CuPy - atomicAdd error

[Hari Haran]

Hi, I tried to use atomic add, but encountered an error. Given below is my CuPy and kernel code.

import cupy as cp

kernel_code = '''

extern "C" __global__

void form_duty_network_adjacency_list_kernel(int *input, int *output, int size, int ct) {

     

    int dp=threadIdx.x + blockIdx.x*blockDim.x;

    if(dp<size && input[dp]==true)

    {

        output[ct]=dp;

        atomicAdd(&ct,1);

    }

    __syncthreads();

}

'''

dn_adj_kernel = cp.RawKernel(kernel_code, 'form_duty_network_adjacency_list_kernel')

adj_list=[]

for i in range(len(duty_network_matrix)):

    print("Currently processing DP = ",i)

    output_array=cp.empty(len(duty_network_matrix[i]),dtype=cp.int32)

    input_array = cp.array(np.array(duty_network_matrix[i]),dtype=cp.int32)

    block=(1,)

    grid=(len(duty_network_matrix),)

    local_ctr=0

    dn_adj_kernel(grid,block,(input_array,output_array,len(duty_network_matrix),local_ctr))

    output_array_cpu = output_array.get()

    output_array=None

    adj_list.append(output_array_cpu)

Please do help.

Thanks