分布式计算-相位算法

一、一般情况

1、节点P的近邻有入近邻（in-neighbor,进程只能向p发送消息）和出近邻(out-neighbor，p只能向进程发送消息)。

p的入近邻存储在集Inp中，p的出近邻邦存储在集Outp中。

2、在相位算法中，每个进程向每个出近邻只发送D条消息。仅当收到来自每个入近邻的i条消息后，才将第(i+1)条消息发送给每个出近邻。

3、通过每条信道至多发送D条消息，使消息复杂度为|E|*D。

4、算法：

cons D:integer =the network diameter;

var  Recp[q] :0...D init 0.for each q∈Inp;

       Sentp    :0...D init 0;

begin if p is initiator then

             begin forall r∈Outp do sen <tok> to r;

                       Sentp:=Sentp+1

            end;

             while minqRecp[q]<D do

                      begin receive <tok> (from neighbor q0);

                                Recp[q0]:=Recp[q0]+1;

                                if   minqRecp[q]:=Recp[q0]+1;

                                     begin forall rOutp do send <tok> to r;

                                               Sentp:=Sentp+1

                                     end

                      end;

              decide

end

二、团网中的相位算法

      如果网络是直径为1的团。在这种情况下，只接收每个近邻的一条消息，每个进程只需计算它所接收消息的总数就足够了。

      var Recp:0..N-1 init 0;

            Sentp:0...1 init 0;

     begin if p isinitiator then

               begin forall rNeighp do send <tok> to r;

                         Sentp:=Sentp+1

               end;

               while Recp<#Neighp do

                        begin receive <tok>;

                                  Recp:=Recp+1;

                                  if Sentp=0 then

                                      begin forall r∈Neighp do send <tok> to r;

                                                  Sentp:=Sentp+1

                                       end

                      end;

               decide

end