import sys readline=sys.stdin.readline from collections import deque mod = 998244353 imag = 911660635 iimag = 86583718 rate2 = (911660635, 509520358, 369330050, 332049552, 983190778, 123842337, 238493703, 975955924, 603855026, 856644456, 131300601, 842657263, 730768835, 942482514, 806263778, 151565301, 510815449, 503497456, 743006876, 741047443, 56250497, 867605899) irate2 = (86583718, 372528824, 373294451, 645684063, 112220581, 692852209, 155456985, 797128860, 90816748, 860285882, 927414960, 354738543, 109331171, 293255632, 535113200, 308540755, 121186627, 608385704, 438932459, 359477183, 824071951, 103369235) rate3 = (372528824, 337190230, 454590761, 816400692, 578227951, 180142363, 83780245, 6597683, 70046822, 623238099, 183021267, 402682409, 631680428, 344509872, 689220186, 365017329, 774342554, 729444058, 102986190, 128751033, 395565204) irate3 = (509520358, 929031873, 170256584, 839780419, 282974284, 395914482, 444904435, 72135471, 638914820, 66769500, 771127074, 985925487, 262319669, 262341272, 625870173, 768022760, 859816005, 914661783, 430819711, 272774365, 530924681) def butterfly(a): n = len(a) h = (n - 1).bit_length() len_ = 0 while len_ < h: if h - len_ == 1: p = 1 << (h - len_ - 1) rot = 1 for s in range(1 << len_): offset = s << (h - len_) for i in range(p): l = a[i + offset] r = a[i + offset + p] * rot % mod a[i + offset] = (l + r) % mod a[i + offset + p] = (l - r) % mod if s + 1 != 1 << len_: rot *= rate2[(~s & -~s).bit_length() - 1] rot %= mod len_ += 1 else: p = 1 << (h - len_ - 2) rot = 1 for s in range(1 << len_): rot2 = rot * rot % mod rot3 = rot2 * rot % mod offset = s << (h - len_) for i in range(p): a0 = a[i + offset] a1 = a[i + offset + p] * rot a2 = a[i + offset + p * 2] * rot2 a3 = a[i + offset + p * 3] * rot3 a1na3imag = (a1 - a3) % mod * imag a[i + offset] = (a0 + a2 + a1 + a3) % mod a[i + offset + p] = (a0 + a2 - a1 - a3) % mod a[i + offset + p * 2] = (a0 - a2 + a1na3imag) % mod a[i + offset + p * 3] = (a0 - a2 - a1na3imag) % mod if s + 1 != 1 << len_: rot *= rate3[(~s & -~s).bit_length() - 1] rot %= mod len_ += 2 def butterfly_inv(a): n = len(a) h = (n - 1).bit_length() len_ = h while len_: if len_ == 1: p = 1 << (h - len_) irot = 1 for s in range(1 << (len_ - 1)): offset = s << (h - len_ + 1) for i in range(p): l = a[i + offset] r = a[i + offset + p] a[i + offset] = (l + r) % mod a[i + offset + p] = (l - r) * irot % mod if s + 1 != (1 << (len_ - 1)): irot *= irate2[(~s & -~s).bit_length() - 1] irot %= mod len_ -= 1 else: p = 1 << (h - len_) irot = 1 for s in range(1 << (len_ - 2)): irot2 = irot * irot % mod irot3 = irot2 * irot % mod offset = s << (h - len_ + 2) for i in range(p): a0 = a[i + offset] a1 = a[i + offset + p] a2 = a[i + offset + p * 2] a3 = a[i + offset + p * 3] a2na3iimag = (a2 - a3) * iimag % mod a[i + offset] = (a0 + a1 + a2 + a3) % mod a[i + offset + p] = (a0 - a1 + a2na3iimag) * irot % mod a[i + offset + p * 2] = (a0 + a1 - a2 - a3) * irot2 % mod a[i + offset + p * 3] = (a0 - a1 - a2na3iimag) * irot3 % mod if s + 1 != (1 << (len_ - 2)): irot *= irate3[(~s & -~s).bit_length() - 1] irot %= mod len_ -= 2 def convolution_naive(a, b): n = len(a) m = len(b) ans = [0] * (n + m - 1) if n < m: for j in range(m): for i in range(n): ans[i + j] = (ans[i + j] + a[i] * b[j]) % mod else: for i in range(n): for j in range(m): ans[i + j] = (ans[i + j] + a[i] * b[j]) % mod return ans def convolution_ntt(a, b): a = a.copy() b = b.copy() n = len(a) m = len(b) z = 1 << (n + m - 2).bit_length() a += [0] * (z - n) butterfly(a) b += [0] * (z - m) butterfly(b) for i in range(z): a[i] = a[i] * b[i] % mod butterfly_inv(a) a = a[:n + m - 1] iz = pow(z, mod - 2, mod) for i in range(n + m - 1): a[i] = a[i] * iz % mod return a def convolution_square(a): a = a.copy() n = len(a) z = 1 << (2 * n - 2).bit_length() a += [0] * (z - n) butterfly(a) for i in range(z): a[i] = a[i] * a[i] % mod butterfly_inv(a) a = a[:2 * n - 1] iz = pow(z, mod - 2, mod) for i in range(2 * n - 1): a[i] = a[i] * iz % mod return a def convolution(a, b): """It calculates (+, x) convolution in mod 998244353. Given two arrays a[0], a[1], ..., a[n - 1] and b[0], b[1], ..., b[m - 1], it calculates the array c of length n + m - 1, defined by > c[i] = sum(a[j] * b[i - j] for j in range(i + 1)) % 998244353. It returns an empty list if at least one of a and b are empty. Complexity ---------- > O(n log n), where n = len(a) + len(b). """ n = len(a) m = len(b) if n == 0 or m == 0: return [] if min(n, m) <= 60: return convolution_naive(a, b) if a is b: return convolution_square(a) return convolution_ntt(a, b) def Extended_Euclid(n,m): stack=[] while m: stack.append((n,m)) n,m=m,n%m if n>=0: x,y=1,0 else: x,y=-1,0 for i in range(len(stack)-1,-1,-1): n,m=stack[i] x,y=y,x-(n//m)*y return x,y class MOD: def __init__(self,p,e=None): self.p=p self.e=e if self.e==None: self.mod=self.p else: self.mod=self.p**self.e def Pow(self,a,n): a%=self.mod if n>=0: return pow(a,n,self.mod) else: #assert math.gcd(a,self.mod)==1 x=Extended_Euclid(a,self.mod)[0] return pow(x,-n,self.mod) def Build_Fact(self,N): assert N>=0 self.factorial=[1] if self.e==None: for i in range(1,N+1): self.factorial.append(self.factorial[-1]*i%self.mod) else: self.cnt=[0]*(N+1) for i in range(1,N+1): self.cnt[i]=self.cnt[i-1] ii=i while ii%self.p==0: ii//=self.p self.cnt[i]+=1 self.factorial.append(self.factorial[-1]*ii%self.mod) self.factorial_inve=[None]*(N+1) self.factorial_inve[-1]=self.Pow(self.factorial[-1],-1) for i in range(N-1,-1,-1): ii=i+1 while ii%self.p==0: ii//=self.p self.factorial_inve[i]=(self.factorial_inve[i+1]*ii)%self.mod def Build_Inverse(self,N): self.inverse=[None]*(N+1) assert self.p>N self.inverse[1]=1 for n in range(2,N+1): if n%self.p==0: continue a,b=divmod(self.mod,n) self.inverse[n]=(-a*self.inverse[b])%self.mod def Inverse(self,n): return self.inverse[n] def Fact(self,N): if N<0: return 0 retu=self.factorial[N] if self.e!=None and self.cnt[N]: retu*=pow(self.p,self.cnt[N],self.mod)%self.mod retu%=self.mod return retu def Fact_Inve(self,N): if self.e!=None and self.cnt[N]: return None return self.factorial_inve[N] def Comb(self,N,K,divisible_count=False): if K<0 or K>N: return 0 retu=self.factorial[N]*self.factorial_inve[K]%self.mod*self.factorial_inve[N-K]%self.mod if self.e!=None: cnt=self.cnt[N]-self.cnt[N-K]-self.cnt[K] if divisible_count: return retu,cnt else: retu*=pow(self.p,cnt,self.mod) retu%=self.mod return retu N,K=map(int,readline().split()) A=list(map(int,readline().split())) P=list(map(int,readline().split())) child=[[] for x in range(N+1)] for i in range(1,N+1): child[P[i-1]].append(i) dp=[deque([]) for x in range(N+1)] MD=MOD(mod) MD.Build_Fact(N+1) fact=1 poly_K=[] for i in range(N+1): poly_K.append(fact*MD.Fact_Inve(i)%mod) fact*=K-i fact%=mod for x in range(N,-1,-1): if child[x]: ma=max(len(dp[y]) for y in child[x]) for y in child[x]: if len(dp[y])==ma: yy=y break dp[x]=dp[yy] for y in child[x]: if y==yy: continue for i in range(len(dp[y])): dp[x][i]+=dp[y][i] dp[x][i]%=mod dp[x].appendleft(A[x]) ans_lst=[None]*(N+1) DP=[None]*(N+1) le=len(dp[0]) DP[0]=deque(convolution(list(dp[0]),list(poly_K[:le][::-1]))[le-1:]) for x in range(N+1): ans_lst[x]=DP[x].popleft() for y in child[x]: if dp[x] is dp[y]: DP[y]=DP[x] else: le=len(dp[y]) DP[y]=deque(convolution(list(dp[y]),list(poly_K[:le][::-1]))[le-1:]) for i in range(le): DP[x][i]-=DP[y][i] DP[x][i]%=mod print(*ans_lst,sep="\n")