#pragma GCC optimize("O3,unroll-loops") /* #pragma GCC optimize("Ofast,unroll-loops") */ // wtf, tera jest 1.37s. czemu /* #pragma GCC optimize("O3,unroll-loops,prefetch-loop-arrays") */ // slower z tą pragmą bez własnych optow interestingly. // -mprefetchwt1 todo -> to czesc avx512 though? /* #pragma GCC target("prefetch-loop-arrays") */ /* #pragma GCC target("avx2") */ /* #pragma GCC optimize("Ofast,unroll-loops") */ #include <bits/stdc++.h> #include <ext/pb_ds/assoc_container.hpp> #include <ext/pb_ds/tree_policy.hpp> #define pb push_back #define mp make_pair #define all(a) begin(a),end(a) #define FOR(x,val,to) for(int x=(val);x<int((to));++x) #define FORE(x,val,to) for(auto x=(val);x<=(to);++x) #define FORR(x,arr) for(auto &x: arr) #define FORS(x,plus,arr) for(auto x = begin(arr)+(plus); x != end(arr); ++x) #define FORREV(x,plus,arr) for(auto x = (arr).rbegin()+(plus); x !=(arr).rend(); ++x) #define REE(s_) {cout<<s_<<'\n';exit(0);} #define GET(arr) for(auto &i: (arr)) sc(i) #define whatis(x) cerr << #x << " is " << (x) << endl; #define e1 first #define e2 second #define INF 0x7f7f7f7f typedef std::pair<int,int> pi; typedef std::vector<int> vi; typedef std::vector<std::string> vs; typedef int64_t ll; typedef uint64_t ull; #define umap unordered_map #define uset unordered_set using namespace std; using namespace __gnu_pbds; #ifdef ONLINE_JUDGE #define whatis(x) ; #endif #ifdef _WIN32 #define getchar_unlocked() _getchar_nolock() #define _CRT_DISABLE_PERFCRIT_LOCKS #endif template<class L, class R> ostream& operator<<(ostream &os, map<L, R> P) { for(auto const &vv: P)os<<"("<<vv.first<<","<<vv.second<<")"; return os; } template<class T> ostream& operator<<(ostream &os, set<T> V) { os<<"[";for(auto const &vv:V)os<<vv<<","; os<<"]"; return os; } template<class T> ostream& operator<<(ostream &os, vector<T> V) { os<<"[";for(auto const &vv:V)os<<vv<<","; os<<"]"; return os; } template<class L, class R> ostream& operator<<(ostream &os, pair<L, R> P) { os<<"("<<P.first<<","<<P.second<<")"; return os; } inline int fstoi(const string &str){auto it=str.begin();bool neg=0;int num=0;if(*it=='-')neg=1;else num=*it-'0';++it;while(it<str.end()) num=num*10+(*it++-'0');if(neg)num*=-1;return num;} inline void getch(char &x){while(x = getchar_unlocked(), x < 33){;}} inline void getstr(string &str){str.clear(); char cur;while(cur=getchar_unlocked(),cur<33){;}while(cur>32){str+=cur;cur=getchar_unlocked();}} template<typename T> inline bool sc(T &num){ bool neg=0; int c; num=0; while(c=getchar_unlocked(),c<33){if(c == EOF) return false;} if(c=='-'){ neg=1; c=getchar_unlocked(); } for(;c>47;c=getchar_unlocked()) num=num*10+c-48; if(neg) num*=-1; return true;}template<typename T, typename ...Args> inline void sc(T &num, Args &...args){ bool neg=0; int c; num=0; while(c=getchar_unlocked(),c<33){;} if(c=='-'){ neg=1; c=getchar_unlocked(); } for(;c>47;c=getchar_unlocked()) num=num*10+c-48; if(neg) num*=-1; sc(args...); } template<typename T> using ordered_set = tree<T, null_type, less<T>, rb_tree_tag, tree_order_statistics_node_update>; //s.find_by_order(), s.order_of_key() <- works like lower_bound template<typename T> using ordered_map = tree<T, int, less<T>, rb_tree_tag, tree_order_statistics_node_update>; #define N 1000001 /* #define MXN */ // wa z tym^^ for some reason.. // -> przez to że dodawałem -1 zamiast jakiegoś -LL_MAX / 2. // guess best just wywalić MXN. #define PRINTS #define assert(x) ; /* #define VERIFY */ #define whatis(x) ; int main(){ ios_base::sync_with_stdio(0);cin.tie(0); /* assert(0); */ /* constexpr int mxn = 10000; */ /* constexpr int mxn = 10000, mxk = 5000; */ /* constexpr int mxn = 10000; */ /* constexpr int mxn = 20000; */ /* constexpr int mxk = 32 * 313 + 3; */ /* constexpr int mxk = 5000; */ /* constexpr int mxk = 16 * 312 + 3; */ // wtf, czemu przez wiekszy mxk nie działa ? xd // -> to było < 5000 xdd. /* constexpr int mxn = 10000; */ /* constexpr int