Samuel Omlin

> So I am thinking that in the interest of reducing the complexity of KA in usage and to align it better with CUDA.jl I would like to remove the...

Thanks for your comment, the JustSayIt recipes I'm going to share soon (https://github.com/omlins/JustSayItRecipes.jl), allow to work very well with Visual Studio Code. Stay tuned!

@luraess : do you have a fix for this meanwhile?

> Currently, ImplicitGlobalGrid automatically takes care of the grid "chunking" by allocating appropriate arrays on each process and ensuring the arrays have a suitable halo around them. Not quite, ImplicitGlobalGrid...

> I was sceptical that this change would do much without a thorough pass over all of Base @maleadt What you did has already a drastic impact on some kernels!...

@maleadt When i remove the UInt32 casting in the above code, then it introduces some i64 operations. Here is the code: ```julia using CUDA function copy3D!(T2, T, Ci) ix =...

> Well, yeah, because you're introducing Int64s. Or what did you expect? I originally did expect literals to be treated as Int64. However, as in the first example there are...

> So if you're interested in this feature, please contribute by taking your code, running it with CUDA.jl from this PR, inspecting the generated code (e.g. with Cthulhu using @device_code_warntype...

@maleadt: I have run some little test codes with increasing complexity (including some examples from https://github.com/omlins/julia-gpu-course/blob/main/solutions/4_datatransfer_optimisations_advanced_part2.ipynb), always casting literal integers to `Int32`. I have have encountered a first issue. The...

Thank you for your interest! I'm sorry that you ran into these errors, which as you noted are in the dependencies. It looks like for some reason it cannot detect...