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

#define SIZE 1024  

void matmul_optimized(float** A, float** B, float** C, int n) {
    for (int i = 0; i < n; i++) {
        for (int j = 0; j < n; j++) {
            float32x4_t sum = vdupq_n_f32(0);  
            for (int k = 0; k < n; k += 4) {
                float32x4_t vecA = vld1q_f32(&A[i][k]);
                float32x4_t vecB = vld1q_f32(&B[k][j]);
                sum = vmlaq_f32(sum, vecA, vecB); 
            }
            C[i][j] = vgetq_lane_f32(sum, 0) + vgetq_lane_f32(sum, 1) + vgetq_lane_f32(sum, 2) + vgetq_lane_f32(sum, 3);  // 提取结果
        }
    }
}

int main() {
    float** A = (float**)malloc(SIZE * sizeof(float*));
    float** B = (float**)malloc(SIZE * sizeof(float*));
    float** C = (float**)malloc(SIZE * sizeof(float*));
    for (int i = 0; i < SIZE; i++) {
        A[i] = (float*)malloc(SIZE * sizeof(float));
        B[i] = (float*)malloc(SIZE * sizeof(float));
        C[i] = (float*)malloc(SIZE * sizeof(float));
    }
    
    for (int i = 0; i < SIZE; i++) {
        for (int j = 0; j < SIZE; j++) {
            A[i][j] = rand() % 100;
            B[i][j] = rand() % 100;
        }
    }

    clock_t start = clock();
    matmul_optimized(A, B, C, SIZE);
    clock_t end = clock();
    
    double time_taken = (double)(end - start) / CLOCKS_PER_SEC;
    printf("优化后的稠密矩阵乘法: %f seconds\n", time_taken);

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

    return 0;
}