strsafe functions performance

Question

hi
strsafe.h functions do more checks to do string operations so I think standard string functions is faster is this correct? or should I use strsafe functions?
thanks

Answer 1

I used following code to measure speed and I post result:

TCHAR buffer[1024];  
LPCTSTR pszText = _T("This is sample text. This is sample text. This is sample text. This is sample text. This is sample text.");  

LARGE_INTEGER start[4] = { 0 };  
LARGE_INTEGER end[4] = { 0 };  

QueryPerformanceCounter(&start[0]);		  
for (UINT i = 0; i < 10000; i++)  
{  
	_tcscpy(buffer, pszText);  
}	  
QueryPerformanceCounter(&end[0]);  


QueryPerformanceCounter(&start[1]);  
for (UINT i = 0; i < 10000; i++)  
{  
	_tcscpy_s(buffer, 1024, pszText);  
}  
QueryPerformanceCounter(&end[1]);  

QueryPerformanceCounter(&start[2]);  
for (UINT i = 0; i < 10000; i++)  
{  
	lstrcpy(buffer, pszText);  
}  
QueryPerformanceCounter(&end[2]);  

QueryPerformanceCounter(&start[3]);  
for (UINT i = 0; i < 10000; i++)  
{  
	StringCchCopy(buffer, 1024, pszText);  
}  
QueryPerformanceCounter(&end[3]);  

CString str;  
str.Format(_T("_tcscpy(): %I64u\n_tcscpy_s(): %I64u\nlstrcpy(): %I64u\nStringCchCopy(): %I64u"),   
	end[0].QuadPart - start[0].QuadPart,  
	end[1].QuadPart - start[1].QuadPart,  
	end[2].QuadPart - start[2].QuadPart,  
	end[3].QuadPart - start[3].QuadPart);  
MessageBox(0, str, _T("Result") , MB_OK | MB_ICONINFORMATION);  

Results: (in 4)

tcscpy() has best speed and StringCchCopy() has worste

Answer 2

For me, the results are as the following for 1'000'000 loop (Ryzen 5600x, Debug build):

auto start = std::chrono::steady_clock::now();
for (std::size_t i = 0; i < 1'000'000; ++i) {
    //StringCchCopy(lpszCurrentVariable, BUFSIZE, TEXT("MySetting=A")); // 32ms
    auto r1 = wcscpy(lpszCurrentVariable, TEXT("MySetting=B"));         // 42ms
}
using namespace std::chrono; // NOLINT(google-build-using-namespace)
std::cout << std::format("Elapsed in XX : {:d}ms\n",
                         duration_cast<milliseconds>(steady_clock::now() - start).count());

For Release builds, both are 0ms.