mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* What am I losing with noapic
@ 2002-04-02 10:53 Tim Kay
  2002-04-02 11:24 ` Tigran Aivazian
  0 siblings, 1 reply; 2+ messages in thread
From: Tim Kay @ 2002-04-02 10:53 UTC (permalink / raw)
  To: linux-kernel

Hi all,
	Does anyone know what I'm actually losing with having to set noapic on 
bootup? I mean in real terms how much harder / slower is an SMP machine 
working when it's doing standard multi-bus xt polling compared to when it's 
in APIC poll state. I appreciate that there can be a reduction in interrupt 
response latency using the damn thing but is this a measurable amount given a 
machine processing about 1200 interrupts/sec? (this figure is a sum rather 
than per processor). As an added complication how do I get around interrupt 
routing conflicts in noapic mode and do the 'routing conflict for xx:xx:xx 
have X want Y' messages make any difference to this performance?

	A useful URL (I couldn't find any) or reference would suffice if this is too 
invloved or boring a topic to explain easily.

TIA

Tim

-- 
----------------
Tim Kay
systems administrator
Advfn.com Plc - http://www.advfn.com/

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: What am I losing with noapic
  2002-04-02 10:53 What am I losing with noapic Tim Kay
@ 2002-04-02 11:24 ` Tigran Aivazian
  0 siblings, 0 replies; 2+ messages in thread
From: Tigran Aivazian @ 2002-04-02 11:24 UTC (permalink / raw)
  To: Tim Kay; +Cc: linux-kernel

When you pass "noapic" the IO-APIC IRQ routing entries are not used. This
is useful if the way MP table is prepared by firmware is not good.

What is most noticeable in this case is that all interrupts will be
handled by the first cpu.

On Tue, 2 Apr 2002, Tim Kay wrote:

> Hi all,
> 	Does anyone know what I'm actually losing with having to set noapic on
> bootup? I mean in real terms how much harder / slower is an SMP machine
> working when it's doing standard multi-bus xt polling compared to when it's
> in APIC poll state. I appreciate that there can be a reduction in interrupt
> response latency using the damn thing but is this a measurable amount given a
> machine processing about 1200 interrupts/sec? (this figure is a sum rather
> than per processor). As an added complication how do I get around interrupt
> routing conflicts in noapic mode and do the 'routing conflict for xx:xx:xx
> have X want Y' messages make any difference to this performance?
>
> 	A useful URL (I couldn't find any) or reference would suffice if this is too
> invloved or boring a topic to explain easily.
>
> TIA
>
> Tim
>
>


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2002-04-02 11:25 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-04-02 10:53 What am I losing with noapic Tim Kay
2002-04-02 11:24 ` Tigran Aivazian

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®