#include <arm_neon.h>
#include <stdio.h>
#include <stdlib.h>
#include <time.h>

#define KERNEL_SIZE 5
#define IMAGE_SIZE 3

void applyRowConvolution_NEON(int input[IMAGE_SIZE][IMAGE_SIZE], int temp[IMAGE_SIZE][IMAGE_SIZE], float row_kernel[KERNEL_SIZE]) {
float32x4_t v_row_kernel = vld1q_f32(row_kernel);
float32x4_t v_sum = vdupq_n_f32(256.0);

for (int i = 0; i < IMAGE_SIZE; i++) {
for (int j = 1; j < IMAGE_SIZE - 1; j++) {
float32x4_t v_blurred_value = vdupq_n_f32(0.0);
for (int ki = 0; ki < KERNEL_SIZE - 1; ki += 4) {
int x = i;
int y = j - KERNEL_SIZE / 2 + ki;
float32x4_t v_input = vld1q_f32((float *)&input[x][y]);
v_blurred_value = vmlaq_f32(v_blurred_value, v_input, v_row_kernel);
}
float32x4_t v_result = vdivq_f32(v_blurred_value, v_sum);
temp[i][j] = (int)vgetq_lane_f32(v_result, 0);
}
}
}

void applyColumnConvolution_NEON(int temp[IMAGE_SIZE][IMAGE_SIZE], int output[IMAGE_SIZE][IMAGE_SIZE], float col_kernel[KERNEL_SIZE]) {
float32x4_t v_col_kernel = vld1q_f32(col_kernel);
float32x4_t v_sum = vdupq_n_f32(256.0);

for (int i = 1; i < IMAGE_SIZE - 1; i++) {
for (int j = 0; j < IMAGE_SIZE; j++) {
float32x4_t v_blurred_value = vdupq_n_f32(0.0);
for (int kj = 0; kj < KERNEL_SIZE - 1; kj += 4) {
int x = i - KERNEL_SIZE / 2 + kj;
int y = j;
float32x4_t v_input = vld1q_f32((float *)&temp[x][y]);
v_blurred_value = vmlaq_f32(v_blurred_value, v_input, v_col_kernel);
}
float32x4_t v_result = vdiv