#include <bits/stdc++.h>
#pragma GCC optimize "tree-vectorize,unroll-loops"
#pragma GCC target "avx512vl,avx512bw,avx512cd"

using namespace std;

constexpr int S = 1 << 12, N = 1 << 16;

alignas(64) uint16_t p[N], tp[N], cmp[S], b0[S], b1[S];
int64_t ans, n, m;

template <int s>
void sand(const uint16_t* __restrict__ a, const uint16_t* __restrict__ b, uint16_t* __restrict__ c) {
  if (s < S) {
    for (int i = 0; i + s < S; ++i) c[i] = a[i + s] & b[i];
    for (int i = S - s; i < S; ++i) c[i] = a[i + s - S] >> 1 & b[i];
  } else {
    for (int i = 0; i < S; ++i) c[i] = a[i] >> s / S & b[i];
  }
}

template <int S = ::N>
void sand(const uint16_t* __restrict__ a, const uint16_t* __restrict__ b, uint16_t* __restrict__ c, int s) {
  if (s == S) sand<S>(a, b, c);
  else if constexpr (S == 1) __builtin_unreachable();
  else sand<S / 2>(a, b, c, s);
}

int main() {
  cin.tie(0)->sync_with_stdio(0);
  cin >> n >> m;
  for (int i = 0; i < n; ++i) cin >> p[i];
  for (int s = 1; s + m <= n; ++s) {
    cmp[n - s] = 0;
    if (n - s < S) fill(cmp + n - s, cmp + S, 0);
    #pragma GCC unroll 99
    for (int t = 0; t < N; t += S) if (t + s < n)
    for (int j = 0; j < S; ++j) cmp[j] |= (p[j + t] < p[j + t + s]) << t / S;
    fill(b0, end(b0), -1);
    uint16_t* r = b0, * a = cmp, * t = b1;
    #pragma GCC unroll 99
    for (int b = 1; b < N; sand(a, a, t, b), swap(a, t), b *= 2) if (b & m) sand(r, a, t, b), swap(r, t);
    for (int i = 0; i < S / 4; ++i) ans += __builtin_popcountll((uint64_t&)r[i * 4]);
  }
  cout << ans << '\n';
}