Re: problem with 5.0-test1 on amd64

Bringing this back to RTAI list. I missed clicking a previous reply to all.

On 12/14/2015 12:55 AM, Sebastian Kuzminsky wrote:
> On 12/13/2015 03:15 PM, Paolo Mantegazza wrote:
>> At a first sight, I see a "<uses LINUX SYSCALLs>"  that shoule be a nomore definable option.
>> Also "timer setup = 1498 ns" is something strange on a machine with "lapic/62504000(Hz)", it should be in the range of tens or few hundreds of ns, and a "GHz CPU.
>> Finally, 0 ns for resched latency likely points at a missing calibration.
> Thanks for the feedback, i'll turn off linux syscalls.
I'm wondering why. Such a parameter is no more configurable with RTAI-5.0.
>
> The kernel output is from a virtual machine, so the timing numbers are
> probably going to be wonky-looking.  I don't have a bare-metal machine
> handy to test on.
>
> I disabled the build-time calibration, since i'm building on a different
> machine than i'm running the kernel on.  What's the process for
> calibrating on a particular machine, once the kernel is booted?
That is a thing we missed when decided for the self calibration. We are 
evaluating a few options for
a possible dynamic setting  of it.

A way is to use previously available  estimates.
Alternatively, at the moment you should run the latency test and get the 
calibration time from its average latency, setting the other parameters 
to 500 ns (roughly). Then you should reconfigure and remake it.
Paolo

_______________________________________________
Rtai mailing list
[email protected]
https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai