計算字符串長度[C++/ASM]

#include <iostream>
#include <windows.h>
#include <chrono>

/*

    \主要使用指令 repne scasb, 如果使用repne scasd應該能更快

*/
__declspec(naked) DWORD WINAPI StringLength1(LPCSTR pszStr)
{
    __asm
    {
            push ebp
            mov ebp, esp
            or ecx, 0xffffffff
            mov edi, [ebp+8]
            xor eax, eax
            test edi, edi
            jz _fexit
            repne scasb
            not ecx
            mov eax, ecx
            dec eax    
    _fexit:
            leave
            retn 4
    }
}

/*

    \常規實現

*/
DWORD StringLength2(LPCSTR pszStr)
{
    if (NULL == pszStr)
        return 0;
    DWORD dwCnt = 0;
    while (*pszStr++)
        ++dwCnt;

    return dwCnt;
}

// MAIN
int main(int, char **)
{
    // 十億個字符
    char* test = new char[1000000000];
    memset(test, 65, 1000000000-1);

    using namespace std;
    using namespace chrono;

    double cz1, cz2;

    // 計算 StringLength1 執行時間
    auto start = system_clock::now();
    DWORD len1 = StringLength1(test);
    auto end = system_clock::now();

    auto dura = duration_cast<microseconds>(end - start);
    cz1 = double(dura.count()) * microseconds::period::num / microseconds::period::den;

    // 計算 StringLength2 執行時間
    start = system_clock::now();
    DWORD len2 = StringLength2(test);
    end = system_clock::now();
    
    dura = duration_cast<microseconds>(end - start);
    cz2 = double(dura.count()) * microseconds::period::num / microseconds::period::den;

    std::cout << "len1: " << len1 << ", cz1: " << cz1 << std::endl;
    std::cout << "len2: " << len2 << ", cz2: " << cz2 << std::endl;

    delete test;

    return 0;
}

結果就是 repne scasb 要快4倍左右.

但也沒啥用...