I submitted two solutions, one with O(m+n) complexity (merge the entire arrays first and then find the median), and one with O(log(m+n)) complexity (divide-and-conquer to find K smallest element of the two arrays and then calculate median) and latter was only 2ms faster. Suggest providing a test case that's sufficiently large to make the difference between logarithmic and linear runtime performance obvious.