Shooting (Hard)

#include <bits/stdc++.h>
#define lowbit(x) ((x) & -(x))
#define ml(a, b) ((1ll * (a) * (b)) % M)
#define tml(a, b) (a) = ((1ll * (a) * (b)) % M)
#define ad(a, b) ((0ll + (a) + (b)) % M)
#define tad(a, b) (a) = ((0ll + (a) + (b)) % M)
#define mi(a, b) ((0ll + M + (a) - (b)) % M)
#define tmi(a, b) (a) = ((0ll + M + (a) - (b)) % M)
#define tmin(a, b) (a) = min((a), (b))
#define tmax(a, b) (a) = max((a), (b))
#define iter(a) (a).begin(), (a).end()
#define riter(a) (a).rbegin(), (a).rend()
#define inin(a, b) memset((a), (b), sizeof(a))
#define cpy(a, b) memcpy((a), (b), sizeof(a))
#define uni(a) a.resize(unique(iter(a)) - a.begin())
#define size(x) (int)x.size()
#define pb emplace_back
#define mpr make_pair
#define ls(i) ((i) << 1)
#define rs(i) ((i) << 1 | 1)
#define INF 0x3f3f3f3f
#define NIF 0xc0c0c0c0
#define eps 1e-9
#define F first
#define S second
#define AC cin.tie(0)->sync_with_stdio(0)
using namespace std;
typedef long long llt;
typedef unsigned long long ull;
typedef __int128_t lll;
typedef double dl;
typedef pair<int, int> pii;
typedef pair<dl, dl> pdd;
typedef pair<llt, llt> pll;
typedef pair<llt, int> pli;
typedef complex<dl> cd;
typedef complex<llt> cll;
// const int M = 998244353;

// random_device rm;
// mt19937 rg(rm());
// default_random_engine rg(rm());
// uniform_int_distribution<int> rnd(INT_MIN, INT_MAX);
// uniform_real_distribution<double> rd(0, M_PI);

void db() { cerr << "\n"; }
template <class T, class... U>
void db(T a, U... b) { cerr << a << " ", db(b...); }

inline char gc()
{
    const static int SZ = 1 << 16;
    static char buf[SZ], *p1, *p2;
    if (p1 == p2 && (p2 = buf + fread(p1 = buf, 1, SZ, stdin), p1 == p2))
        return -1;
    return *p1++;
}
void rd() {}
template <typename T, typename... U>
void rd(T &x, U &...y)
{
    x = 0;
    bool f = 0;
    char c = gc();
    while (!isdigit(c))
        f ^= !(c ^ 45), c = gc();
    while (isdigit(c))
        x = (x << 1) + (x << 3) + (c ^ 48), c = gc();
    f && (x = -x), rd(y...);
}

template <typename T>
void prt(T x)
{
    if (x < 0)
        putchar('-'), x = -x;
    if (x > 9)
        prt(x / 10);
    putchar((x % 10) ^ 48);
}

pll operator+(const pll &a, const pll &b)
{
    return pll(a.F + b.F, a.S + b.S);
}

pll operator-(const pll &a, const pll &b)
{
    return pll(a.F - b.F, a.S - b.S);
}

pll operator*(const pll &a, const llt &b)
{
    return pll(a.F * b, a.S * b);
}

signed main()
{
    int t, n, m, x;
    rd(t);
    while (t--)
    {
        rd(n, m);
        vector<pll> c1(n + m), c2(n + m), a1(n + m), a2(n + m);
#define fx(x, y) (x + y)
#define fy(x, y) (x - y + m - 1)
        for (int i = 0; i < n; i++)
        {
            for (int j = 0; j < m; j++)
            {
                rd(x);
                if (x == 1 || x == 3)
                    c1[fx(i, j)].F++, c2[fy(i, j)].F++;
                if (x == 2 || x == 3)
                    c1[fx(i, j)].S++, c2[fy(i, j)].S++;
            }
        }
        auto slv = [&](vector<pll> &c, vector<pll> &a) -> void
        {
            pll cr(0, 0), d(0, 0);
            for (int i = 0; i < n + m; i++)
                cr = cr + c[i] * i, d = d + c[i];
            for (int i = 0; i < n + m; i++)
            {
                a[i] = cr, d = d - c[i] * 2;
                cr = cr - d;
            }
        };
        slv(c1, a1), slv(c2, a2);
        for (int i = 0; i < n; i++)
        {
            for (int j = 0; j < m; j++)
            {
                pll ans = a1[fx(i, j)] + a2[fy(i, j)];
                prt(abs(ans.S - ans.F) >> 1);
                putchar(" \n"[j == m - 1]);
            }
        }
    }
}