Kernel 2.6.28

[Free Ekanayaka-3]

Hi,

this is an experimental package for kernel 2.6.28, which is supposed
to have a good realtime performance (sorry i386 only for now):

http://people.64studio.com/~free/linux-image-2.6.28-1-686_2.6.28-1~experimental.1_i386.deb

It would be great if 64 Studio testers could check it out and report
about it. It should work both on 64 Studio 2.1 and 3.0 alpha.

To install it just download the package and run:

sudo dpkg -i linux-image-2.6.28-1-686_2.6.28-1~experimental.1_i386.deb

Thanks!

Free
_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[Gabriel M. Beddingfield]

On Tue, January 13, 2009 4:16 pm, Free Ekanayaka wrote:
> Hi,
>
> this is an experimental package for kernel 2.6.28, which is supposed
> to have a good realtime performance (sorry i386 only for now):

Cool!  Thanks!  Is ALSA compiled in-kernel, or do we need to compile it
separate like for 2.6.21?

Thanks,
Gabriel

--
               G a b r i e l   M   B e d d i n g f i e l d

_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[Free Ekanayaka-3]

|--==> On Tue, 13 Jan 2009 16:31:45 -0600 (CST), "Gabriel M. Beddingfield" <gabriel@...> said:

  GMB> On Tue, January 13, 2009 4:16 pm, Free Ekanayaka wrote:
  >>Hi,
  >>
  >>this is an experimental package for kernel 2.6.28, which is supposed
  >>to have a good realtime performance (sorry i386 only for now):

  GMB> Cool!  Thanks!  Is ALSA compiled in-kernel, or do we need to compile it
  GMB> separate like for 2.6.21?

Compiled it.

Ciao!

Free
_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[Jaromír Mikeš]

<    GMB> On Tue, January 13, 2009 4:16 pm, Free Ekanayaka wrote:
<    >>Hi,
<    >>
<    >>this is an experimental package for kernel 2.6.28, which is supposed
<    >>to have a good realtime performance (sorry i386 only for now):
<  
<    GMB> Cool!  Thanks!  Is ALSA compiled in-kernel, or do we need to compile it
<    GMB> separate like for 2.6.21?
<  
<  Compiled it.
<  

Many thanks .... sorry .. how to compile alsa for it?

Cheers

mira
_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[Free Ekanayaka-3]

|--==> On Wed, 14 Jan 2009 00:41:47 +0100 (CET), Jaromír Mikeš <mira.mikes@...> said:

  JM> <    GMB> On Tue, January 13, 2009 4:16 pm, Free Ekanayaka wrote:
  JM> <    >>Hi,
  JM> <    >>
  JM> <    >>this is an experimental package for kernel 2.6.28, which is supposed
  JM> <    >>to have a good realtime performance (sorry i386 only for now):
  JM> <  
  JM> <    GMB> Cool!  Thanks!  Is ALSA compiled in-kernel, or do we need to compile it
  JM> <    GMB> separate like for 2.6.21?
  JM> <  
  JM> <  Compiled it.
  JM> <  

  JM> Many thanks .... sorry .. how to compile alsa for it?

Sorry, I actually meant "Compiled in", no need to compile it
separately :)

Ciao!

Free
_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[Jaromír Mikeš]

<    JM> <    GMB> On Tue, January 13, 2009 4:16 pm, Free Ekanayaka wrote:
<    JM> <    >>Hi,
<    JM> <    >>
<    JM> <    >>this is an experimental package for kernel 2.6.28, which is
<  supposed
<    JM> <    >>to have a good realtime performance (sorry i386 only for now):
<    JM> <  
<    JM> <    GMB> Cool!  Thanks!  Is ALSA compiled in-kernel, or do we need to
<  compile it
<    JM> <    GMB> separate like for 2.6.21?
<    JM> <  
<    JM> <  Compiled it.
<    JM> <  
<  
<    JM> Many thanks .... sorry .. how to compile alsa for it?
<  
<  Sorry, I actually meant "Compiled in", no need to compile it
<  separately :)
<
 
Thank you ;-)
_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[dave-322]

Free Ekanayaka wrote:

> Hi,
>
> this is an experimental package for kernel 2.6.28, which is supposed
> to have a good realtime performance (sorry i386 only for now):
>
> http://people.64studio.com/~free/linux-image-2.6.28-1-686_2.6.28-1~experimental.1_i386.deb
>
> It would be great if 64 Studio testers could check it out and report
> about it. It should work both on 64 Studio 2.1 and 3.0 alpha.
>
> To install it just download the package and run:
>
> sudo dpkg -i linux-image-2.6.28-1-686_2.6.28-1~experimental.1_i386.deb
>
> Thanks!
>
> Free
>  

Hi,

Tested on two 3.0 alpha hardy installs (different computers), both ok.

Still need to test on 2.1 lenny dist-upgrade.

