Chiel van Heerwaarden

UCX incompatible with CUDA.jl memory pool

4

I am not sure whether this is a MPI.jl issue or something from our local supercomputer, but I have a failing `Alltoall` in my Julia code, whereas the identical code...

c-indexing and dimension ordering in `attach_scale` function

2

I have implemented a simple example adding dimensions to an HDF5 file with the code below. I noticed that I need to swap the dimensions to c-ordering and use c-indexing...

LoopVectorization returning zeros in @tturbo accelerated region

3

I have written a macro that generates a fast loop for interpolations. Without `@tturbo` and with `Threads.@threads` the code runs fine, but it fails with `@tturbo` which does not do...

Why default tolerance at 0 in indexing on axis coordinate value?

1

I was playing around AxisArrays for the first time and did this: ``` using AxisArrays x = 0:10:100 |> collect a = AxisArray(rand(length(x)); x=x) println(a[x=29..33]) println(a[x=30..30]) println(a[atvalue(29; atol=5)]) println(a[atvalue(29)]) #...

I opened an issue here following the discourse discussion: https://discourse.julialang.org/t/how-to-control-threads-in-combination-of-loopvectorization-and-spawn/75466 In short, I would like to be able to run two tasks with each an own thread in my code...

Make minimum value k depend on floating point precision

7

Make the minimum value of k in the two-stream kernels relative to the precision of the floating point numbers. This makes the solver work in single precision with sufficient accuracy.

Link to HDF plugin documentation is broken, and I cannot retrieve the correct information from the HDF documentation.