From 970f7bc84e3ff9dbd08ebad35dac625a53334192 Mon Sep 17 00:00:00 2001
From: paz4r2p6k <2252672714@qq.com>
Date: Fri, 29 Nov 2024 19:25:39 +0800
Subject: [PATCH] ADD file via upload

---
 ...化稠密矩阵乘法及时间比较 (1).c | 104 ++++++++++++++++++
 1 file changed, 104 insertions(+)
 create mode 100644 neon优化稠密矩阵乘法及时间比较 (1).c

diff --git a/neon优化稠密矩阵乘法及时间比较 (1).c b/neon优化稠密矩阵乘法及时间比较 (1).c
new file mode 100644
index 0000000..a3dae89
--- /dev/null
+++ b/neon优化稠密矩阵乘法及时间比较 (1).c	
@@ -0,0 +1,104 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <time.h>
+#include <arm_neon.h>
+
+void matmul(float** A, float** B, float** C, int n) {
+	int i,k,j; 
+    for (i = 0; i < n; i++) {
+        for (j = 0; j < n; j++) {
+            C[i][j] = 0;
+            for (k = 0; k < n; k++) {
+                C[i][j] += A[i][k] * B[k][j];
+            }
+        }
+    }
+}
+
+// NEON�Ż��ľ���˷�����
+void matmul_optimized(float** A, float** B, float** C, int n) {
+	int i,j,k;
+    for (i = 0; i < n; i++) {
+        for (j = 0; j < n; j++) {
+            // ��ʼ���������C�ĵ�ǰԪ��Ϊ0
+            C[i][j] = 0;
+
+            // �����ۼӵ������Ĵ�������ʼ��Ϊ0
+            float32x4_t vecC = vdupq_n_f32(0);
+
+            for (k = 0; k < n; k += 4) {
+                // ���ؾ���A��һ���е�4������Ԫ�ص������Ĵ���
+                float32x4_t vecA = vld1q_f32(&A[i][k]);
+                // ���ؾ���B��һ���е�4������Ԫ�ص������Ĵ�����ע������Ҫת�õ��߼���ʵ���ǰ���ȡԪ�أ�
+                float32x4_t vecB = vld1q_f32(&B[k][j]);
+
+                // ��ӦԪ����˲��ۼӵ�vecC
+                vecC = vmlaq_f32(vecC, vecA, vecB);
+            }
+
+            // ���ۼӽ���������Ĵ�����ȡ���ۼӵ�C[i][j]
+            C[i][j] += vgetq_lane_f32(vecC, 0) + vgetq_lane_f32(vecC, 1) + vgetq_lane_f32(vecC, 2) + vgetq_lane_f32(vecC, 3);
+        }
+    }
+}
+
+int main() {
+    // ��������С
+    int n = 1024;
+
+    // ��̬���������������A��B���Լ��������C���ڴ�
+    float** A = (float**)malloc(n * sizeof(float*));
+    float** B = (float**)malloc(n * sizeof(float*));
+    float** C = (float**)malloc(n * sizeof(float*));
+
+	int i,j;
+    for (i = 0; i < n; i++) {
+        A[i] = (float*)malloc(n * sizeof(float));
+        B[i] = (float*)malloc(n * sizeof(float));
+        C[i] = (float*)malloc(n * sizeof(float));
+    }
+
+    // ��ʼ���������ݣ���A��B�����ÿ��Ԫ�������ʼ��
+    srand((unsigned int)time(NULL));
+    for (i = 0; i < n; i++) {
+        for (j = 0; j < n; j++) {
+            A[i][j] = (float)(rand() % 100);
+            B[i][j] = (float)(rand() % 100);
+        }
+    }
+
+    clock_t start, end;
+
+    // ������ͨ����˷�����������ʱ��
+    start = clock();
+    matmul(A, B, C, n);
+    end = clock();
+    double time_taken_normal = ((double)(end - start)) / CLOCKS_PER_SEC;
+    printf("normal time: %lf s\n", time_taken_normal);
+
+    // ���³�ʼ���������CΪ0���Ա��ٴ�ʹ��
+    for (i = 0; i < n; i++) {
+        for (j = 0; j < n; j++) {
+            C[i][j] = 0;
+        }
+    }
+
+    // ����NEON�Ż�����˷�����������ʱ��
+    start = clock();
+    matmul_optimized(A, B, C, n);
+    end = clock();
+    double time_taken_optimized = ((double)(end - start)) / CLOCKS_PER_SEC;
+    printf("NEON time: %lf s\n", time_taken_optimized);
+
+    // �ͷŶ�̬������ڴ�ռ�
+    for ( i = 0; i < n; i++) {
+        free(A[i]);
+        free(B[i]);
+        free(C[i]);
+    }
+    free(A);
+    free(B);
+    free(C);
+
+    return 0;
+}