Subject: Re: [linux-audio-dev] interesting 2.3.4* kernel statistic
From: Benno Senoner (sbenno_AT_gardena.net)
Date: ti helmi 22 2000 - 17:52:54 EST
On Tue, 22 Feb 2000, Juhana Sadeharju wrote:
> >From: Roger Larsson <roger.larsson_AT_norran.net>
> >
> >I.e with the version tested your worst case will be at least 34 ms.
> >Since several locks might be taken during one kernel operation.
>
> This means MS Windows is better platform for audio with 20 ms latency?!
> And Linux has 34 ms minimum latency?!
>
> If that is not true, please tell slowly what this all _really_ means.
> I wonder what was the point of low latency pathes if the situation
> is really that bad as you say.
>
> I really wonder does Linus kernel authors really know what they are doing.
>
> Juhana
--Don't worry, the folks on the linux-kernel list are just debugging the 2.2.14 which has still some some races in it.
I can say nothing about 2.3.x lowlatency performance, since Ingo has yet to release a patch. But anyway if 2.2.10 is able to deliver rock-solid 3-5ms latencies, I don't see why 2.3.x shouldn't ? As far I know 2.3.x has more finegrained locking than 2.2.x.
"Linus kernel authors" or "Linux kernel authors" ? :-)
Anyway Linus is aware of the low-latency patches, and Ingo told me that we have not to worry about the inclusion of the low-latency patches into 2.4.
Meanwhile the best kernel to test your lowlatency userspace stuff is 2.2.10N6B. (the only drawback is that it does not deliver lowlatency while in SMP mode due to some locking, but in UP miode it works wonderfully)
Ingo told me that it will not be a big problem to get this performance on SMP 2.3.x kernels. (In fact his initial latency tests were run on a quad Xeon !
Benno.
This archive was generated by hypermail 2b28 : pe maalis 10 2000 - 07:23:27 EST