cuda: practice: first easy exercises

labs/cuda/practice/exercises/mergesort/.gitignore

+mergesort

labs/cuda/practice/exercises/mergesort/Makefile

+EXEC := $(notdir $(CURDIR))
+
+include ../../../Makefile

labs/cuda/practice/exercises/mergesort/mergesort.cu

+#include <stdio.h>
+
+#define NUM_ELEMENTS 16
+
+__device__ void merge(int* arr, int* temp, int left, int middle, int right) {
+  int i = left;
+  int j = middle;
+  int k = left;
+
+  while (i < middle && j < right) {
+    // TODO 1: Get the smaller element
+  }
+
+  while (i < middle) {
+    // TODO 2: Copy any remaining elements from the left subarray
+  }
+  while (j < right) {
+    // TODO 3: Copy any remaining elements from the right subarray
+  }
+
+  for (int x = left; x < right; x++) {
+    // TODO 4: Copy the sorted elements
+  }
+}
+
+__global__ void mergeSortGPU(int* arr, int* temp, int n, int width) {
+  int tid = threadIdx.x + blockDim.x * blockIdx.x;
+  int left = tid * width;
+  int middle = left + width / 2;
+  int right = left + width;
+
+  if (left < n && middle < n) {
+    merge(arr, temp, left, middle, right);
+  }
+}
+
+void generateData(int* data, int size) {
+  srand(time(0));
+
+  for (int i = 0; i < size; i++) {
+    data[i] = rand() % 14 + 1;
+  }
+}
+
+int main() {
+  int* array = NULL;
+  array = (int*)malloc(NUM_ELEMENTS * sizeof(int));
+  generateData(array, NUM_ELEMENTS);
+
+  printf("Original Array: ");
+  for (int i = 0; i < NUM_ELEMENTS; i++) {
+    printf("%d ", array[i]);
+  }
+  printf("\n");
+
+  int *d_array, *d_tmp;
+  cudaMalloc(&d_array, NUM_ELEMENTS * sizeof(int));
+  cudaMemcpy(d_array, array, NUM_ELEMENTS * sizeof(int),
+             cudaMemcpyHostToDevice);
+  cudaMalloc(&d_tmp, NUM_ELEMENTS * sizeof(int));
+
+  int block_size = 256;
+  size_t blocks_no = NUM_ELEMENTS / block_size;
+
+  if (NUM_ELEMENTS % block_size) ++blocks_no;
+
+  for (int width = 1; width < NUM_ELEMENTS; width *= 2) {
+    // TODO 0: Call the mergeSortGPU kernel with the appropriate arguments
+  }
+  cudaDeviceSynchronize();
+
+  cudaMemcpy(array, d_array, NUM_ELEMENTS * sizeof(int),
+             cudaMemcpyDeviceToHost);
+  cudaFree(d_array);
+
+  printf("Sorted Array: ");
+  for (int i = 0; i < NUM_ELEMENTS; i++) {
+    printf("%d ", array[i]);
+  }
+  printf("\n");
+
+  return 0;
+}

labs/cuda/practice/exercises/oets/.gitignore

+oets

labs/cuda/practice/exercises/oets/Makefile

+EXEC := $(notdir $(CURDIR))
+
+include ../../../Makefile

labs/cuda/practice/exercises/oets/oets.cu

+#include <stdio.h>
+#include <stdlib.h>
+
+#define NUM_ELEMENTS 16
+
+__device__ void swap(int *a, int *b) {
+  int temp = *a;
+  *a = *b;
+  *b = temp;
+}
+
+// TODO 2: define parameters
+__global__ void oddEvenTranspositionSort() {
+  int tid = blockIdx.x * blockDim.x + threadIdx.x;
+
+  for (int i = 0; i < n; i++) {
+    if (i % 2 == 0) {  // Even phase
+                       // TODO 2: Compare and swap elements if thread id is even
+    } else {           // Odd phase
+                       // TODO 3: Compare and swap elements if thread id is odd
+    }
+    // TODO 4: Sync threads
+  }
+}
+
+void generateData(int *data, int size) {
+  srand(time(0));
+
+  for (int i = 0; i < size; i++) {
+    data[i] = rand() % 14 + 1;
+  }
+}
+
+int main() {
+  int *array = NULL;
+  array = (int *)malloc(NUM_ELEMENTS * sizeof(int));
+  generateData(array, NUM_ELEMENTS);
+
+  printf("Original Array: ");
+  for (int i = 0; i < NUM_ELEMENTS; i++) {
+    printf("%d ", array[i]);
+  }
+  printf("\n");
+
+  int *d_array;
+  // TODO 0: Allocate device array and copy host elements to it
+
+  // TODO 1: Calculate blocks_no and block_size
+  oddEvenTranspositionSort<<<blocks_no, block_size>>>(d_array, NUM_ELEMENTS);
+  cudaDeviceSynchronize();
+
+  cudaMemcpy(array, d_array, NUM_ELEMENTS * sizeof(int),
+             cudaMemcpyDeviceToHost);
+  cudaFree(d_array);
+
+  printf("Sorted Array: ");
+  for (int i = 0; i < NUM_ELEMENTS; i++) {
+    printf("%d ", array[i]);
+  }
+  printf("\n");
+
+  free(array);
+  return 0;
+}