mxk = 16 * 313 + 3; */ /* constexpr int mxk = 16 * 313 + 3; */ /* constexpr int mxk = 16 * 313 + 1; */ // +1, +3 both dobrze. // +2, +4 bad. // todo podzielne przez 64? constexpr int mxn = 100000; constexpr int mxk = 50000; int n,k; sc(n,k); #ifdef MXN /* int _a[mxn], _b[mxn]; */ int _a[mxn], _b[mxn]; // XDDD, kox bug też. memset(_a,-1,sizeof _a); memset(_b,-1,sizeof _b); FOR(i,0,n){ sc(_a[i]); } FOR(i,0,n){ sc(_b[i]); } #else int _a[n], _b[n]; GET(_a); GET(_b); #endif // xd. /* const int *a = a; */ /* const int *b = b; */ const int *a = _a; const int *b = _b; bool swpd = 0; if(k > n - k){ k = n - k; swap(a, b); swpd = 1; } // todo opt z high = sumą maxvali. // -> kox opt. /* ll low = 0, high = n * (ll)1000000000, mid, ans; */ // ??? high = 0 ll low = 0, high = 0, mid, ans; // XDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD /* FOR(i,0,n) */ //* FOR(i,0,n) */ /* high += max(a[i], b[i]); // down to 1.5s from 1.7s with this opt. */ FOR(i,0,n) high += max({0, a[i], b[i]}); // not overkill btw. /* whatis(high) */ /* ll low = 4, high = 4, mid, ans = -1; */ constexpr ll bad = 0x7f7f7f7f7f7f7f7f; while(low <= high){ mid = (low + high) >> 1; const auto norm = [&](ll f){ return f < 0 ? 0 : f > mid ? bad : f; }; /* whatis(mid) */ // todo n, ?k? constexpr; podzielne przez jakieś 8 najlepiej /* ll dp[mxk + 1]; */ /* alignas(64) ll dp[mxk + 1]; // o eps slower btw z alignas(64). */ /* alignas(16) ll dporig[mxk + 17]; // o eps slower btw z alignas(64). */ vector<pair<ll,int>> vec; vec.emplace_back(0, 0); /* alignas(64) ll dporig[mxk + 17]; // o eps slower btw z alignas(64). */ /* memset(dporig, 127, sizeof dporig); */ /* ll *dp = dporig + 16; */ /* alignas(16) ll dp[mxk + 1]; // o eps slower btw z alignas(64). */ // z 16 eq speed. /* ll dp[mxk + 1]; */ /* ll nxdp[mxk + 1]; */ /* memset(dp, -1, sizeof dp); */ // XXX don't forget this memset xddddd /* alignas(16) ll dp[mxk + 1]; // o eps slower btw z alignas(64). */ /* memset(dp, 127, sizeof dp); */ /* // and don't forget to set 0 as well xd. */ /* dp[0] = 0; */ /* int cl = -1, cr = -1; */ /* int cl = 0, cr = mxk; */ /* ll d1sum = 0, d2sum = 0; */ /* ll difsum = 0; */ /* vector<pair<ll,int>> nwvec; */ /* nwvec.reserve(100); */ #ifdef MXN for(int i = 0; i < mxn; ++i){ #else for(int i = 0; i < n; ++i){ #endif /* set<int> iledif; */ const int cra = a[i]; // k + 1 const int crb = b[i]; // k int nxk = mxk; /* for(; nxk > 3;){ */ vector<pair<ll,int>> nwvec; nwvec.emplace_back(norm(vec[0].first + crb), 0); /* nwvec = {{norm(vec[0].first + crb), 0}}; */ /* int cl = 0; */ auto nwvecemplace = [&](ll val, int tillr){ if(nwvec.back().first == val) nwvec.back().second = tillr; else nwvec.emplace_back(val, tillr); }; for(int i = 0; i < vec.size(); ++i){ /* if( */ /* if(cl != vec[i].second){ */ if(nwvec.back().second != vec[i].second){ /* nwvec.emplace_back(norm(vec[i].first + min(cra, crb)), vec[i].second); */ nwvecemplace(norm(vec[i].first + min(cra, crb)), vec[i].second); } if(vec[i].second != mxk){ if(i + 1 == vec.size()) /* nwvec.emplace_back(norm(vec[i].first + cra), vec[i].second + 1); */ nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1); else /* nwvec.emplace_back(norm(min(vec[i].first + cra, vec[i + 1].first + crb)), vec[i].second + 1); */ nwvecemplace(norm(min(vec[i].first + cra, vec[i + 1].first + crb)), vec[i].second + 1); } } vec = std::move(nwvec); whatis(vec) /* for(; nxk > 3;){ */ /* /1* for(; nxk >= 3;){ *1/ */ /* // wtf, tera jest 1.37s. czemu */ /* // ale zajebioza. */ /* // ale czemu, wczesniej tak dobrze nie było, miałem jakies 1.6s xd. */ /* // jest bez alignas btw. */ /* /1* __builtin_prefetch(&dp[nxk - 3], 1); *1/ */ /* // to samo, ale bez prefetchu 1.7s xd (vs 1.5s sam loop poniżej) */ /* __builtin_prefetch(&dp[nxk - 3], 1); */ /* /1* __builtin_prefetch(&dp[nxk - 3], 0); *1/ */ /* // ten sam czas gdy either 0 lub 1, ale 1.5s gdy oba. */ /* // n20k -> 6.6s no prefetch (and no this loop) vs 5.4s