Kernel with dynamic parallelism seems to be calling CPU functions

Specific Domains GPU
2

I think you’re using the wrong synchronization call. Try device_synchronize(): CUDA.jl/src/device/intrinsics/dynamic_parallelism.jl at e561e7a106684f8e4be59cad98a51cc304c671d2 · JuliaGPU/CUDA.jl · GitHub