We should use https://github.com/JuliaGPU/GPUToolbox.jl/blob/f971bcdeb2cd8096333549e6563cf03a76459ae2/src/ccalls.jl#L110 for Wait and allreduce
We should use https://github.com/JuliaGPU/GPUToolbox.jl/blob/f971bcdeb2cd8096333549e6563cf03a76459ae2/src/ccalls.jl#L110 for Wait and allreduce