fork download
copy 
  1. #include <bits/stdc++.h>
  2. using namespace std;
  3.  
  4. using ull = uint64_t;
  5. using ll = int64_t;
  6.  
  7. template<typename complex_t>
  8. class Fft{
  9.     static constexpr double PI(){return 3.1415926535897932384626;};
  10.     static constexpr int SPLIT_BITS = 15;
  11.     using cvec = std::vector<complex_t>;
  12.     using ivec = std::vector<int64_t>;
  13.  
  14.     static int lg(int x){
  15.         return __builtin_clz(1) - __builtin_clz(x);
  16.     }
  17.     static int next_2pow(int x){
  18.         return x>1 ? 1<<lg(2*x-1): 1;
  19.     }
  20.  
  21.     static std::vector<complex_t> roots;
  22.     static void gen_roots(int const&depth){
  23.         if(roots.size() < (1u<<depth)){
  24.             if(roots.empty()){
  25.                 roots.emplace_back(std::polar(0.0, 0.0));
  26.                 roots.emplace_back(std::polar(1.0, 0.0));
  27.             }
  28.             int cur_d = lg(roots.size())-1;
  29.             roots.resize(1<<(depth+1));
  30.             for(int i = cur_d;i<depth;++i){
  31.                 int n = 1 << i;
  32.                 complex_t ome = std::polar(1.0, PI()/2.0/n);
  33.                 for(int j=n;j<2*n;++j){
  34.                     roots[2*j] = roots[j];
  35.                     roots[2*j+1] = roots[j]*ome;
  36.                 }
  37.             }
  38.         }
  39.     }
  40.     static void bitflip(cvec &v){
  41.         int n = v.size();
  42. 		for(int i=1, j=0;i<n;++i){
  43. 			int m = n >> 1;
  44. 			for(;j>=m;m >>= 1)
  45. 				j -= m;
  46. 			j += m;
  47. 			if(i < j) std::swap(v[i], v[j]);
  48. 		}
  49.     }
  50.     template<bool is_rev>
  51.     static void fft_regular(cvec& vec){
  52.         bitflip(vec);
  53.         int n = vec.size();
  54.         gen_roots(lg(n));
  55.         for(int iter=1, sh=0;iter<n;iter*=2, ++sh){
  56. 			for(int x=0;x<n;x+=2*iter){
  57. 				for(int y=0;y<iter;++y){
  58. 					complex_t ome = roots[(1 << sh) + y];
  59. 					if (is_rev) ome = conj(ome);
  60. 					complex_t v = vec[x+y], w=vec[x+y+iter];
  61. 					vec[x+y] = v+ome*w;
  62. 					vec[x+y+iter] = v-ome*w;
  63. 				}
  64. 			}
  65. 		}
  66.     }
  67.     template<bool is_rev>
  68.     static void fft_inplace(cvec&v){
  69.         fft_regular<is_rev>(v);
  70.     }
  71. public:
  72.     static void three_mul(ivec &vout, ivec const&v1, ivec const& v2, const size_t mod = 0){
  73.         constexpr int BITS = 17, MASK = (1<<BITS) - 1;
  74.         const size_t out_size = v1.size() + v2.size() - 1;
  75.         const size_t n = next_2pow(out_size);
  76.  
  77.         cvec c1(n, complex_t(0.0, 0.0)), c2(n, complex_t(0.0, 0.0)), c3(n, complex_t(0.0, 0.0));
  78.         for(size_t i=0;i<v1.size();++i){
  79.             int64_t a = v1[i] >> (2*BITS), b = (v1[i] >> BITS) & MASK, c = v1[i] & MASK;
  80.             if(MASK-c < c){
  81.                 c-= 1<<BITS;
  82.                 ++b;
  83.             }
  84.             if(MASK-b < b){
  85.                 b-= 1<<BITS;
  86.                 ++a;
  87.             }
  88.             c1[i].real(a);
  89.             c1[i].imag(b);
