Timing 250 versus 1000

Timing 250 versus 1000

[Alex Stone]

No doubt i'm likely to get hammered for asking this on a technical ML, 
but i'm brave enough and interested enough to risk it.

Is there an important reason why timing in the kernel is set to 250 by 
default?

I'm using linux to write music with. With the addition of many of the RT 
patches in the standard kernel, recording audio at low latencies with a 
standard kernel is no longer a problem, but for those of us who write a 
lot of midi driven work, we're still more or less required to use some 
sort of RT kernel to get any degree of playback timing accuracy.

I appreciate my use case is just one among many, but i've done a lot of 
research on the interlink, and unless i'm missing something really 
simple here, i can't find a reason why the default timer can't be set at 
1000, and be done with it.

I appreciate you chaps are busy, and elbow deep in code, just a "yes it 
could" or "no we won't" would suffice.

Thanks,

Alex.

Re: Timing 250 versus 1000

[Clemens Ladisch]

Alex Stone wrote:
> Is there an important reason why timing in the kernel is set to 250 by default?

It's a compromise.  Read kernel/Kconfig.hz.

> those of us who write a lot of midi driven work, we're still more or
> less required to use some sort of RT kernel to get any degree of
> playback timing accuracy.

The ALSA sequencer timer stopped using the system timer years ago.
And the difference between RT and 'normal' kernels is not the timer
frequency.


Regards,
Clemens

Re: Timing 250 versus 1000

[Clemens Ladisch]

Alex Stone wrote:
> On 04/08/2012 05:08 PM, Clemens Ladisch wrote:
>> Alex Stone wrote:
>>> those of us who write a lot of midi driven work, we're still more or
>>> less required to use some sort of RT kernel to get any degree of
>>> playback timing accuracy.
>>
>> The ALSA sequencer timer stopped using the system timer years ago.
>> And the difference between RT and 'normal' kernels is not the timer
>> frequency.
>
> I'm using jack with seq, so are you saying the ALSA sequencer timer is
> using a higher timer rate than than is set in the kernel,

It's using different timers (RTC or hrtimer), which support higher
precision.

> and there's no need for me to bother with this,

Yes.  If you want to be sure, check that G3 appears in /proc/asound/timers
when you're using the sequencer.


Regards,
Clemens

Re: Timing 250 versus 1000

[Alex Stone]

On 04/10/2012 02:04 PM, Clemens Ladisch wrote:
> Alex Stone wrote:
>> On 04/08/2012 05:08 PM, Clemens Ladisch wrote:
>>> Alex Stone wrote:
>>>> those of us who write a lot of midi driven work, we're still more or
>>>> less required to use some sort of RT kernel to get any degree of
>>>> playback timing accuracy.
>>> The ALSA sequencer timer stopped using the system timer years ago.
>>> And the difference between RT and 'normal' kernels is not the timer
>>> frequency.
>> I'm using jack with seq, so are you saying the ALSA sequencer timer is
>> using a higher timer rate than than is set in the kernel,
> It's using different timers (RTC or hrtimer), which support higher
> precision.
>
>> and there's no need for me to bother with this,
> Yes.  If you want to be sure, check that G3 appears in /proc/asound/timers
> when you're using the sequencer.
>
>
> Regards,
> Clemens
It does.

Thanks for the info.

Regards,
Alex.