Passing Symbol as a an argument fails

[vchuravy]

After JuliaGPU/GPUCompiler.jl#650
we should be able to pass Symbols as arguments.

This fails in

CUDA.jl/src/compiler/compilation.jl

Line 284 in 860eb88

param_usage = sum(sizeof, argtypes)

due to Julia telling us the wrong thing...

I guess we will need to pass them as Ptr{Cvoid} equivalent.

 ERROR: LoadError: Type Symbol does not have a definite size.
 Stacktrace:
   [1] sizeof
     @ ./essentials.jl:780 [inlined]
   [2] _mapreduce(f::typeof(sizeof), op::typeof(Base.add_sum), ::IndexLinear, A::Vector{DataType})
     @ Base ./reduce.jl:440
   [3] _mapreduce_dim(f::Function, op::Function, ::Base._InitialValue, A::Vector{DataType}, ::Colon)
     @ Base ./reducedim.jl:337
   [4] mapreduce
     @ ./reducedim.jl:329 [inlined]
   [5] _sum
     @ ./reducedim.jl:987 [inlined]
   [6] sum(f::Function, a::Vector{DataType})
     @ Base ./reducedim.jl:983
   [7] compile(job::GPUCompiler.CompilerJob)
     @ CUDA ~/.julia/packages/CUDA/BZS7Q/src/compiler/compilation.jl:284
   [8] actual_compilation(cache::Dict{Any, CUDA.CuFunction}, src::Core.MethodInstance, world::UInt64, cfg::GPUCompiler.CompilerConfig{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}, compiler::typeof(CUDA.compile), linker::typeof(CUDA.link))
     @ GPUCompiler ~/.julia/packages/GPUCompiler/8Yisz/src/execution.jl:237
   [9] cached_compilation(cache::Dict{Any, CUDA.CuFunction}, src::Core.MethodInstance, cfg::GPUCompiler.CompilerConfig{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}, compiler::Function, linker::Function)
     @ GPUCompiler ~/.julia/packages/GPUCompiler/8Yisz/src/execution.jl:151
  [10] macro expansion
     @ ~/.julia/packages/CUDA/BZS7Q/src/compiler/execution.jl:380 [inlined]
  [11] macro expansion
     @ ./lock.jl:273 [inlined]
  [12] cufunction(f::typeof(gpu_apply_collision_kernel), tt::Type{Tuple{KernelAbstractions.CompilerMetadata{KernelAbstractions.NDIteration.DynamicSize, KernelAbstractions.NDIteration.DynamicCheck, Nothing, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}, KernelAbstractions.NDIteration.NDRange{1, KernelAbstractions.NDIteration.DynamicSize, KernelAbstractions.NDIteration.StaticSize{(64,)}, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}, Nothing}}, CUDA.CuDeviceVector{Float64, 1}, CUDA.CuDeviceMatrix{Float64, 1}, CUDA.CuDeviceMatrix{Float64, 1}, Int64, Float64, Symbol}}; kwargs::@Kwargs{always_inline::Bool, maxthreads::Int64})