Lowlatency kernel testing (Re: A feature for -lowlatency kernel?)

[ailo]

On 04/05/2011 09:13 PM, David Henningsson wrote:
>> I'm testing the 2.6.38-7 -generic, and it is in fact working better than
>> before. Almost as well as -lowlatency.
> 
> Only almost?
> Anyway, that's kind of why I've losing faith in lowlatency personally -
> I have yet to see someone showing me that it actually performs better
> than the generic kernel.
> 

I have not been able to achieve as low latencies with -generic as with
-lowlatency, except for once, which was on 2.6.38-1. At the time,
-generic seemed to behave exactly the same as -lowlatency, however after
a system update, the same -generic kernel would not give me low
latencies without audio dropouts anymore.
Two of us got the same initial result with 2.6.38-1, so I'm pretty sure
it was not something I dreamt :P.

After that, for a good while, -generic was not usable with jackd for
latencies that are required for playing soft instruments, or monitoring
realtime audio processing.
Now, with 2.6.38-7, -generic seems to behave a good deal better, but I
can still not get the same low latency as I can with -lowlatency.

Also, during this whole period, -lowlatency has been more or less,
perfectly stable, and virtually as reliable as a realtime patched kernel.
The diff in my tests may be more related to me picking up problems
during later tests, that I didn't pick up during my first tests. Or
something in Ubuntu other than the kernel is affecting that in later tests.

So, my opinion is, I have still to see proof of -generic behaving as
reliably as -lowlatency.

-- 
ailo