#pragma GCC optimize("O3")
#pragma GCC optimize("unroll-loops")
#pragma GCC target("avx2")
#include <bits/stdc++.h>
int ri() {
int n;
scanf("%d", &n);
return n;
}
struct BIT {
int n;
std::vector<int> data;
int all_sum = 0;
BIT (int n) : n(n), data(n + 1) {}
void add(int i, int val) {
for (i++; i <= n; i += i & -i) data[i] += val;
all_sum += val;
}
int sum_left(int r) {
int res = 0;
for (; r; r -= r & -r) res += data[r];
return res;
}
int sum_right(int l) { return all_sum - sum_left(l); }
};
struct SegTree {
int n;
std::vector<int> data;
SegTree (int n_) {
for (n = 1; n < n_; n <<= 1);
data.resize(n << 1);
}
void add(int i, int val) {
for (i += n; i; i >>= 1) data[i] += val;
}
int sum(int l, int r) {
int res = 0;
for (l += n, r += n; l < r; l >>= 1, r >>= 1) {
if (r & 1) res += data[--r];
if (l & 1) res += data[l++];
}
return res;
}
};
long long count_rectangles(std::vector<std::vector<int> > a) {
int n = a.size();
int m = a[0].size();
int64_t res = 0;
if (n == 3) {
auto solve_sub = [&] (std::vector<int> a) {
int n = a.size();
std::deque<int> all;
for (int i = n - 1; i >= 0; i--) {
if (i <= n - 3) {
int tmp = std::lower_bound(all.begin(), all.end(), a[i]) - all.begin();
res += std::max(0, tmp);
}
while (all.size() && all.front() <= a[i]) all.pop_front();
all.push_front(a[i]);
}
};
std::vector<int> all{a[1][0]};
for (int i = 1; i < n; i++) {
bool ok = i != n - 1 && a[1][i] < a[0][i] && a[1][i] < a[2][i];
if (!ok) {
if (all.size() > 1) {
all.push_back(a[1][i]);
solve_sub(all);
}
all = {a[1][i]};
} else all.push_back(a[1][i]);
}
} else if (n <= 700 && m <= 700) {
assert(0);
assert(n <= 700 && m <= 700);
std::vector<std::vector<std::vector<int> > > r1(n, std::vector<std::vector<int> > (m));
for (int i = 1; i + 1 < m; i++) {
std::vector<int> cur(m - i, 0);
for (int k = n - 2; k >= 1; k--) {
int max = -1;
for (int j = i + 1; j < m; j++) {
max = std::max(max, a[k][j - 1]);
if (max < a[k][j] && max < a[k][i - 1]) cur[j - i]++;
else cur[j - i] = 0;
}
r1[k][i] = cur;
}
}
auto calc = [&] (const std::vector<int> &r0, const std::vector<int> &r1) {
std::vector<int> all[r0.size()];
for (int k = 1; k < (int) r1.size(); k++) if (r1[k]) all[r1[k]].push_back(k);
BIT tree(r1.size());
for (int k = 1; k < (int) r0.size(); k++) {
tree.add(r0[k], 1);
for (auto l : all[k]) res += tree.sum_right(l);
}
};
for (int i = 1; i + 1 < n; i++) {
bool ok[n][m];
std::vector<int> max(m, -1);
for (int j = i + 1; j < n; j++) {
for (int k = 1; k + 1 < m; k++) {
max[k] = std::max(max[k], a[j - 1][k]);
ok[j][k] = max[k] < a[j][k] && max[k] < a[i - 1][k];
}
}
std::vector<int> cur(n - i, 0);
for (int j = m - 2; j >= 1; j--) {
for (int k = i + 1; k < n; k++) {
if (ok[k][j]) cur[k - i]++;
else cur[k - i] = 0;
}
calc(cur, r1[i][j]);
}
}
}
return res;
}
std::random_device rnd_dev;
std::mt19937 rnd(rnd_dev() ^ clock());
int rnd_int(int l, int r) { return std::uniform_int_distribution<>(l, r)(rnd); }
#ifdef LOCAL
int main() {
int n = ri();
int m = ri();
std::vector<std::vector<int> > a(n, std::vector<int>(m));
for (auto &i : a) for (auto &j : i) j = ri();
printf("%lld\n", count_rectangles(a));
{
int n, m;
n = m = 700;
std::vector<std::vector<int> > a(n, std::vector<int>(m));
for (auto &i : a) for (auto &j : i) j = rnd_int(1, 1000000);
clock_t r0 = clock();
int res = count_rectangles(a);
clock_t r1 = clock();
printf("%d\n", res);
printf("took %f ms\n", (double) (r1 - r0) * 1000 / CLOCKS_PER_SEC);
}
return 0;
}
#endif
