对Erlang使用SMP模式效率更低的猜想

erl -noshell +P 1000000 -s processes max 800000 -s init stop
Maximum allowed processes:1000000
Process spawn time=4.16250 (8.02000) microseconds

erl -smp -noshell +P 1000000 -s processes max 800000 -s init stop
Maximum allowed processes:1000000
Process spawn time=19.7000 (28.0012) microseconds

max(N) ->
  Max = erlang:system_info(process_limit),
  io:format("Maximum allowed processes:~p~n" ,[Max]),
  statistics(runtime),
  statistics(wall_clock),
  L = for(1, N, fun() -> spawn(fun() -> wait() end) end),
 {_, Time1} = statistics(runtime),
 {_, Time2} = statistics(wall_clock),
 lists:foreach(fun(Pid) -> Pid ! die end, L),
 U1 = Time1 * 1000 / N,
 U2 = Time2 * 1000 / N,
 io:format("Process spawn time=~p (~p) microseconds" ,
 [U1, U2]).

wait() ->
 receive
  die -> void
 end.
 for(N, N, F) -> [F()];
 for(I, N, F) -> [F()|for(I+1, N, F)].

-module(message).
-export([main/1, loop/1]).

main(Times) when is_integer(Times) ->
    Last = start(500, self()),
    io:fwrite("~p~n", [sendtimes(Times, Last, 0)]);
main([Arg]) ->
    N = list_to_integer(Arg),
    Last = start(500, self()),
    io:fwrite("~p~n", [sendtimes(N, Last, 0)]),
    erlang:halt(0).

start(0, LastPID) -> LastPID;

start(X, LastPID) -> start(X-1, spawn(message, loop, [LastPID])).

loop(LastPID) ->
receive
N ->
LastPID ! N+1,
loop(LastPID)
end.

sendtimes(0,_ ,X) -> X;

sendtimes(N, Last, X) ->
Last ! 0,
receive Y -> Y end,
sendtimes(N-1, Last, X+Y).

对待multi-processor的情况Erlang通常的策略是，为每一个CPU分配一个工作线程.在不同的native thread中的processor进行Message Passing 就不是一个原子操作了.在这个问题上Erlang通过一个非常巧妙的方法来保证Message Passing的同步.
当一个Sender Process 要准备向Reciver Process发送一个Message的时候，Scheduler首先会检查这两个Process是不是在一个native thread中，如果是则按常规方法进行Message Passing,如果不是Scheduler会把reciver从一个native thread调度到Sender 所在的native thread里，当recive结束后根据调度策略决定是否再将reciver调度回原来的native thread中. 这样就保证了多CPU上的Message passing仍然是原子的.