Question

4A. Write a CUDA host program that reads a character type matrix A and integer type matrix B of size mxn. It produces an output string STR such that, every character of A is repeated n times (where n is the integer value in matrix B which is having the same index as that of the character taken in A). Solve this using 1D Block. Example: 1 2 4 3 e X a M Output String STR: pCCaaaaPPPeeXXXXaaaMM 4B. Write the kernel code for Q4A.such that, every column of input matrix must be produced required number of times by one thread. 4M 3M

Answer 1

#include <stdio.h>

__global__void cuda_print(int m, int n, int **x, char **y)
{
   int k=0;
   char *STR;
  
  
for(int i=0;i<m;i++)
{
   for(int j =0;j<n;j++)
   {
       while(yx[i][j]]--)
       {
           STR[k++] = y[i][j];
       }
   }
}
  
printf("%s",STR);
}

int main(void)
{
int m,n;
printf("Enter the values of m and n : ");
scanf("%d %d",&m,&n);
  
char** str = new char*[10000];
for (int i=0;i<10000;i++)
   str[i] = new char[20];
  
int** x = new int*[10000];
for (int i=0;i<10000;i++)
   x[i] = new int[20];

int **d_x;
char **d_str;
char **STR;
  
cudaMalloc(&d_x, m*n*sizeof(int));
cudaMalloc(&d_str, m*n*sizeof(char));
cudaMalloc(&STR, 10000*sizeof(char));

for (int i = 0; i < m; i++) {
   for(int j = 0; j < n; j++ )
       scanf("%d",x[i][j]]);
       scanf("%c",str[i][j]]);
   }
}

cudaMemcpy(d_x, x, m*n*sizeof(int), cudaMemcpyHostToDevice);
cudaMemcpy(d_str, str, m*n*sizeof(char), cudaMemcpyHostToDevice);

// Kernel for device
cuda_print<<<1,1>>>(m, n, d_x, d_str);

cudaFree(d_x);
cudaFree(d_str);
free(x);
free(str);
}