prefetch */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* --nxk; */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* --nxk; */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* --nxk; */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* --nxk; */ /* } */ /* for(; nxk > 0; --nxk){ */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* } */ /* dp[0] = dp[0] + crb; */ /* if(dp[0] < 0) */ /* dp[0] = 0; */ /* else if(dp[0] > mid) */ /* dp[0] = bad; */ /* for(int c = 0; c <= mxk; ++c) */ /* /1* for(int c = 0; c <= k; ++c) *1/ */ /* iledif.insert(dp[c]); */ /* /1* difsum += iledif.size(); *1/ */ /* whatis(iledif.size()) */ /* /1* /2* whatis(nwvec.size()) *2/ *1/ */ /* whatis(vec.size()) */ /* /1* whatis(vec) *1/ */ /* /1* whatis(iledif) *1/ */ /* /1* for(int c = 0; c <= k; ++c) *1/ */ /* /1* cout << dp[c] << ' '; *1/ */ /* /1* cout << endl; *1/ */ /* /1* assert(vec.size() == iledif.size()); *1/ */ /* if(vec.size() > iledif.size() + 4){ */ /* /1* if(vec.size() != iledif.size()){ *1/ */ /* whatis(vec) */ /* whatis(iledif) */ /* } */ // kozak, doesn't trigger now. } // not overkill at all. /* ll vecres = lower_bound(all(vec), k, [](auto &f, auto &s){return f < s.second;})->second; */ /* ll vecres = (upper_bound(all(vec), k, [](auto &f, auto &s){return s > f.second;}) - 1)->second; */ /* ll vecres = (upper_bound(all(vec), k, [](auto &f, auto &s){return f < s.second;}) - 1)->second; */ /* assert(vecres == dp[k]); */ ll vecres = -1; for(auto &i: vec){ if(i.second >= k){ vecres = i.first; break; } } assert(vecres != -1); /* assert(vecres == dp[k]); */ /* whatis(difsum) */ /* if(dp[k] <= mid){ */ whatis(vec) whatis(mid) whatis(vecres) whatis(vecres <= mid) if(vecres <= mid){ ans = mid; high = mid - 1; } else{ low = mid + 1; } } #ifdef PRINTS static vector<pair<int, bool>> p[mxn]; /* static vector<pair<ll, int>> pastvec[mxn]; */ { /* vector<pair<int, bool>> p[maxn + 1]; */ // omg, wait, czy bug przez to że tutaj porównywałem do mid vs ans? xdddd // (ale jak w legacy wersji było w takim razie dobrze, skoro tam tak samo..) // -> still źle, ale at least one future bug less. const auto norm = [&](ll f){ /* return f < 0 ? 0 : f > mid ? bad : f; */ return f < 0 ? 0 : f > ans ? bad : f; }; vector<pair<ll,int>> vec; vec.emplace_back(0, 0); #ifdef MXN for(int i = 0; i < mxn; ++i){ #else for(int i = 0; i < n; ++i){ #endif /* pastvec[i] = vec; */ /* set<int> iledif; */ const int cra = a[i]; // k + 1 const int crb = b[i]; // k int nxk = mxk; /* for(; nxk > 3;){ */ vector<pair<ll,int>> nwvec; nwvec.emplace_back(norm(vec[0].first + crb), 0); p[i].emplace_back(0, 0); /* int cl = 0; */ auto nwvecemplace = [&](ll val, int tillr, bool nwp){ /* whatis("--nwvecemplace") */ /* whatis(val) */ /* whatis(tillr) */ /* whatis(nwp) */ /* whatis("--nwvecemplace") */ // prevy chyba wciąż mogą być jednak inne, więc tak czy inaczej // muszę zrobić push_back. // ewentualnie opt że jak nwp równe p[i].back().second to nie muszę. // guess nawet oddzielnie to mogę btw. if(nwvec.back().first == val){ nwvec.back().second = tillr; /* p[i].emplace_back(tillr, nwp); */ } else{ nwvec.emplace_back(val, tillr); /* p[i].emplace_back(tillr, nwp); */ } if(p[i].back().second == nwp){ p[i].back().first = tillr; } else{ p[i].emplace_back(tillr, nwp); } }; /* whatis(cra) */ /* whatis(crb) */ for(int i = 0; i < vec.size(); ++i){ /* whatis(i) */ /* if( */ /* if(cl != vec[i].second){ */ if(nwvec.back().second != vec[i].second){ /* nwvec.emplace_back(norm(vec[i].first + min(cra, crb)), vec[i].second); */ // ohhh my goddd. // jakim cudem tak prosty bug zajął mi O(1h) na // znalezenie.... // patrzałem na wszystkie branche poza tym jednym ofc. /* nwvecemplace(norm(vec[i].first + min(cra, crb)), vec[i].second, 0); */ if(cra < crb) nwvecemplace(norm(vec[i].first + cra), vec[i].second, 1); else nwvecemplace(norm(vec[i].first + crb), vec[i].second, 0); /* p[i].emplace_back(vec[i].second, 0); */ } if(vec[i].second != mxk){ if(i + 1 == vec.size()){ /* nwvec.emplace_back(norm(vec[i].first + cra), vec[i].second + 1); */ /* nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1); */ nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1, 1); /* p[i].emplace_back(vec[i].second + 1, 1); */ } else{ /* nwvec.emplace_back(norm(min(vec[i].first + cra, vec[i + 1].first + crb)), vec[i].second + 1); */ if(vec[i].first + cra < vec[i + 1].first + crb){ /* nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1); */ /* whatis("b1") */ nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1, 1); /* p[i].emplace_back(vec[i].second + 1, 1); */ } else{ /* whatis("b2") */ nwvecemplace(norm(vec[i + 1].first + crb), vec[i].second + 1, 0); } } } } /* if(i <= n) */ /* for(int x = 0; x < p[i].size(); ++x){ */ /* for(int j = x ? p[i][x - 1].first + 1 : 0; j <= p[i][x].first; ++j){ */ /* assert(j <= mxk); */ /* if(!j) */ /* continue; */ /* ll nwvecres = -1; */ /* for(auto &i: nwvec){ */ /* if(i.second >= j){ */ /* nwvecres = i.first; */ /* break; */ /* } */ /* } */ /* ll vecm1 = -1; */ /* for(auto &i: vec){ */ /* if(i.second >= j - 1){ */ /* vecm1 = i.first; */ /* break; */ /* } */ /* } */ /* ll vecm0 = -1; */ /* for(auto &i: vec){ */ /* if(i.second >= j){ */ /* vecm0 = i.first; */ /* break; */ /* } */ /* } */ /* /1* whatis(i) *1/ */ /* /1* whatis(p[i]) *1/ */ /* /1* whatis(vec[i]) *1/ */ /* /1* whatis(nwvec[i]) *1/ */ /* /1* whatis(vec) *1/ */ /* /1* whatis(nwvec) *1/ */ /* /1* whatis(x) *1/ */ /* /1* whatis(j) *1/ */ /* assert(nwvecres != -1); */ /* ll expres = norm(p[i][x].second ? vecm1 + cra : vecm0 + crb); */ /* /1* whatis(expres) *1/ */ /* /1* whatis(nwvecres) *1/ */ /* assert(nwvecres == expres); */ /* } */ /* } */ vec = std::move(nwvec); } ll vecres = -1; for(auto &i: vec){ if(i.second >= k){ vecres = i.first; break; } } assert(vecres != -1); // hmm. assert(vecres <= ans); } /* static bool p[mxn + 1][mxk + 1]; */ /* { */ /* /1* static unsigned short p[mxn][mxk]; *1/ */ /* // todo bitset def. */ /* alignas(16) ll dp[mxk + 1]; // o eps slower btw z alignas(64). */ /* memset(dp, 127, sizeof dp); */ /* dp[0] = 0; */ /* #ifdef MXN */ /* for(int i = 0; i < mxn; ++i){ */ /* #else */ /* for(int i = 0; i < n; ++i){ */ /* #endif */ /* const int cra = a[i]; // k + 1 */ /* const int crb = b[i]; // k */ /* int nxk = mxk; */ /* for(; nxk > 0; --nxk){ */ /* /1* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); *1/ */ /* if(dp[nxk - 1] + cra < dp[nxk] + crb){ */ /* dp[nxk] = dp[nxk - 1] + cra; */ /* p[i][nxk] = 1; */ /* } */ /* else{ */ /* dp[nxk] = dp[nxk] + crb; */ /* } */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* } */ /* dp[0] = dp[0] + crb; */ /* if(dp[0] < 0) */ /* dp[0] = 0; */ /* else if(dp[0] > mid) */ /* dp[0] = bad; */ /* } */ /* } */ #endif cout << ans << '\n'; #ifdef PRINTS string s; int crk = k; /* for(int crn = n; crn > 0; --crn){ */ /* for(int crn = n - 1; crn >= 0; --crn){ */ /* s += "BA"[p[crn][crk]^swpd]; */ /* crk -= p[crn][crk]; */ /* } */ for(int crn = n - 1; crn >= 0; --crn){ ll vecres = -1; /* whatis(crn) */ /* whatis(crk) */ /* whatis(pastvec[crn]) */ /* whatis(p[crn]) */ for(auto &i: p[crn]){ if(i.first >= crk){ vecres = i.second; break; } } assert(vecres != -1); s += "BA"[vecres^swpd]; crk -= vecres; } reverse(all(s)); whatis(s) #ifdef VERIFY ll sum = 0; ll mx = 0; int ilek = 0; int it = 0; /* whatis(s) */ /* whatis(ans) */ FORR(i,s){ /* if((i == 'A') ^ swpd){ */ /* whatis(it) */ /* whatis(sum) */ if(i == 'A'){ /* whatis(_a[it]) */ /* if(!swpd) */ ++ilek; sum += _a[it]; } else{ /* whatis(_b[it]) */ /* if(swpd) */ /* ++ilek; */ sum += _b[it]; } /* whatis(it) */ /* whatis(sum) */ mx = max(mx, sum); sum = max(sum, 0l); ++it; } if(swpd) ilek = n - ilek; /* assert(sum == ans); */ /* whatis(mx) */ assert(mx == ans); assert(ilek == k); #endif /* if(0) */ cout << s << '\n'; #endif // 8m12s without opts and pragmas. }
