关于dinic的当前弧优化

为什么这样

ll dinic(int x,ll In){
    if(x == t)return In;
    ll Bk = 0,res;
    for(int i = st[x];i && In;i = Next[i]){
        int y = ver[i];
        if(edge[i] && d[y] == d[x] + 1){
            res = dinic(y,min(In,edge[i]));
            In -= res,Bk += res;
            edge[i] -= res,edge[i ^ 1] += res;
        }
        st[x] = Next[i];//here
    }
    if(Bk == 0)
        d[x] = 0;
    return Bk;
}

如此之慢 提交记录

而这样

ll dinic(int x,ll In){
    if(x == t)return In;
    ll Bk = 0,res;
    for(int i = st[x];i && In;i = Next[i]){
        int y = ver[i];
        if(edge[i] && d[y] == d[x] + 1){
            res = dinic(y,min(In,edge[i]));
            In -= res,Bk += res;
            edge[i] -= res,edge[i ^ 1] += res;
        }
        if(!edge[i])st[x] = Next[i];//here
    }
    if(Bk == 0)
        d[x] = 0;
    return Bk;
}

却快了不少提交记录