From: Shashank Date <ADATE@...>
Date: 2002-06-12T15:01:29+09:00
Subject: Re: threads question (newbie)

Ok, then, replace the block

10.times { |i|
    commands << "ping 1.1.1.1 -n 10 -w 1000 > junk#{i}.txt"
}

with the block:

3.times { |i|
     commands << "dir /s > nul"
     commands << "ping 1.1.1.1 -n 10 -w 1000 > nul"
     commands << "find \"nokada\" %windir%\\*.* > nul"
}

and you will get the same results !!
But to get all the 9 threads to start simultaneously, I had to bump
sleep(0.01) to sleep(1).

I guess, I did not understand the dependency  on "winsock limitation".

I am sure, we will be able to repeat the experiment without the ping (or any
commands dependant on winsock);
Thanks,
-- Shanko

<nobu.nokada@softhome.net> wrote in message
news:200206120424.g5C4O9214711@sharui.nakada.kanuma.tochigi.jp...
> Hi,
>
> At Wed, 12 Jun 2002 11:21:17 +0900,
> Shashank Date wrote:
> > This is using Ruby 1.6.7 on Windows (mswin32).
> >
> > The following program without the sleep(0.01) stmt
> > seems to hang/stall after creating 3 threads until one thread is done.
> > Thus it takes over 35 seconds to finish.
>
> IO multiplexing works with only sockets in mswin32 -- this
> comes from winsock limitation.  And new thread would get the
> control immediatly, so `#{myCmd}` blocks before the main thread
> would run again.
>
> > But with the sleep stmt it starts all the 10 threads concurrently and
the
> > entire
> > process is over in about 10 to 12 seconds as expected.
>
> In that case, the main thread gets the control before the
> backticks.  And invoked commands can run simultaneously.
>
> With Thread.stop/wakeup it may also work.
>
> > threads = []
> > for command in commands
> >    threads << Thread.new(command){ |myCmd|
>             Thread.stop
> >           `#{myCmd}`
> >    }
> > end
>   Thread.exclusive {threads.each {|aThread| aThread.wakeup}}
> > threads.each { |aThread| aThread.join }
> > puts `echo %TIME%`
>
> --
> Nobu Nakada
>