Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.lang.python > #77192

Re: Thread terminate

From Marko Rauhamaa <marko@pacujo.net>
Newsgroups comp.lang.python
Subject Re: Thread terminate
Date 2014-08-28 12:02 +0300
Organization A noiseless patient Spider
Message-ID <87sikhf0m2.fsf@elektro.pacujo.net> (permalink)
References <20140827185514.GA18660@arxnet.hu> <CALvWhxsJW+pDKYnf9iZzRHsUg6V9wbNjp=3VqEjGSomez_WQEw@mail.gmail.com> <mailman.13546.1409215846.18130.python-list@python.org>

Show all headers | View raw


Ervin Hegedüs <airween@gmail.com>:

> at this time there is only one thread, as you wrote. I just try
> to prepare it to higher load, when one thread will not enough...

Threads are a necessary evil when dealing with blocking function calls,
but evil they remain. I would generally *not* associate a thread for
each parallel context but work with a limited thread pool, whose size is
independent of the number of contexts. Often, a good rule of thumb is to
have two threads per CPU core to get most out of the hardware.

Since you talked about forking, you might be better served with a pool
of processes instead of threads since processes have less surprising
semantics and, if push came to shove, you can kill them with a signal.


Marko

Back to comp.lang.python | Previous | NextPrevious in thread | Next in thread | Find similar | Unroll thread


Thread

Re: Thread terminate Ervin Hegedüs <airween@gmail.com> - 2014-08-28 10:52 +0200
  Re: Thread terminate Marko Rauhamaa <marko@pacujo.net> - 2014-08-28 12:02 +0300
    Re: Thread terminate Ervin Hegedüs <airween@gmail.com> - 2014-08-28 11:19 +0200
    Re: Thread terminate Cameron Simpson <cs@zip.com.au> - 2014-08-29 08:30 +1000

csiph-web