perl - 如何在 fork 进程中获取随机数？

Question

我 fork 了几个进程，我需要每个进程对随机选择的项目执行任务。 我发现 fork 进程选择完全相同的随机数。 我尝试生成种子并调用 srand()，但没有多大帮助。实际上，我读过的大多数文档都建议避免使用 srand() “除非您确切知道自己在做什么”。

这是我的代码:

#!/usr/bin/perl
use strict;
use warnings;
use Getopt::Long qw(GetOptions);
use Time::HiRes qw(time);
use MongoDB;
#use Math::Random::Secure qw(rand);

my $num_clients = shift;
my $num_loops = 50_000_000;

sub test_forked_sub {
    my $sum=0;
    my $random_offset = rand(120);
    printf "time: %10.4f -  Random offset: %6.2f at pid: %s  \n", time(), $random_offset, $$;
    for (my $i=0; $i < $num_loops  ; $i++) {
        $sum += rand(120);
    }
}

my $nub_processes = 0;
for (my $client_id =0; $client_id < $num_clients; $client_id++ ) {
    if (my $pid = fork) {     #Parent
        $nub_processes++;
    }
    else {         # child
        die "cannot fork: $!" unless defined $pid;
        test_forked_sub();
        exit(0);
    }
}

while ($nub_processes){
    wait;
    $nub_processes--;
}

当我运行它时，我在每个进程中得到完全相同的“随机”数字:

$ time ./test_rand_fork.pl 10
time: 1569510011.6891 -  Random offset:  46.64 at pid: 2091
time: 1569510011.6937 -  Random offset:  46.64 at pid: 2092
time: 1569510011.6987 -  Random offset:  46.64 at pid: 2093
time: 1569510011.7028 -  Random offset:  46.64 at pid: 2094
time: 1569510011.7070 -  Random offset:  46.64 at pid: 2095
time: 1569510011.7097 -  Random offset:  46.64 at pid: 2096
time: 1569510011.7144 -  Random offset:  46.64 at pid: 2097
time: 1569510011.7203 -  Random offset:  46.64 at pid: 2098
time: 1569510011.7230 -  Random offset:  46.64 at pid: 2099
time: 1569510011.7249 -  Random offset:  46.64 at pid: 2100

real    0m3.974s
user    0m32.955s
sys     0m1.444s

一个可能但丑陋的解决方案，它不是 fork 。相反，从我的 shell 运行多个实例，如下所示:

# shell processes
$ for i in `seq 1 10`; do  ./test_rand_fork.pl 1 &  done
time: 1569511908.7708 -  Random offset:   7.44 at pid: 4129
time: 1569511908.8070 -  Random offset:  19.50 at pid: 4131
time: 1569511908.8068 -  Random offset:  97.59 at pid: 4132
time: 1569511908.8073 -  Random offset:  14.51 at pid: 4133
time: 1569511908.8077 -  Random offset:  16.70 at pid: 4134
time: 1569511908.8080 -  Random offset: 108.63 at pid: 4138
time: 1569511908.8079 -  Random offset:  69.44 at pid: 4137
time: 1569511908.8080 -  Random offset:  83.25 at pid: 4136
time: 1569511908.8080 -  Random offset:  43.33 at pid: 4135
time: 1569511908.8203 -  Random offset:  33.82 at pid: 4139

由于“shell 进程”是一个丑陋的解决方案，我尝试使用 Math::Random::Secure。 以下是我根据需要获得随机性的结果，但它比“shell 进程”方法慢 50 倍:

# using: Math::Random::Secure qw(rand);
$ time ./test_rand_fork.pl 10
time: 1569510036.9331 -  Random offset: 112.48 at pid: 2128
time: 1569510036.9470 -  Random offset:  47.15 at pid: 2129
time: 1569510036.9501 -  Random offset:  20.77 at pid: 2130
time: 1569510036.9517 -  Random offset:  40.98 at pid: 2131
time: 1569510036.9521 -  Random offset:  13.84 at pid: 2132
time: 1569510036.9538 -  Random offset:  20.43 at pid: 2133
time: 1569510036.9543 -  Random offset:  48.48 at pid: 2134
time: 1569510036.9563 -  Random offset: 109.29 at pid: 2135
time: 1569510036.9579 -  Random offset:  70.30 at pid: 2136
time: 1569510036.9601 -  Random offset:  24.31 at pid: 2137

real    3m17.251s
user    32m31.129s
sys     0m1.054s

我需要的随机性并不是出于安全目的，我只是需要良好的传播。 有没有一种方法可以为 fork 进程生成足够好的种子，并且仍然使用标准 rand() 或替代更快的方法？或者至少不需要安装额外的库？

Answer 1

确实是srand你要的那个。虽然确实一般情况下人们不需要花太多时间来处理它，但它有一些用途

However, there are a few situations where programs are likely to want to call srand. One is for generating predictable results, generally for testing or debugging. There, you use srand($seed), with the same $seed each time.

而且，正是为了您的目的

Another case is that you may want to call srand after a fork to avoid child processes sharing the same seed value as the parent (and consequently each other).

一个简单的例子

perl -wE'
    say "Parent $$ rand: ", rand;  
    for (1..4) { 
        $pid = fork // die "Cant fork: $!"; 
        if ($pid == 0) { srand(); say "\tchild $$ rand: ", rand; exit } 
    }'

打印

Parent 23375 rand: 0.502774688721733
        child 23376 rand: 0.914527430039929
        child 23377 rand: 0.97985713889739
        child 23378 rand: 0.127702740327553
        child 23379 rand: 0.187181734786467

正是上面的srand发挥了作用；如果没有它，数字是相同的(因为它们必须是相同的)。请注意，它也可以使用种子运行，这为我们提供了更好的诊断/测试控制。

---

注意，避免传播坏习惯。

在上面的快速程序中，子进程会被系统清理(由init回收)。然而，一个人确实应该总是收获，所以上面应该确实有

for (1..4) { 
    $pid = fork // die "Cant fork: $!"; 
    if ($pid == 0) { srand(); say "\tchild $$ rand: ", rand; exit } 
    push @procs, $pid;
}
for (@procs) { $gone = wait; say "$gone exited with $?" }

使用 waitpid 会更好和非阻塞以实现更好的监控，也许在 signal handler 中.