KernelAbstractions 0.10 support

[christiangnrd]

JLArrays support done in this way to keep lack of support for scalar indexing.

KA 0.9 support can be easily reverted if not desired

Replaces and closes #595

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic master) to apply these changes.

Click here to view the suggested changes.

diff --git a/lib/JLArrays/src/JLArrays.jl b/lib/JLArrays/src/JLArrays.jl
index 4b238fa..3a3d1e3 100644
--- a/lib/JLArrays/src/JLArrays.jl
+++ b/lib/JLArrays/src/JLArrays.jl
@@ -57,7 +57,7 @@ Adapt.adapt_structure(to::Adaptor, r::Base.RefValue) = JlRefValue(adapt(to, r[])
     Base.sizeof(x::JLDeviceArray) = Base.elsize(x) * length(x)
 
     Base.unsafe_convert(::Type{Ptr{T}}, x::JLDeviceArray{T}) where {T} =
-        convert(Ptr{T}, pointer(x.data)) + x.offset*Base.elsize(x)
+        convert(Ptr{T}, pointer(x.data)) + x.offset * Base.elsize(x)
 
     # conversion of untyped data to a typed Array
     function typed_data(x::JLDeviceArray{T}) where {T}
@@ -340,10 +340,10 @@ end
 ## GPUArrays interfaces
 
 @static if !isdefined(JLArrays.KernelAbstractions, :POCL) # KA v0.9
-    Adapt.adapt_storage(::Adaptor, x::JLArray{T,N}) where {T,N} =
-      JLDeviceArray{T,N}(x.data[], x.offset, x.dims)
+    Adapt.adapt_storage(::Adaptor, x::JLArray{T, N}) where {T, N} =
+        JLDeviceArray{T, N}(x.data[], x.offset, x.dims)
 else
-    function Adapt.adapt_storage(::Adaptor, x::JLArray{T,N}) where {T,N}
+    function Adapt.adapt_storage(::Adaptor, x::JLArray{T, N}) where {T, N}
         arr = typed_data(x)
         Adapt.adapt_storage(POCL.KernelAdaptor([pointer(arr)]), arr)
     end

[maleadt]

Conceptually fine, but CI doesn't look happy. CUDA.jl failure is unrelated, though.
cc @vchuravy

[christiangnrd]

This is ready for final review. I added tests to KA main branch to confirm that this works for both KA 0.9 and 0.10. I don't know if that should stay, and it seems like Windows CI fails with KA 0.10. Not sure if that's my fault or not.

[christiangnrd]

I wasn't able to reproduce the Windows failure locally, so I still don't understand what's happening, but I got JuliaGPU/KernelAbstractions.jl#638 out of it so not a total waste of time.