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

#define SIZE 1024

// 基础矩阵乘法函数
void matmul(float A[SIZE][SIZE], float B[SIZE][SIZE], float C[SIZE][SIZE], int n) {
    for (int i = 0; i < n; i++) {
        for (int j = 0; j < n; j++) {
            float sum = 0;
            for (int k = 0; k < n; k++) {
                sum += A[i][k] * B[k][j];
            }
            C[i][j] = sum;
        }
    }
}

// NEON优化的矩阵乘法函数
void matmul_optimized(float A[SIZE][SIZE], float B[SIZE][SIZE], float C[SIZE][SIZE], int n) {
    int i, j, k;
    for (i = 0; i < n; i++) {
        for (j = 0; j < n; j += 4) {
            float32x4_t sum = vdupq_n_f32(0);
            for (k = 0; k < n; k++) {
                float32x4_t a = vld1q_f32(&A[i][k]); // Load 4 elements from A
                float32x4_t b = vld1q_f32(&B[k][j]); // Load 4 elements from B
                sum = vmlaq_f32(sum, a, b); // Multiply and accumulate
            }
            vst1q_f32(&C[i][j], sum); // Store the result
        }
    }
}

int main() {
    // 矩阵内存分配
    float (*A)[SIZE] = malloc(SIZE * SIZE * sizeof(float));
    float (*B)[SIZE] = malloc(SIZE * SIZE * sizeof(float));
    float (*C)[SIZE] = malloc(SIZE * SIZE * sizeof(float));

    if (A == NULL || B == NULL || C == NULL) {
        printf("Memory allocation failed!\n");
        return 1;
    }

    // 初始化矩阵数据
    srand(time(NULL)); // 初始化随机数种子
    for (int i = 0; i < SIZE; i++) {
        for (int j = 0; j < SIZE; j++) {
            A[i][j] = rand() % 100;
            B[i][j] = rand() % 100;
        }
    }


    for (int i = 0; i < SIZE; i++) {
        for (int j = 0; j < SIZE; j++) {
            C[i][j] = 0;
        }
    }

    // 计时并输出基础矩阵乘法的运行时间
    clock_t start, end;
    start = clock();
    matmul(*A, *B, *C, SIZE);
    end = clock();
    double time_spent = (double)(end - start) / CLOCKS_PER_SEC;
    printf("基础矩阵乘法的运行时间为: %lf 秒\n", time_spent);


    for (int i = 0; i < SIZE; i++) {
        for (int j = 0; j < SIZE; j++) {
            C[i][j] = 0;
        }
    }

    // 计时并输出NEON优化矩阵乘法的运行时间
    start = clock();
    matmul_optimized(*A, *B, *C, SIZE);
    end = clock();
    time_spent = (double)(end - start) / CLOCKS_PER_SEC;
    printf("NEON优化矩阵乘法的运行时间为: %lf 秒\n", time_spent);


    free(A);
    free(B);
    free(C);

    return 0;
}