实现功能：同最大流 1
这里面主要是把前面的邻接矩阵改成了邻接表，相比之下速度大大提高——本人实测，当M=1000000 N=10000 时，暂且不考虑邻接矩阵会不会MLE，新的程序速度快了很多倍（我们家这个很弱的电脑上耗时0.3s）；而当M=300000 N=10000时，优势更加明显（几乎是秒出），别的没了，尤其当遇到稀疏图的时候这样子是大大划算的！！！
1 type 2 point=^node; 3 node=record 4 g,w:longint; 5 next:point; 6 end; 7 8 var 9 i,j,k,l,m,n,tmp,ans,aug,mi,s,t:longint; 10 di,a:array[0..10005] of point; 11 pre,his,dis,vh:array[0..10005] of longint; 12 flag:boolean;p,jl:point; 13 function min(x,y:longint):longint;inline; 14 begin 15 if x<y then min:=x else min:=y; 16 end; 17 function add(x,y,z:longint):longint;inline; 18 var p:point; 19 begin 20 new(p);p^.w:=z;p^.g:=y; 21 p^.next:=a[x];a[x]:=p; 22 end; 23 procedure op(x,y,z:longint);inline; 24 var p:point; 25 begin 26 p:=a[x]; 27 while p<>nil do 28 begin 29 if (p^.g=y) and ((p^.w+z)>=0) then 30 begin 31 p^.w:=p^.w+z; 32 break; 33 end; 34 p:=p^.next; 35 end; 36 end; 37 begin 38 readln(n,m,s,t); 39 for i:=1 to n do a[i]:=nil; 40 for i:=1 to m do 41 begin 42 readln(j,k,l); 43 add(j,k,l);add(k,j,0); 44 end; 45 for i:=1 to n do di[i]:=a[i]; 46 fillchar(dis,sizeof(dis),0); 47 fillchar(pre,sizeof(pre),0); 48 fillchar(his,sizeof(his),0); 49 fillchar(vh,sizeof(vh),0); 50 i:=s;vh[0]:=n;ans:=0;aug:=maxlongint; 51 while dis[s]<n do 52 begin 53 flag:=false;his[i]:=aug; 54 p:=a[i]; 55 while p<>nil do 56 begin 57 if (p^.w>0) and (dis[i]=(dis[p^.g]+1)) then 58 begin 59 aug:=min(aug,p^.w); 60 pre[p^.g]:=i;di[i]:=p; 61 flag:=true;i:=p^.g; 62 if i=t then 63 begin 64 ans:=ans+aug; 65 while i<>s do 66 begin 67 tmp:=i; 68 i:=pre[i]; 69 op(i,tmp,-aug); 70 op(tmp,i,aug); 71 end; 72 aug:=maxlongint; 73 end; 74 break; 75 end; 76 p:=p^.next; 77 end; 78 if flag then continue; 79 jl:=nil;mi:=n-1; 80 p:=a[i]; 81 while p<>nil do 82 begin 83 if (p^.w>0) and (dis[p^.g]<mi) then 84 begin 85 jl:=p;mi:=dis[p^.g]; 86 end; 87 p:=p^.next; 88 end; 89 di[i]:=jl; 90 dec(vh[dis[i]]); 91 if vh[dis[i]]=0 then break; 92 dis[i]:=mi+1; 93 inc(vh[dis[i]]); 94 if i<>s then 95 begin 96 i:=pre[i]; 97 aug:=his[i]; 98 end; 99 end; 100 writeln(ans); 101 end. 102