Indexing with empty array fails

[DrChainsaw]

Sanity checks (read this first, then remove this section)
I don't know if this works by accident for normal arrays, but I did find it quite useful.

Describe the bug
Trying to index an array with an empty index array produces an error.

To Reproduce
The Minimal Working Example (MWE) for this bug:

julia> aa = randn(3,3);

julia> aa[Int[],1:3]
0×3 Array{Float64,2}

julia> cu(aa)[Int[],1:3]
ERROR: DivideError: integer division error
Stacktrace:
 [1] div at .\int.jl:230 [inlined]
 [2] div at .\div.jl:215 [inlined]
 [3] div at .\div.jl:270 [inlined]
 [4] cld at .\div.jl:227 [inlined]
 [5] configurator at E:\Programs\julia\.julia\packages\CuArrays\e8PLr\src\gpuarrays.jl:27 [inlined]
 [6] #cudacall#216 at E:\Programs\julia\.julia\packages\CUDAnative\ierw8\src\execution.jl:259 [inlined]
 [7] macro expansion at E:\Programs\julia\.julia\packages\CUDAnative\ierw8\src\execution.jl:242 [inlined]
 [8] call(::CUDAnative.HostKernel{GPUArrays.index_kernel,Tuple{CuArrays.CuKernelContext,CUDAnative.CuDeviceArray{Float32,2,CUDAnative.AS.Global},CUDAnative.CuDeviceArray{Float32,2,CUDAnative.AS.Global},Tuple{Int64,Int64},Tuple{CUDAnative.CuDeviceArray{Int64,1,CUDAnative.AS.Global},UnitRange{Int64}}}}, ::CuArrays.CuKernelContext, ::CUDAnative.CuDeviceArray{Float32,2,CUDAnative.AS.Global}, ::CUDAnative.CuDeviceArray{Float32,2,CUDAnative.AS.Global}, ::Tuple{Int64,Int64}, ::Tuple{CUDAnative.CuDeviceArray{Int64,1,CUDAnative.AS.Global},UnitRange{Int64}}; call_kwargs::Base.Iterators.Pairs{Symbol,CuArrays.var"#configurator#50"{Int64},Tuple{Symbol},NamedTuple{(:config,),Tuple{CuArrays.var"#configurator#50"{Int64}}}}) at E:\Programs\julia\.julia\packages\CUDAnative\ierw8\src\execution.jl:219
 [9] (::CUDAnative.HostKernel{GPUArrays.index_kernel,Tuple{CuArrays.CuKernelContext,CUDAnative.CuDeviceArray{Float32,2,CUDAnative.AS.Global},CUDAnative.CuDeviceArray{Float32,2,CUDAnative.AS.Global},Tuple{Int64,Int64},Tuple{CUDAnative.CuDeviceArray{Int64,1,CUDAnative.AS.Global},UnitRange{Int64}}}})(::CuArrays.CuKernelContext, ::Vararg{Any,N} where N; kwargs::Base.Iterators.Pairs{Symbol,CuArrays.var"#configurator#50"{Int64},Tuple{Symbol},NamedTuple{(:config,),Tuple{CuArrays.var"#configurator#50"{Int64}}}}) at E:\Programs\julia\.julia\packages\CUDAnative\ierw8\src\execution.jl:472
 [10] macro expansion at E:\Programs\julia\.julia\packages\CUDAnative\ierw8\src\execution.jl:158 [inlined]
 [11] gpu_call(::CuArrays.CuArrayBackend, ::Function, ::Tuple{CuArray{Float32,2,Nothing},CuArray{Float32,2,Nothing},Tuple{Int64,Int64},Tuple{CuArray{Int64,1,Nothing},UnitRange{Int64}}}, ::Int64; name::Nothing) at E:\Programs\julia\.julia\packages\CuArrays\e8PLr\src\gpuarrays.jl:32
 [12] #gpu_call#1 at E:\Programs\julia\.julia\packages\GPUArrays\QDGmr\src\device\execution.jl:60 [inlined]
 [13] gpu_call at E:\Programs\julia\.julia\packages\GPUArrays\QDGmr\src\device\execution.jl:46 [inlined]
 [14] _unsafe_getindex! at E:\Programs\julia\.julia\packages\GPUArrays\QDGmr\src\host\indexing.jl:125 [inlined]
 [15] _unsafe_getindex(::IndexLinear, ::CuArray{Float32,2,Nothing}, ::Array{Int64,1}, ::UnitRange{Int64}) at .\multidimensional.jl:741
 [16] _getindex at .\multidimensional.jl:727 [inlined]
 [17] getindex(::CuArray{Float32,2,Nothing}, ::Array{Int64,1}, ::UnitRange{Int64}) at .\abstractarray.jl:980
 [18] top-level scope at none:0

Expected behavior
Same behaviour as for the normal array: An empty array with the "correct" dimensions.

Build log

# post the output of Pkg.build()
# make sure the error still reproduces after that.
julia> Pkg.build()
   Building NNlib → `E:\Programs\julia\.julia\packages\NNlib\FAI3o\deps\build.log`

julia> aa = randn(3,3);

julia> cu(aa)[Int[],1:3]
ERROR: DivideError: integer division error
...

Environment details (please complete this section)
Details on Julia:

# please post the output of:
julia> versioninfo()
Julia Version 1.4.0
Commit b8e9a9ecc6 (2020-03-21 16:36 UTC)
Platform Info:
  OS: Windows (x86_64-w64-mingw32)
  CPU: Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-8.0.1 (ORCJIT, haswell)

Julia packages:

• CuArrays.jl:
• CUDAnative.jl:
• ...

CUDA: toolkit and driver version

Additional context
If the right solution is to just add something like any(isempty, indices) && return CuArray(length(indices)...) (I know this example does not work as is) in e.g. ´_unsafe_getindex` I can probably contribute with that :)

[maleadt]

Thanks for the report! Looks like we are calling a kernel with zero threads, and this needs a fix like #259.