#include // {{{ // https://codeforces.com/blog/entry/96344 #pragma GCC optimize("O2,unroll-loops") #pragma GCC target("avx2,bmi,bmi2,lzcnt,popcnt") using namespace std; using i32 = int32_t; using u32 = uint32_t; using i64 = int64_t; using u64 = uint64_t; using f64 = double; using f128 = long double; #if __cplusplus >= 202002L template constexpr T MIN = std::numeric_limits::min(); template constexpr T MAX = std::numeric_limits::max(); template [[nodiscard]] static T sc(auto&& x) { return static_cast(x); } template [[nodiscard]] static T sz(auto&& x) { return static_cast(x.size()); } #endif static void NO() { std::cout << "NO\n"; } static void YES() { std::cout << "YES\n"; } template using vec = std::vector; #define all(x) (x).begin(), (x).end() #define rall(x) (x).rbegin(), (x).rend() #define ff first #define ss second #ifdef LOCAL #define db(...) std::print(__VA_ARGS__) #define dbln(...) std::println(__VA_ARGS__) #else #define db(...) #define dbln(...) #endif // }}} void solve() { u32 n; cin >> n; vector grid(n); for (auto& e : grid) cin >> e; /* core strategy: walk every char in tL region, find other 3 matching chars rotationally - find formula sort, dist is sum of char distances for each smaller than biggest (can only inc char) this is ans optimizes walking "rin" impl: fill in grid as we go fore easer coverage actually, just walk top side */ u32 ans = 0; for (i32 r = 0; r < n / 2; ++r) { // NOTE: missed this condition + other vars, made unecessary // // no need for pairs // why is auto comment on, fix the autogroup for (i32 c = r; c < n - 1 - r; ++c) { char big = max({grid[r][c], grid[c][n - 1 - r], grid[n - 1 - r][n - 1 - c], grid[n - 1 - c][r]}); ans += big - grid[r][c]; ans += big - grid[c][n - 1 - r]; ans += big - grid[n - 1 - r][n - 1 - c]; ans += big - grid[n - 1 - c][r]; } } cout << ans << endl; } int main() { // {{{ cin.tie(nullptr)->sync_with_stdio(false); cin.exceptions(cin.failbit); u32 tc = 1; cin >> tc; for (u32 t = 0; t < tc; ++t) { solve(); } return 0; } // }}}