Cheers!
dave.
_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[Quentin Harley]

> Free Ekanayaka wrote:
>> Hi,
>>
>> this is an experimental package for kernel 2.6.28, which is supposed
>> to have a good realtime performance (sorry i386 only for now):
>>
>> http://people.64studio.com/~free/linux-image-2.6.28-1-686_2.6.28-1~experimental.1_i386.deb
>>
>> It would be great if 64 Studio testers could check it out and report
>> about it. It should work both on 64 Studio 2.1 and 3.0 alpha.
>>
>> To install it just download the package and run:
>>
>> sudo dpkg -i linux-image-2.6.28-1-686_2.6.28-1~experimental.1_i386.deb
>>
>> Thanks!
>>
>> Free
>>
>
> Hi,
>
> Tested on two 3.0 alpha hardy installs (different computers), both ok.
>
> Still need to test on 2.1 lenny dist-upgrade.

I finally had chance to test performance on my older 32bit laptop Centrino
Duo.

2.1 with 2.6.21-multimedia works well, with Ardour and Jamin running at 32
% and no x-runs

the same setup with the new kernel starts dropping x-runs just after
starting jamin.

3.0 Alpha as is also produces x-runs, and with the new kernel even more...
I'll have to stay on 2.1 for the moment on my laptop. My 64bit hardware
runs nicely on the 2.1 upgraded to 3.0 alpha at the moment.

2.6.28 might be ok, but is is still not as powerful as our 2.6.21 kernel
in terms of realtime performance.

Cheers, Q


_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[Gabriel M. Beddingfield]

On Wed, January 14, 2009 1:47 pm, qharley@... wrote:
> the same setup with the new kernel starts dropping x-runs just after
> starting jamin.
[snip]
>
> 2.6.28 might be ok, but is is still not as powerful as our 2.6.21 kernel
> in terms of realtime performance.

Are you sure you're getting RT performance?  When I played with a 2.6.28
kernel a few months ago, I had trouble getting the scheduler configured
correctly.  (Between 2.6.21 and now, the Completely Fair Scheduler was
implemented.)  Now, being in the 'audio' group wasn't sufficient to have
RT permission.

IIRC, the kernel two scheduling paradigms... and neither one works well
with the old limits.conf system.  However, I don't remember what I *did*
do to get it working (that PC is at home).

--
               G a b r i e l   M   B e d d i n g f i e l d

_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[Quentin Harley-2]

>
> On Wed, January 14, 2009 1:47 pm, qharley@... wrote:
>> the same setup with the new kernel starts dropping x-runs just after
>> starting jamin.
> [snip]
>>
>> 2.6.28 might be ok, but is is still not as powerful as our 2.6.21 kernel
>> in terms of realtime performance.
>
> Are you sure you're getting RT performance?

Not at all sure... RT is flashing in qjackctl, so I assumed it was running
RT.

> When I played with a 2.6.28
> kernel a few months ago, I had trouble getting the scheduler configured
> correctly.  (Between 2.6.21 and now, the Completely Fair Scheduler was
> implemented.)  Now, being in the 'audio' group wasn't sufficient to have
> RT permission.
>
> IIRC, the kernel two scheduling paradigms... and neither one works well
> with the old limits.conf system.  However, I don't remember what I *did*
> do to get it working (that PC is at home).

When you have time, please let me know what you did so I can try it on
this lower powered laptop of mine.  It serves as a good benchmark to new
setups and kernels as it runs well under 2.1 with 2.6.21, but anything
worse fails.

My real system does not have much troubles, but is also not with me as I
am on a business trip on a different continent at the moment

Thanks,
Quentin
>
> --
>                G a b r i e l   M   B e d d i n g f i e l d
>
> _______________________________________________
> 64studio-devel mailing list
> 64studio-devel@...
> http://lists.64studio.com/mailman/listinfo/64studio-devel
>


_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel

[Gabriel M. Beddingfield]

On Thu, January 15, 2009 3:12 am, quentin@... wrote:
>
> Not at all sure... RT is flashing in qjackctl, so I assumed it was running
> RT.

Then you probably were.  IIRC, jackd wouldn't start for me with the -R
option.

> When you have time, please let me know what you did so I can try it on
> this lower powered laptop of mine.  It serves as a good benchmark to new
> setups and kernels as it runs well under 2.1 with 2.6.21, but anything
> worse fails.

OK, I dug up all my old memories.  The kernel I used was 2.6.26.5-rt9.  I
flipped a coin and chose CONFIG_RT_GROUP_SCHED and CONFIG_USER_SCHED
instead of CONFIG_CGROUP_SCHED.  In order to have RT permission, I would
have to do something like:

$ sudo echo -n 75 > /sys/kernel/uids/$UID/cpu_rt_runtime_us

Then I could set stuff with realtime priority and it worked OK.  Machine
is 500MHz PIII with about 384MB of memory.

However, it looks like Free has used the _OTHER_ scheduling paradigm.
When I started JACK, it appeared to take the -R flag just fine.  I haven't
had a chance to try any audio projects, yet.

In the end, I abandoned the 2.6.26 kernel because I was having other
issues (which I can't remember any more).  I want to say that I was
getting periodic freeze-ups (or something) on two different machines...
but I can't find the key to that memory.  That problem was probably
between the chair and the monitor, though.

--
               G a b r i e l   M   B e d d i n g f i e l d

_______________________________________________
64studio-devel mailing list
64studio-devel@...
http://lists.64studio.com/mailman/listinfo/64studio-devel