Hi there I have a CUDA program that has a global that has an array as input which contains my constants for running several kernels. If I set them inside the global without the array it runs at 6400ms if I set them from the array it slows right down to about 420000ms. Any ideas? E.g

Slow:-

__global__ void(int* array, int iter){

const int one = array[iter*2+0];

const int two = array[iter*2+1];

}

Fast:-

__global__ void(int* array,int iter){

const int one = 2;

const int two = 1;

}

I havent tried anything yet.