$ cat t269.cu
#include <stdio.h>

struct matl1
{
    static const double cond;
};

const double matl1::cond = 420.5;

__global__ void kernel(matl1* d_matl1)
{
double cond = d_matl1->cond;
printf("cond = %lf\n", cond);
}


int main(){

  matl1 * h_matl1 = (matl1*)malloc(sizeof(matl1));
  matl1 * d_matl1;
  cudaMalloc((void**)&d_matl1, sizeof(matl1));
  cudaMemcpy(d_matl1, h_matl1, sizeof(matl1), cudaMemcpyHostToDevice);
  kernel<<<1,1>>>(d_matl1);
  cudaDeviceSynchronize();
  return 0;
}
$ nvcc -arch=sm_20 -o t269 t269.cu
$ ./t269
cond = 420.500000
$