Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 9 Mar 1997 23:54:11 +0100
From:      j@uriah.heep.sax.de (J Wunsch)
To:        hackers@FreeBSD.ORG
Cc:        bmcgover@cisco.com (Brian McGovern)
Subject:   Re: Performance questions (Interrupt vs. polled)
Message-ID:  <19970309235411.LH28573@uriah.heep.sax.de>
In-Reply-To: <3.0.32.19970307084514.00905c60@visigenic.com>; from Tim Oneil on Mar 7, 1997 08:45:15 -0800
References:  <3.0.32.19970307084514.00905c60@visigenic.com>

next in thread | previous in thread | raw e-mail | index | archive | help
As Tim Oneil wrote:

> No flame. My learned opinion however is that a system with efficiant thread
> scheduling (such as freeBSD) is best served by polling. Interrupted tasking
> is best for serial, brain-dead systems, like DOS.

You've just got this reversed.

A decent multiprocessing system benefits a lot from hardware that can
do interrupt notification right when there is something to do.  The
worst that could happen are CPU spin loops where the CPU is waiting
for some bit to be set by a peripheral device.  (That's why IDE drives
are ugly.)

To the contrary, DOS often doesn't use interrupt-controlled drivers,
so you can even work with peripherals that have a broken-by-design
interrupt behaviour (like the IRQ clash for COM1/3 or COM2/4).  This
doesn't do much harm there, since it's a single-process system, so the
CPU has to wait anyway until the current operation has completed.

-- 
cheers, J"org

joerg_wunsch@uriah.heep.sax.de -- http://www.sax.de/~joerg/ -- NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19970309235411.LH28573>