  90.             c2[i].real(c);
  91.         }
  92.         for(size_t i=0;i<v2.size();++i){
  93.             int64_t a = v2[i] >> (2*BITS), b = (v2[i] >> BITS) & MASK, c = v2[i] & MASK;
  94.             if(MASK-c < c){
  95.                 c-= 1<<BITS;
  96.                 ++b;
  97.             }
  98.             if(MASK-b < b){
  99.                 b-= 1<<BITS;
  100.                 ++a;
  101.             }
  102.             c2[i].imag(a);
  103.             c3[i].real(b);
  104.             c3[i].imag(c);
  105.         }
  106.         fft_inplace<false>(c1);
  107.         fft_inplace<false>(c2);
  108.         fft_inplace<false>(c3);
  109.  
  110.         const double factor = 0.25 / static_cast<double>(c1.size());
  111.         const complex_t IMAG (0.0, 1.0);
  112.         const complex_t NIMAG = conj(IMAG);
  113.  
  114.         for(int i=0;i<=n/2;++i){
  115.             int j = (n-i)&(n-1); //reverse index
  116.             complex_t rx = (c1[i] + conj(c1[j]));
  117.             complex_t ix = (c1[i] - conj(c1[j])) * NIMAG;
  118.             complex_t ry = (c2[i] + conj(c2[j]));
  119.             complex_t iy = (c2[i] - conj(c2[j])) * NIMAG;
  120.             complex_t rz = (c3[i] + conj(c3[j]));
  121.             complex_t iz = (c3[i] - conj(c3[j])) * NIMAG;
  122.             //cerr << rx << " " << ix << " " << ry << "\n";
  123.             c1[i] = (ry*iz + (ix*iz + ry*rz) * IMAG) * factor;
  124.             c2[i] = (rx*iz + ix*rz + ry*iy) * factor;
  125.             c1[j] = conj(ry*iz + (ix*iz + ry*rz) * NIMAG) * factor;
  126.             c2[j] = conj(rx*iz + ix*rz + ry*iy) * factor;
  127.         }
  128.         fft_inplace<true>(c1);
  129.         fft_inplace<true>(c2);
  130.         vout.resize(out_size);
  131.         //cerr << c1[0] << " " << c2[0] << "\n";
  132.         for(int i=0;i<out_size;++i){
  133.             uint64_t a = llround(c1[i].real()), b = llround(c1[i].imag()), c = llround(c2[i].real());
  134.             vout[i] = (a + (b<<BITS) + (c<<(2*BITS))) % mod;
  135.         }
  136.  
  137.     }
  138. };
  139. template<typename complex_t>
  140. std::vector<complex_t> Fft<complex_t>::roots;
  141.  
  142. using fft = Fft<std::complex<double>>;
  143.  
  144.  
  145. template<int64_t mod>
  146. struct NT64 {
  147.     static_assert(mod >= 0, "mod should be non-negative.");
  148.     static int64_t add(int64_t const&a, int64_t const&b){
  149.         int64_t ret = a+b;
  150.         if(ret>=mod) ret-=mod;
  151.         return ret;
  152.     }
  153.     static int64_t& xadd(int64_t& a, int64_t const&b){
  154.         a+=b;
  155.         if(a>=mod) a-=mod;
  156.         return a;
  157.     }
  158.     static int64_t sub(int64_t const&a, int64_t const&b){
  159.         return add(a, mod-b);
  160.     }
  161.     static int64_t& xsub(int64_t& a, int64_t const&b){
  162.         return xadd(a, mod-b);
  163.     }
  164.     static int64_t mul(int64_t const&a, int64_t const&b){
  165.         return a*static_cast<int64_t>(b)%mod;
  166.     }
  167.     static int64_t& xmul(int64_t &a, int64_t const&b){
  168.         return a=mul(a, b);
