---
title: Intrinsics实现示例
description: "如果需要在代码中调用NEON Intrinsics函数，需要加入头文件\"arm_neon.h\"。以数组加法为例。"
url: https://www.hikunpeng.com/document/detail/zh/perftuning/progtuneg/kunpengprogramming_05_0028.html
sourcePath: /source/zh/perftuning/progtuneg/kunpengprogramming_05_0028.html
indexId: 562998976c9f3bd50845c070452ba57e689a31cdf76d79846f78713bd1fd24cf63
---
# Intrinsics实现示例

如果需要在代码中调用NEON Intrinsics函数，需要加入头文件"arm_neon.h"。以数组加法为例。

C语言实现：

```
using namespace std;
void add(int* out, int* input1, int* input2, int count)
{
for(int i = 0; i < count; i += 1)
{
out[i] = input1[i] + input2[i];
}
}
int main()
{
int count;
count = 10000 * 4;
int a[count];
int b[count];
int c[count];
clock_t start, finish;
double duration;
for(int i = 0; i < count; i += 1)
{
a[i] = rand();
}
for(int i = 0; i < count; i += 1)
{
b[i] = rand();
}
start = clock();
for(int i = 0; i < count; i += 1)
{
add(c, a, b, count);
}
finish = clock();
duration = (double)(finish - start) / CLOCKS_PER_SEC;
printf( "%f seconds\n", duration);
return 0;
}
```

输出结果如下：

```
1.910000 seconds
```

NEON Intrinsics实现：

```
using namespace std;
void add_neon(int* out, int* input1, int* input2, int count)
{
int32x4_t input1_neon, input2_neon, out_neon;
for(int i = 0; i < count; i += 4)
{
input1_neon = vld1q_s32(input1);
input1 += 4;
input2_neon = vld1q_s32(input2);
input2 += 4;
out_neon = vaddq_s32(input1_neon, input2_neon);
vst1q_s32(out, out_neon);
out += 4;
}
}
int main()
{
int count;
count = 10000 * 4;
int a[count];
int b[count];
int c[count];
clock_t start, finish;
double duration;
for(int i = 0; i < count; i += 1)
{
a[i] = rand();
}
for(int i = 0; i < count; i += 1)
{
b[i] = rand();
}
start = clock();
for(int i = 0; i < count; i += 1)
{
add_neon(c, a, b, count);
}
finish = clock();
duration = (double)(finish - start) / CLOCKS_PER_SEC;
printf( "%f seconds for neon\n", duration);
return 0;
}
```

输出结果：

```
0.360000 seconds for neon
```

可以看出，使用NEON Intrinsics实现，性能有了明显的提升。