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 | #pragma GCC optimize("O3,unroll-loops") /* #pragma GCC optimize("Ofast,unroll-loops") */ // wtf, tera jest 1.37s. czemu /* #pragma GCC optimize("O3,unroll-loops,prefetch-loop-arrays") */ // slower z tą pragmą bez własnych optow interestingly. // -mprefetchwt1 todo -> to czesc avx512 though? /* #pragma GCC target("prefetch-loop-arrays") */ /* #pragma GCC target("avx2") */ /* #pragma GCC optimize("Ofast,unroll-loops") */ #include <bits/stdc++.h> #include <ext/pb_ds/assoc_container.hpp> #include <ext/pb_ds/tree_policy.hpp> #define pb push_back #define mp make_pair #define all(a) begin(a),end(a) #define FOR(x,val,to) for(int x=(val);x<int((to));++x) #define FORE(x,val,to) for(auto x=(val);x<=(to);++x) #define FORR(x,arr) for(auto &x: arr) #define FORS(x,plus,arr) for(auto x = begin(arr)+(plus); x != end(arr); ++x) #define FORREV(x,plus,arr) for(auto x = (arr).rbegin()+(plus); x !=(arr).rend(); ++x) #define REE(s_) {cout<<s_<<'\n';exit(0);} #define GET(arr) for(auto &i: (arr)) sc(i) #define whatis(x) cerr << #x << " is " << (x) << endl; #define e1 first #define e2 second #define INF 0x7f7f7f7f typedef std::pair<int,int> pi; typedef std::vector<int> vi; typedef std::vector<std::string> vs; typedef int64_t ll; typedef uint64_t ull; #define umap unordered_map #define uset unordered_set using namespace std; using namespace __gnu_pbds; #ifdef ONLINE_JUDGE #define whatis(x) ; #endif #ifdef _WIN32 #define getchar_unlocked() _getchar_nolock() #define _CRT_DISABLE_PERFCRIT_LOCKS #endif template<class L, class R> ostream& operator<<(ostream &os, map<L, R> P) { for(auto const &vv: P)os<<"("<<vv.first<<","<<vv.second<<")"; return os; } template<class T> ostream& operator<<(ostream &os, set<T> V) { os<<"[";for(auto const &vv:V)os<<vv<<","; os<<"]"; return os; } template<class T> ostream& operator<<(ostream &os, vector<T> V) { os<<"[";for(auto const &vv:V)os<<vv<<","; os<<"]"; return os; } template<class L, class R> ostream& operator<<(ostream &os, pair<L, R> P) { os<<"("<<P.first<<","<<P.second<<")"; return os; } inline int fstoi(const string &str){auto it=str.begin();bool neg=0;int num=0;if(*it=='-')neg=1;else num=*it-'0';++it;while(it<str.end()) num=num*10+(*it++-'0');if(neg)num*=-1;return num;} inline void getch(char &x){while(x = getchar_unlocked(), x < 33){;}} inline void getstr(string &str){str.clear(); char cur;while(cur=getchar_unlocked(),cur<33){;}while(cur>32){str+=cur;cur=getchar_unlocked();}} template<typename T> inline bool sc(T &num){ bool neg=0; int c; num=0; while(c=getchar_unlocked(),c<33){if(c == EOF) return false;} if(c=='-'){ neg=1; c=getchar_unlocked(); } for(;c>47;c=getchar_unlocked()) num=num*10+c-48; if(neg) num*=-1; return true;}template<typename T, typename ...Args> inline void sc(T &num, Args &...args){ bool neg=0; int c; num=0; while(c=getchar_unlocked(),c<33){;} if(c=='-'){ neg=1; c=getchar_unlocked(); } for(;c>47;c=getchar_unlocked()) num=num*10+c-48; if(neg) num*=-1; sc(args...); } template<typename T> using ordered_set = tree<T, null_type, less<T>, rb_tree_tag, tree_order_statistics_node_update>; //s.find_by_order(), s.order_of_key() <- works like lower_bound template<typename T> using ordered_map = tree<T, int, less<T>, rb_tree_tag, tree_order_statistics_node_update>; #define N 1000001 /* #define MXN */ // wa z tym^^ for some reason.. // -> przez to że dodawałem -1 zamiast jakiegoś -LL_MAX / 2. // guess best just wywalić MXN. #define PRINTS #define assert(x) ; /* #define VERIFY */ #define whatis(x) ; int main(){ ios_base::sync_with_stdio(0);cin.tie(0); /* assert(0); */ /* constexpr int mxn = 10000; */ /* constexpr int mxn = 10000, mxk = 5000; */ /* constexpr int mxn = 10000; */ /* constexpr int mxn = 20000; */ /* constexpr int mxk = 32 * 313 + 3; */ /* constexpr int mxk = 5000; */ /* constexpr int mxk = 16 * 312 + 3; */ // wtf, czemu przez wiekszy mxk nie działa ? xd // -> to było < 5000 xdd. /* constexpr int mxn = 10000; */ /* constexpr