  169.     }
  170.     static int64_t inv_rec(int64_t const&a, int64_t const&m){
  171.         assert(a!=0);
  172.         if(a==1) return 1;
  173.         int64_t ret = m+(1-inv_rec(m%a, a)*static_cast<int64_t>(m))/a;
  174.         return ret;
  175.     }
  176. 	// this is soooo great, can even be used for a sieve
  177.     static int64_t inv_rec_2(int64_t const&a, int64_t const&m){
  178.         assert(a!=0);
  179.         if(a==1) return 1;
  180.         int64_t ret = m-mul((m/a), inv_rec_2(m%a, m));
  181.         return ret;
  182.     }
  183.     static int64_t inv(int64_t const&a){
  184.         return inv_rec(a, mod);
  185.     }
  186.     static int64_t div(int64_t const&a, int64_t const&b){
  187.         return mul(a, inv(b));
  188.     }
  189. };
  190.  
  191.  
  192. using nt = NT64<1ll<<32>;
  193.  
  194.  
  195.  
  196. const int max_log = 32 + 18;
  197. const ll MODMOD = 1ll << max_log;
  198.  
  199.  
  200. int lg(int x){
  201.     return __builtin_clz(1) - __builtin_clz(x);
  202. }
  203.  
  204.  
  205. vector<unsigned int> facts, ifacts, v2;
  206.  
  207. void precalc_facts(int lim){
  208.     facts.resize(lim, 1);
  209.     ifacts.resize(lim, 1);
  210.     v2.resize(lim, 0);
  211.     for(int i=1;i<lim;++i){
  212.         const int j = i >> __builtin_ctz(i);
  213.         v2[i] = v2[i-1] + __builtin_ctz(i);
  214.         facts[i] = facts[i-1] * j;
  215.     }
  216.     ifacts.back() = nt::inv(facts.back());
  217.     for(int i=lim-1;i>0;--i){
  218.         const int j = i >> __builtin_ctz(i);
  219.         ifacts[i-1] = ifacts[i] * j;
  220.     }
  221. }
  222.  
  223.  
  224. void twofy(vector<ll> &v){
  225.     for(size_t i=0;i<v.size();++i){
  226.         v[i] *= 1ll<<(i-v2[i]);
  227.         v[i] *= ifacts[i];
  228.         v[i] %= MODMOD;
  229.     }
  230. }
  231.  
  232.  
  233. void vec_mul(vector<ll> &c, vector<ll> const&a, vector<ll> const&b){
  234.     const int bits = 25;
  235.     fft::three_mul(c, a, b, MODMOD);
  236. }
  237.  
  238. signed main()
  239. {
  240.     cin.tie(0); cout.tie(0); ios_base::sync_with_stdio(false);
  241.     precalc_facts(2e5+10);
  242.     int n;
  243.     cin >> n;
  244.     ++n;
  245.     const int m = lg(n)+1;
  246.     vector<ll> a(n), b(n), c;
  247.     for(int i=0;i<n;++i)
  248.         cin >> a[i];
  249.     for(int i=0;i<n;++i)
  250.         cin >> b[i];
  251.     twofy(a);
  252.     twofy(b);
  253.     vec_mul(c, a, b);
  254.     for(int i=0;i<n;++i){
  255.         ll val = c[i];
  256.         val >>= (i-v2[i]);
  257.         uint32_t out = val;
  258.         out*= facts[i];
  259.         cout << out << " ";
  260.     }
  261.     cout << "\n";
  262.     return 0;
  263. }
  264.  
Success #stdin #stdout 0s 15248KB
comments (?)
stdin
copy 
5
0 1 2 3 4 5
6 7 8 9 10 11
stdout
copy 
0 6 26 84 240 640
https://ideone.com/hkdjI4
language:
C++14 (gcc 8.3)
created:
6 years ago
visibility:
secret

Share or Embed source code

Discover > Sphere Engine API

The brand new service which powers Ideone!

Discover > IDE Widget

Widget for compiling and running the source code in a web browser!