add parallel

src/radix_sort_par.cu

@@ -47,16 +47,31 @@ void countSort(int arr[], int n, int exp)
 // The main function to that sorts arr[] of size n using  
 // Radix Sort
 
-void radixsort(int arr[], int n) 
+void radixsort(int *arr, int n) 
 { 
+    int *d_arr;
+
     // Find the maximum number to know number of digits 
     int m = getMax(arr, n); 
   
     // Do counting sort for every digit. Note that instead 
     // of passing digit number, exp is passed. exp is 10^i 
     // where i is current digit number 
+
+    // allocate device memory
+    cudaMalloc((void**)&d_arr,sizeof(int)*n);
+
+    cudaMemcpy(d_arr, arr, sizeof(int)*n,cudaMemcpyHostToDevie);
     for (int exp = 1; m/exp > 0; exp *= 10) 
-        countSort<<1,32>>(arr, n, exp); 
+        countSort<<<1,32>>>(d_arr, n, exp); 
+
+    //transfer data back to host memory
+    cudaMemcpy(arr, d_arr, sizeof(int)*n, cudaMemcpyDeviceToHost);
+
+    //deallocate device memory
+    cudaFree(d_arr);
+
+
 } 
   
 // A utility function to print an array 
@@ -89,13 +104,12 @@ timespec diff(timespec start, timespec end)
 }
 
 // Driver program to test above functions 
-int main() 
+int main(int argc, char *argv[]) 
 { 
     timespec start, stop;
     
     int n;
-    cout<<"Masukkan nilai N\n";
-    cin>>n;
+    n= atoi(argv[1]);
     int arr[n];
     rng(arr,n);
     clock_gettime(CLOCK_REALTIME, &start);
@@ -107,5 +121,7 @@ int main()
     long time = duration.tv_sec * 1000000 + duration.tv_nsec/1000;
     printf("\n%d.%09d s\n", duration.tv_sec, duration.tv_nsec);
     
+    //deallocate host memory
+    free(arr);
     return 0; 
 } 
\ No newline at end of file