int mxk = 16 * 313 + 3; */ /* constexpr int mxk = 16 * 313 + 3; */ /* constexpr int mxk = 16 * 313 + 1; */ // +1, +3 both dobrze. // +2, +4 bad. // todo podzielne przez 64? constexpr int mxn = 100000; constexpr int mxk = 50000; int n,k; sc(n,k); #ifdef MXN /* int _a[mxn], _b[mxn]; */ int _a[mxn], _b[mxn]; // XDDD, kox bug też. memset(_a,-1,sizeof _a); memset(_b,-1,sizeof _b); FOR(i,0,n){ sc(_a[i]); } FOR(i,0,n){ sc(_b[i]); } #else int _a[n], _b[n]; GET(_a); GET(_b); #endif // xd. /* const int *a = a; */ /* const int *b = b; */ const int *a = _a; const int *b = _b; bool swpd = 0; if(k > n - k){ k = n - k; swap(a, b); swpd = 1; } // todo opt z high = sumą maxvali. // -> kox opt. /* ll low = 0, high = n * (ll)1000000000, mid, ans; */ // ??? high = 0 ll low = 0, high = 0, mid, ans; // XDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD /* FOR(i,0,n) */ //* FOR(i,0,n) */ /* high += max(a[i], b[i]); // down to 1.5s from 1.7s with this opt. */ FOR(i,0,n) high += max({0, a[i], b[i]}); // not overkill btw. /* whatis(high) */ /* ll low = 4, high = 4, mid, ans = -1; */ constexpr ll bad = 0x7f7f7f7f7f7f7f7f; while(low <= high){ mid = (low + high) >> 1; const auto norm = [&](ll f){ return f < 0 ? 0 : f > mid ? bad : f; }; /* whatis(mid) */ // todo n, ?k? constexpr; podzielne przez jakieś 8 najlepiej /* ll dp[mxk + 1]; */ /* alignas(64) ll dp[mxk + 1]; // o eps slower btw z alignas(64). */ /* alignas(16) ll dporig[mxk + 17]; // o eps slower btw z alignas(64). */ vector<pair<ll,int>> vec; vec.emplace_back(0, 0); /* alignas(64) ll dporig[mxk + 17]; // o eps slower btw z alignas(64). */ /* memset(dporig, 127, sizeof dporig); */ /* ll *dp = dporig + 16; */ /* alignas(16) ll dp[mxk + 1]; // o eps slower btw z alignas(64). */ // z 16 eq speed. /* ll dp[mxk + 1]; */ /* ll nxdp[mxk + 1]; */ /* memset(dp, -1, sizeof dp); */ // XXX don't forget this memset xddddd /* alignas(16) ll dp[mxk + 1]; // o eps slower btw z alignas(64). */ /* memset(dp, 127, sizeof dp); */ /* // and don't forget to set 0 as well xd. */ /* dp[0] = 0; */ /* int cl = -1, cr = -1; */ /* int cl = 0, cr = mxk; */ /* ll d1sum = 0, d2sum = 0; */ /* ll difsum = 0; */ /* vector<pair<ll,int>> nwvec; */ /* nwvec.reserve(100); */ #ifdef MXN for(int i = 0; i < mxn; ++i){ #else for(int i = 0; i < n; ++i){ #endif /* set<int> iledif; */ const int cra = a[i]; // k + 1 const int crb = b[i]; // k int nxk = mxk; /* for(; nxk > 3;){ */ vector<pair<ll,int>> nwvec; nwvec.emplace_back(norm(vec[0].first + crb), 0); /* nwvec = {{norm(vec[0].first + crb), 0}}; */ /* int cl = 0; */ auto nwvecemplace = [&](ll val, int tillr){ if(nwvec.back().first == val) nwvec.back().second = tillr; else nwvec.emplace_back(val, tillr); }; for(int i = 0; i < vec.size(); ++i){ /* if( */ /* if(cl != vec[i].second){ */ if(nwvec.back().second != vec[i].second){ /* nwvec.emplace_back(norm(vec[i].first + min(cra, crb)), vec[i].second); */ nwvecemplace(norm(vec[i].first + min(cra, crb)), vec[i].second); } if(vec[i].second != mxk){ if(i + 1 == vec.size()) /* nwvec.emplace_back(norm(vec[i].first + cra), vec[i].second + 1); */ nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1); else /* nwvec.emplace_back(norm(min(vec[i].first + cra, vec[i + 1].first + crb)), vec[i].second + 1); */ nwvecemplace(norm(min(vec[i].first + cra, vec[i + 1].first + crb)), vec[i].second + 1); } } vec = std::move(nwvec); whatis(vec) /* for(; nxk > 3;){ */ /* /1* for(; nxk >= 3;){ *1/ */ /* // wtf, tera jest 1.37s. czemu */ /* // ale zajebioza. */ /* // ale czemu, wczesniej tak dobrze nie było, miałem jakies 1.6s xd. */ /* // jest bez alignas btw. */ /* /1* __builtin_prefetch(&dp[nxk - 3], 1); *1/ */ /* // to samo, ale bez prefetchu 1.7s xd (vs 1.5s sam loop poniżej) */ /* __builtin_prefetch(&dp[nxk - 3], 1); */ /* /1* __builtin_prefetch(&dp[nxk - 3], 0); *1/ */ /* // ten sam czas gdy either 0 lub 1, ale 1.5s gdy oba. */ /* // n20k -> 6.6s no prefetch (and no this loop) vs 