CODEWITHSUNDEEP
c++stdvector
jaanq
jaanq

Reputation: 55

C++: get int from any place of vector<byte>

I have big enough

std::vector<byte> source

and I need to get four bytes from any offset in vector (for example, 10-13 bytes) and convert it to integer.

int ByteVector2Int(std::vector &source, int offset)
{
return (source[offset] | source[offset + 1] << 8 | source[offset + 2] << 16 | source[offset + 3] << 24);
}

This method called too offen, how I can do that with maximum perfomance?

Upvotes: 1

Views: 1073

Answers (1)

Daniel Langr
Daniel Langr

Reputation: 23497

Use memcpy. You might be tempted to use reinterpret_cast, but then you can easily end up with undefined behavior (for instance due to alignment issues). Also, pass a vector by a const reference:

int f(const std::vector<std::byte>& v, size_t n)
{
    int temp;
    memcpy(&temp, v.data() + n, sizeof(int));
    return temp;  
}

Note that compilers are very good in optimizations. In my case, GCC with -O2 resulted in:

mov     rax, qword ptr [rdi]
mov     eax, dword ptr [rax + rsi]
ret

So, there is no memcpy invoked and the assembly is minimal. Live demo: https://godbolt.org/z/oWGqej

UPDATE (based on question update)

After edit, you may also notice that the generated assembly is the very same (in my case) as for your approach:

int f2(const std::vector<std::byte>& v, size_t n)
{
  return (int)(
     (unsigned int)v[n]
     + ((unsigned int)v[n + 1] << 8)
     + ((unsigned int)v[n + 2] << 16)
     + ((unsigned int)v[n + 3] << 24) );
}

Live demo: https://godbolt.org/z/c9dE9W

Note that your code is not correct. First, bitwise operations are performed with std::byte which overflows, and second, there is no implicit conversion of std::byte to int.

Upvotes: 6

Related Questions