This is the mail archive of the
glibc-linux@ricardo.ecn.wfu.edu
mailing list for the glibc project.
Re: Slow pthread_create() under high load
- To: Alan Cox <alan at lxorguk dot ukuu dot org dot uk>
- Subject: Re: Slow pthread_create() under high load
- From: Ulrich Drepper <drepper at redhat dot com>
- Date: 26 Mar 2000 18:22:42 -0800
- Cc: kaz at ashi dot footprints dot net (Kaz Kylheku), dpw at doc dot ic dot ac dot uk (David Wragg), glibc-linux at ricardo dot ecn dot wfu dot edu, linux-kernel at vger dot rutgers dot edu
- References: <E12ZOlG-0006AC-00@the-village.bc.nu>
- Reply-To: glibc-linux at ricardo dot ecn dot wfu dot edu
Alan Cox <alan@lxorguk.ukuu.org.uk> writes:
> You mean create a thread which starts by doing a pause syscall in a loop
> until you poke it ? Why does that need kernel help
Because you would need extra synchronization. How else can the
creater be sure the SIGCONT arrives when the new thread already sent
the SIGSTOP? This would make the startup process yet slower and we
were just talking about the already high overhead. This is a two-line
change, nothing which bloats up the kernel or makes it slow.
--
---------------. drepper at gnu.org ,-. 1325 Chesapeake Terrace
Ulrich Drepper \ ,-------------------' \ Sunnyvale, CA 94089 USA
Red Hat `--' drepper at redhat.com `------------------------