" QR decomposition. See section for rectangular matrices here: https://en.wikipedia.org/wiki/QR_decomposition This discussion is relevant on how to compute...

> Thank you for your comments! From Algorithm 7 I understand that the information in node_to_aggregate would be sufficient to compute the tentative prolongator. You are correct, but note that...

> Besides, do you have an idea how to circumvent the QR decomposition if the number of rows of the near null space B is smaller than its number of...

@GeliezaK @amartinhuertas Regarding the bug discussed on the meeting on 2024-08-15. This works for me with PartitionedArrays v0.5.2: ```julia using PartitionedArrays nodes_per_dir = (4,4,4) parts_per_dir = (2,2,2) p = prod(parts_per_dir)...

@GeliezaK Can you please send your minimal reproducer?

I can reproduce the error. I know hot to fix it.

Fixed in JuliaLang/julia#166

Take a look how the dof partition is build from a node partition here: https://github.com/fverdugo/PartitionedArrays.jl/blob/11ded45c2a486dee20f63ca69c0b4b767038a7b0/src/gallery.jl#L389 Maybe you can even use the same code, or just calling the `node_to_dof_partition` function.