5.4s prefetch */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* --nxk; */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* --nxk; */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* --nxk; */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* --nxk; */ /* } */ /* for(; nxk > 0; --nxk){ */ /* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* } */ /* dp[0] = dp[0] + crb; */ /* if(dp[0] < 0) */ /* dp[0] = 0; */ /* else if(dp[0] > mid) */ /* dp[0] = bad; */ /* for(int c = 0; c <= mxk; ++c) */ /* /1* for(int c = 0; c <= k; ++c) *1/ */ /* iledif.insert(dp[c]); */ /* /1* difsum += iledif.size(); *1/ */ /* whatis(iledif.size()) */ /* /1* /2* whatis(nwvec.size()) *2/ *1/ */ /* whatis(vec.size()) */ /* /1* whatis(vec) *1/ */ /* /1* whatis(iledif) *1/ */ /* /1* for(int c = 0; c <= k; ++c) *1/ */ /* /1* cout << dp[c] << ' '; *1/ */ /* /1* cout << endl; *1/ */ /* /1* assert(vec.size() == iledif.size()); *1/ */ /* if(vec.size() > iledif.size() + 4){ */ /* /1* if(vec.size() != iledif.size()){ *1/ */ /* whatis(vec) */ /* whatis(iledif) */ /* } */ // kozak, doesn't trigger now. } // not overkill at all. /* ll vecres = lower_bound(all(vec), k, [](auto &f, auto &s){return f < s.second;})->second; */ /* ll vecres = (upper_bound(all(vec), k, [](auto &f, auto &s){return s > f.second;}) - 1)->second; */ /* ll vecres = (upper_bound(all(vec), k, [](auto &f, auto &s){return f < s.second;}) - 1)->second; */ /* assert(vecres == dp[k]); */ ll vecres = -1; for(auto &i: vec){ if(i.second >= k){ vecres = i.first; break; } } assert(vecres != -1); /* assert(vecres == dp[k]); */ /* whatis(difsum) */ /* if(dp[k] <= mid){ */ whatis(vec) whatis(mid) whatis(vecres) whatis(vecres <= mid) if(vecres <= mid){ ans = mid; high = mid - 1; } else{ low = mid + 1; } } #ifdef PRINTS static vector<pair<int, bool>> p[mxn]; /* static vector<pair<ll, int>> pastvec[mxn]; */ { /* vector<pair<int, bool>> p[maxn + 1]; */ // omg, wait, czy bug przez to że tutaj porównywałem do mid vs ans? xdddd // (ale jak w legacy wersji było w takim razie dobrze, skoro tam tak samo..) // -> still źle, ale at least one future bug less. const auto norm = [&](ll f){ /* return f < 0 ? 0 : f > mid ? bad : f; */ return f < 0 ? 0 : f > ans ? bad : f; }; vector<pair<ll,int>> vec; vec.emplace_back(0, 0); #ifdef MXN for(int i = 0; i < mxn; ++i){ #else for(int i = 0; i < n; ++i){ #endif /* pastvec[i] = vec; */ /* set<int> iledif; */ const int cra = a[i]; // k + 1 const int crb = b[i]; // k int nxk = mxk; /* for(; nxk > 3;){ */ vector<pair<ll,int>> nwvec; nwvec.emplace_back(norm(vec[0].first + crb), 0); p[i].emplace_back(0, 0); /* int cl = 0; */ auto nwvecemplace = [&](ll val, int tillr, bool nwp){ /* whatis("--nwvecemplace") */ /* whatis(val) */ /* whatis(tillr) */ /* whatis(nwp) */ /* whatis("--nwvecemplace") */ // prevy chyba wciąż mogą być jednak inne, więc tak czy inaczej // muszę zrobić push_back. // ewentualnie opt że jak nwp równe p[i].back().second to nie muszę. // guess nawet oddzielnie to mogę btw. if(nwvec.back().first == val){ nwvec.back().second = tillr; /* p[i].emplace_back(tillr, nwp); */ } else{ nwvec.emplace_back(val, tillr); /* p[i].emplace_back(tillr, nwp); */ } if(p[i].back().second == nwp){ p[i].back().first = tillr; } else{ p[i].emplace_back(tillr, nwp); } }; /* whatis(cra) */ /* whatis(crb) */ for(int i = 0; i < vec.size(); ++i){ /* whatis(i) */ /* if( */ /* if(cl != vec[i].second){ */ if(nwvec.back().second != vec[i].second){ /* nwvec.emplace_back(norm(vec[i].first + min(cra, crb)), vec[i].second); */ // ohhh my goddd. // jakim cudem tak prosty bug zajął mi O(1h) na // znalezenie.... // patrzałem na wszystkie branche poza tym jednym ofc. /* nwvecemplace(norm(vec[i].first + min(cra, crb)), vec[i].second, 0); */ if(cra < crb) nwvecemplace(norm(vec[i].first + cra), vec[i].second, 1); else nwvecemplace(norm(vec[i].first + crb), vec[i].second, 0); /* p[i].emplace_back(vec[i].second, 0); */ } if(vec[i].second != mxk){ if(i + 1 == vec.size()){ /* nwvec.emplace_back(norm(vec[i].first + cra), vec[i].second + 1); */ /* nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1); */ nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1, 1); /* p[i].emplace_back(vec[i].second + 1, 1); */ } else{ /* nwvec.emplace_back(norm(min(vec[i].first + cra, vec[i + 1].first + crb)), vec[i].second + 1); */ if(vec[i].first + cra < vec[i + 1].first + crb){ /* nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1); */ /* whatis("b1") */ nwvecemplace(norm(vec[i].first + cra), vec[i].second + 1, 1); /* p[i].emplace_back(vec[i].second + 1, 1); */ } else{ /* whatis("b2") */ nwvecemplace(norm(vec[i + 1].first + crb), vec[i].second + 1, 0); } } } } /* if(i <= n) */ /* for(int x = 0; x < p[i].size(); ++x){ */ /* for(int j = x ? p[i][x - 1].first + 1 : 0; j <= p[i][x].first; ++j){ */ /* assert(j <= mxk); */ /* if(!j) */ /* continue; */ /* ll nwvecres = -1; */ /* for(auto &i: nwvec){ */ /* if(i.second >= j){ */ /* nwvecres = i.first; */ /* break; */ /* } */ /* } */ /* ll vecm1 = -1; */ /* for(auto &i: vec){ */ /* if(i.second >= j - 1){ */ /* vecm1 = i.first; */ /* break; */ /* } */ /* } */ /* ll vecm0 = -1; */ /* for(auto &i: vec){ */ /* if(i.second >= j){ */ /* vecm0 = i.first; */ /* break; */ /* } */ /* } */ /* /1* whatis(i) *1/ */ /* /1* whatis(p[i]) *1/ */ /* /1* whatis(vec[i]) *1/ */ /* /1* whatis(nwvec[i]) *1/ */ /* /1* whatis(vec) *1/ */ /* /1* whatis(nwvec) *1/ */ /* /1* whatis(x) *1/ */ /* /1* whatis(j) *1/ */ /* assert(nwvecres != -1); */ /* ll expres = norm(p[i][x].second ? vecm1 + cra : vecm0 + crb); */ /* /1* whatis(expres) *1/ */ /* /1* whatis(nwvecres) *1/ */ /* assert(nwvecres == expres); */ /* } */ /* } */ vec = std::move(nwvec); } ll vecres = -1; for(auto &i: vec){ if(i.second >= k){ vecres = i.first; break; } } assert(vecres != -1); // hmm. assert(vecres <= ans); } /* static bool p[mxn + 1][mxk + 1]; */ /* { */ /* /1* static unsigned short p[mxn][mxk]; *1/ */ /* // todo bitset def. */ /* alignas(16) ll dp[mxk + 1]; // o eps slower btw z alignas(64). */ /* memset(dp, 127, sizeof dp); */ /* dp[0] = 0; */ /* #ifdef MXN */ /* for(int i = 0; i < mxn; ++i){ */ /* #else */ /* for(int i = 0; i < n; ++i){ */ /* #endif */ /* const int cra = a[i]; // k + 1 */ /* const int crb = b[i]; // k */ /* int nxk = mxk; */ /* for(; nxk > 0; --nxk){ */ /* /1* dp[nxk] = min(dp[nxk] + crb, dp[nxk - 1] + cra); *1/ */ /* if(dp[nxk - 1] + cra < dp[nxk] + crb){ */ /* dp[nxk] = dp[nxk - 1] + cra; */ /* p[i][nxk] = 1; */ /* } */ /* else{ */ /* dp[nxk] = dp[nxk] + crb; */ /* } */ /* if(dp[nxk] < 0) */ /* dp[nxk] = 0; */ /* else if(dp[nxk] > mid) */ /* dp[nxk] = bad; */ /* } */ /* dp[0] = dp[0] + crb; */ /* if(dp[0] < 0) */ /* dp[0] = 0; */ /* else if(dp[0] > mid) */ /* dp[0] = bad; */ /* } */ /* } */ #endif cout << ans << '\n'; #ifdef PRINTS string s; int crk = k; /* for(int crn = n; crn > 0; --crn){ */ /* for(int crn = n - 1; crn >= 0; --crn){ */ /* s += "BA"[p[crn][crk]^swpd]; */ /* crk -= p[crn][crk]; */ /* } */ for(int crn = n - 1; crn >= 0; --crn){ ll vecres = -1; /* whatis(crn) */ /* whatis(crk) */ /* whatis(pastvec[crn]) */ /* whatis(p[crn]) */ for(auto &i: p[crn]){ if(i.first >= crk){ vecres = i.second; break; } } assert(vecres != -1); s += "BA"[vecres^swpd]; crk -= vecres; } reverse(all(s)); whatis(s) #ifdef VERIFY ll sum = 0; ll mx = 0; int ilek = 0; int it = 0; /* whatis(s) */ /* whatis(ans) */ FORR(i,s){ /* if((i == 'A') ^ swpd){ */ /* whatis(it) */ /* whatis(sum) */ if(i == 'A'){ /* whatis(_a[it]) */ /* if(!swpd) */ ++ilek; sum += _a[it]; } else{ /* whatis(_b[it]) */ /* if(swpd) */ /* ++ilek; */ sum += _b[it]; } /* whatis(it) */ /* whatis(sum) */ mx = max(mx, sum); sum = max(sum, 0l); ++it; } if(swpd) ilek = n - ilek; /* assert(sum == ans); */ /* whatis(mx) */ assert(mx == ans); assert(ilek == k); #endif /* if(0) */ cout << s << '\n'; #endif // 8m12s without opts and pragmas. } |