ADD file via upload

未命名2.cpp

@@ -0,0 +1,97 @@
+#include <stdio.h>
+#include <time.h>
+
+// 原始的高斯模糊函数
+void applyGaussianBlur(float src[][5], float dst[][5], int h, int w, float kernel[3][3]) {
+    for (int i = 1; i < h - 1; i++) {
+        for (int j = 1; j < w - 1; j++) {
+            float sum = 0.0;
+            for (int ki = -1; ki <= 1; ki++) {
+                for (int kj = -1; kj <= 1; kj++) {
+                    sum += src[i + ki][j + kj] * kernel[ki + 1][kj + 1];
+                }
+            }
+            dst[i][j] = sum;
+        }
+    }
+}
+
+// 优化后的可分离高斯模糊函数
+void applySeparableGaussianBlur(float src[][5], float dst[][5], int h, int w, float kx[3], float ky[3]) {
+    // 用于存储行内卷积的中间结果
+    float buf[3][5];
+    // 宏定义用于循环利用 buf 数组的 3 个行缓冲区
+#define BUF_ROW(r) buf[(r) % 3]
+    // 循环计算前两行的行内卷积并存储在 buf 中
+    for (int j = 1; j < w - 1; j++) {
+        BUF_ROW(0)[j] = src[1][j] * kx[1] + src[1][j - 1] * kx[0] + src[1][j + 1] * kx[2];
+        BUF_ROW(1)[j] = src[2][j] * kx[1] + src[2][j - 1] * kx[0] + src[2][j + 1] * kx[2];
+    }
+    // 每算一行新的，凑齐 3 行，做一次行间的卷积。期间更新 buf
+    for (int i = 2; i < h - 1; i++) {
+        for (int j = 1; j < w - 1; j++) {
+            BUF_ROW(2)[j] = src[i][j] * kx[1] + src[i][j - 1] * kx[0] + src[i][j + 1] * kx[2];
+            float sum = BUF_ROW(0)[j] * ky[0] + BUF_ROW(1)[j] * ky[1] + BUF_ROW(2)[j] * ky[2];
+            dst[i][j] = sum;
+        }
+        // 更新 buf
+        for (int j = 1; j < w - 1; j++) {
+            BUF_ROW(0)[j] = BUF_ROW(1)[j];
+            BUF_ROW(1)[j] = BUF_ROW(2)[j];
+        }
+    }
+}
+
+int main() {
+    float src[5][5] = {
+            {1, 2, 3, 4, 5},
+            {6, 7, 8, 9, 10},
+            {11, 12, 13, 14, 15},
+            {16, 17, 18, 19, 20},
+            {21, 22, 23, 24, 25}
+    };
+    float kernel[3][3] = {
+            {1.0 / 16, 2.0 / 16, 1.0 / 16},
+            {2.0 / 16, 4.0 / 16, 2.0 / 16},
+            {1.0 / 16, 2.0 / 16, 1.0 / 16}
+    };
+    // 拆分 kernel 为行方向和列方向两个向量
+    float kx[3] = {kernel[0][0], kernel[0][1], kernel[0][2]};
+    float ky[3] = {kernel[0][0], kernel[1][0], kernel[2][0]};
+    float dst1[5][5];
+    float dst2[5][5];
+
+    // 使用原始高斯模糊函数并计时
+    clock_t start1 = clock();
+    applyGaussianBlur(src, dst1, 5, 5, kernel);
+    clock_t end1 = clock();
+    double elapsedTime1 = (double)(end1 - start1) / CLOCKS_PER_SEC;
+
+    // 使用优化后的高斯模糊函数并计时
+    clock_t start2 = clock();
+    applySeparableGaussianBlur(src, dst2, 5, 5, kx, ky);
+    clock_t end2 = clock();
+    double elapsedTime2 = (double)(end2 - start2) / CLOCKS_PER_SEC;
+
+    // 输出原始高斯模糊结果和时间
+    printf("Original Gaussian Blur Output matrix:\n");
+    for (int i = 0; i < 5; i++) {
+        for (int j = 0; j < 5; j++) {
+            printf("%f ", dst1[i][j]);
+        }
+        printf("\n");
+    }
+    printf("Time taken by original function: %f seconds\n", elapsedTime1);
+
+    // 输出优化后的高斯模糊结果和时间
+    printf("\nOptimized Separable Gaussian Blur Output matrix:\n");
+    for (int i = 0; i < 5; i++) {
+        for (int j = 0; j < 5; j++) {
+            printf("%f ", dst2[i][j]);
+        }
+        printf("\n");
+    }
+    printf("Time taken by optimized function: %f seconds\n", elapsedTime2);
+
+    return 0;
+}