Re: polling's future [was: Re: Dynamic Ticks/HZ]

From: Andre Oppermann <andre_at_freebsd.org>
Date: Tue, 06 Nov 2012 11:38:44 +0100
On 06.11.2012 11:27, Poul-Henning Kamp wrote:
> --------
> In message <5098E526.6070101_at_freebsd.org>, Andre Oppermann writes:
>
>> Hi Luigi,
>>
>> do you agree on polling having outlived its usefulness in the light
>> of interrupt moderating NIC's and SMP complications/disadvantages?
>
> Can I just point out, that what we have is not in fact "device-polling"
> but only "some-kinds-of-network-interface-polling" ?
>
> I think it should go away, and if there still is a relevant
> usage segment, be replaced by _real_ "device-polling" which is
> not tied to the network stack.

Don't we already have the equivalent with a fast interrupt thread
that simply acknowledges and disables the interrupt and schedules
a dedicated taskqueue thread which then continues to process X as
long it comes in?  Only after all items have been processed the
interrupt is enabled again for the next batch.  The effect is hybrid
polling.  When there is no work the poller doesn't have to run.  If
work starts coming in the dedicated taskqueue picks up and continues
until finished.

> Amongst the sensible uses for that would be high-speed serial
> interfaces etc.

-- 
Andre
Received on Tue Nov 06 2012 - 09:37:16 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:32 UTC