diff --git a/neon优化稀疏矩阵乘法 (3).c b/neon优化稀疏矩阵乘法 (3).c
new file mode 100644
index 0000000..42229fc
--- /dev/null
+++ b/neon优化稀疏矩阵乘法 (3).c	
@@ -0,0 +1,171 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <time.h>
+#include <arm_neon.h>
+
+// Ԥ������������������
+#define ROWS 10
+#define COLS 10
+
+// ϡ�����ṹ��
+typedef struct {
+    float *values;
+    int *rowIndex;
+    int *colIndex;
+    int nonZeroCount;
+} SparseMatrix;
+
+// ��ϡ�����ת��Ϊ��ͨ����
+void sparseToDense(SparseMatrix *sparse, float denseMatrix[ROWS][COLS]) {
+    for (int i = 0; i < ROWS; i++) {
+        for (int j = 0; j < COLS; j++) {
+            denseMatrix[i][j] = 0;
+        }
+    }
+
+    for (int k = 0; k < sparse->nonZeroCount; k++) {
+        int row = sparse->rowIndex[k];
+        int col = sparse->colIndex[k];
+        denseMatrix[row][col] = sparse->values[k];
+    }
+}
+
+// NEON�Ż��ľ���˷�����
+void matmul_optimized(float A[ROWS][COLS], float B[ROWS][COLS], float C[ROWS][COLS]) {
+    for (int i = 0; i < ROWS; i++) {
+        for (int j = 0; j < COLS; j++) {
+            // ��ʼ���������C�ĵ�ǰԪ��Ϊ0
+            C[i][j] = 0;
+
+            // �����ۼӵ������Ĵ�������ʼ��Ϊ0
+            float32x4_t vecC = vdupq_n_f32(0);
+
+            for (int k = 0; k < ROWS; 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);
+        }
+    }
+}
+
+// ��ӡ����
+void printMatrix(float matrix[ROWS][COLS]) {
+    for (int i = 0; i < ROWS; i++) {
+        for (int j = 0; j < COLS; j++) {
+            printf("%f ", matrix[i][j]);
+        }
+        printf("\n");
+    }
+}
+
+int main() {
+    // ��ʼ��ϡ�����A
+    SparseMatrix sparseA;
+    sparseA.nonZeroCount = 5;
+    sparseA.values = (float *)malloc(sparseA.nonZeroCount * sizeof(float));
+    sparseA.rowIndex = (int *)malloc(sparseA.nonZeroCount * sizeof(int));
+    sparseA.colIndex = (int *)malloc(sparseA.nonZeroCount * sizeof(int));
+
+    if (sparseA.values == NULL || sparseA.rowIndex == NULL || sparseA.colIndex == NULL) {
+        free(sparseA.values);
+        free(sparseA.rowIndex);
+        free(sparseA.colIndex);
+        fprintf(stderr, "Memory allocation failed for sparse matrix A!\n");
+        return 1;
+    }
+
+    sparseA.values[0] = 1;
+    sparseA.values[1] = 2;
+    sparseA.values[2] = 3;
+    sparseA.values[3] = 4;
+    sparseA.values[4] = 5;
+
+    sparseA.rowIndex[0] = 0;
+    sparseA.rowIndex[1] = 1;
+    sparseA.rowIndex[2] = 2;
+    sparseA.rowIndex[3] = 1;
+    sparseA.rowIndex[4] = 2;
+
+    sparseA.colIndex[0] = 0;
+    sparseA.colIndex[1] = 1;
+    sparseA.colIndex[2] = 2;
+    sparseA.colIndex[3] = 0;
+    sparseA.colIndex[4] = 1;
+
+    // ��ʼ��ϡ�����B
+    SparseMatrix sparseB;
+    sparseB.nonZeroCount = 4;
+    sparseB.values = (float *)malloc(sparseB.nonZeroCount * sizeof(float));
+    sparseB.rowIndex = (int *)malloc(sparseB.nonZeroCount * sizeof(int));
+    sparseB.colIndex = (int *)malloc(sparseB.nonZeroCount * sizeof(int));
+
+    if (sparseB.values == NULL || sparseB.rowIndex == NULL || sparseB.colIndex == NULL) {
+        free(sparseB.values);
+        free(sparseB.rowIndex);
+        free(sparseB.colIndex);
+        fprintf(stderr, "Memory allocation failed for sparse matrix B!\n");
+        return 1;
+    }
+
+    sparseB.values[0] = 2;
+    sparseB.values[1] = 3;
+    sparseB.values[2] = 4;
+    sparseB.values[3] = 5;
+
+    sparseB.rowIndex[0] = 0;
+    sparseB.rowIndex[1] = 1;
+    sparseB.rowIndex[2] = 1;
+    sparseB.rowIndex[3] = 2;
+
+    sparseB.colIndex[0] = 0;
+    sparseB.colIndex[1] = 1;
+    sparseB.colIndex[2] = 0;
+    sparseB.colIndex[3] = 1;
+
+    // ��������ʼ���������
+    float denseA[ROWS][COLS];
+    float denseB[ROWS][COLS];
+    float denseC[ROWS][COLS];
+
+    // ��ϡ�����Aת��Ϊ�������denseA
+    sparseToDense(&sparseA, denseA);
+
+    // ��ϡ�����Bת��Ϊ�������denseB
+    sparseToDense(&sparseB, denseB);
+
+    // ��ӡת����ij������A
+    printf("matrixA:\n");
+    printMatrix(denseA);
+
+    // ��ӡת����ij������B
+    printf("matrixB:\n");
+    printMatrix(denseB);
+
+    // ��¼NEON�Ż��ľ���˷�����ʱ��
+    clock_t startMul, endMul;
+    startMul = clock();
+    matmul_optimized(denseA, denseB, denseC);
+    endMul = clock();
+    double time_taken_Mul = ((double)(endMul - startMul)) / CLOCKS_PER_SEC;
+    printf("NEON time: %lfs\n", time_taken_Mul);
+    
+    // �ͷ�ϡ�����A���ڴ�
+    free(sparseA.values);
+    free(sparseA.rowIndex);
+    free(sparseA.colIndex);
+
+    // �ͷ�ϡ�����B���ڴ�
+    free(sparseB.values);
+    free(sparseB.rowIndex);
+    free(sparseB.colIndex);
+
+    return 0;
+}