For any of the examples below, simply use a different GPU array and AcceleratedKernels.jl will pick the right backend:

# Intel Graphics
using oneAPI
v = oneArray{Int32}(undef, 100_000)             # Empty array

# AMD ROCm
using AMDGPU
v = ROCArray{Float64}(1:100_000)                # A range converted to Float64

# Apple Metal
using Metal
v = MtlArray(rand(Float32, 100_000))            # Transfer from host to device

# NVidia CUDA
using CUDA
v = CuArray{UInt32}(0:5:100_000)                # Range with explicit step size

# Transfer GPU array back
v_host = Array(v)