UPD: I've reduced the code size.

I've recently found that the following code will generate wrong output.

#include <bitset>
#include <iostream>
const int N = 105;
std::bitset<N> ok[N][N];
int n = 5;
int main() {
  ok[2][2].set(2);
  for (int i = n; i; i--)
    for (int j = i; j <= n; j++) {
      ok[i][j] = ok[i][j] | ok[i + 1][j] | ok[i][j - 1];
    }
  std::cout << ok[2][5][2] << '\n';
  return 0;
}

Compiled with -O3 -mtune=skylake -march=skylake, the code outputs 0.

However if you simulate the code you will know that the correct answer should be 1.

Note that the compiler seems to generate wrong sse instruction.

Godbolt link

Again, I believe this code is ub-free, and has nothing to do with implementation-defined stuff.