[ACCEPTED]-CUDA: Calling a device function from a kernel-cuda

Accepted answer

Score: 24

CUDA actually inlines all functions by default 8 (although Fermi and newer architectures 7 do also support a proper ABI with function 6 pointers and real function calls). So your 5 example code gets compiled to something 4 like this

__global__ void Kernel(int *ptr)
{
    if(threadIdx.x<2)
        if(ptr[threadIdx.x]==threadIdx.x)
            ptr[threadIdx.x]++;
}

Execution happens in parallel, just 3 like normal code. If you engineer a memory 2 race into a function, there is no serialization 1 mechanism that can save you.

Source: stackoverflow.com