[C语言] memset的效率分析

测试1G元素的初始化时间长度作为效率对比参考值。

最后封装一个优化后的自定义双字节memset函数。

#include <stdio.h>
#include <stdint.h>
#include <time.h>

//一般32位的编译器,静态数据区只分配大约2g的内存

#define MAX_SIZE (1024*1024*1024)  //1024MiByte=1GByte
uint8_t pBuff[MAX_SIZE] = { 0 };

int main()
{
    clock_t start, end;
    double seconds = 0;
    uint64_t len = MAX_SIZE;
    uint16_t value = 0x1234;
    uint8_t ch_High = value >> 8;
    uint8_t ch_Low = value & 0xFF;

    printf("Exam MAX_SIZE=0x%lx.\n", MAX_SIZE);
    //测试1：传统for运行耗时0.374
    start = clock();
    for (uint64_t i = 0; i < (len / 2); i++)
    {
        pBuff[i] = ch_High;
        pBuff[i + 1] = ch_Low;
    }
    end = clock();
    seconds = (double)(end - start) / CLOCKS_PER_SEC;
    printf("Use time1: %.8fs \n", seconds);

    //测试2：指针for运行耗时0.256
    uint8_t* pt = pBuff;
    start = clock();
    for (uint64_t i = 0; i < (len / 2); i++)
    {
        *(uint16_t*)pt = value;
        pt += 2;
    }
    end = clock();
    seconds = (double)(end - start) / CLOCKS_PER_SEC;
    printf("Use time2: %.8fs \n", seconds);
    return 0;

}

以上运行结果是基于VS2022,Release 版运行的结果。CPU: 【双核】Intel i7@3.10GHZ

可以看出用指针赋值比传统数组赋值快50%

//优化后的my_memset函数，用于初始化双字节函数。
 
void my_memset(uint8_t *buf,uint16_t value,uint64_t len)
{
    for (uint64_t i = 0; i < len/2; i ++)
    {
        *(uint16_t*)buf = value;//关键：指针类型的转换
        buf += 2;
    }   

}