RTAI 3.2 on Athlon 1.3 Ghz

Paolo Mantegazza mantegazza at aero.polimi.it
Wed Jan 18 19:29:35 CET 2006


Tobias Bruckmann wrote:
> ----- Original Message -----
> From: "Paolo Mantegazza" <mantegazza at aero.polimi.it>
> To: "Tobias Bruckmann" <tobi at email.com>
> Subject: Re: RTAI 3.2 on Athlon 1.3 Ghz
> Date: Wed, 18 Jan 2006 16:46:14 +0100
> 
> 
>>Tobias Bruckmann wrote:
>>
>>>----- Original Message -----
>>>From: "Tobias Bruckmann" <tobi at email.com>
>>>To: rtai at rtai.org
>>>Subject: Re: RTAI 3.2 on Athlon 1.3 Ghz
>>>Date: Tue, 17 Jan 2006 17:24:23 -0500
>>>
>>>
>>>
>>>>Tobias Bruckmann wrote:
>>>>
>>>>
>>>>>Hi everyone,
>>>>>
>>>>>I'm presently starting my experiences with RTAI.
>>>>>
>>>>>As a first try, I wanted to install RTAI 3.2 on an Athlon 1.3 
>>>>>Ghz machine having an Asus mainboard with ALI chipset. 
>>>>>Installation works fine with a kernel v2.4.27 and ADEOS patch. 
>>>>>I made the following selections in the "critical" part of the 
>>>>>kernel configuration:
>>>>>
>>>>>- K7/Athlon Architecture
>>>>>- Deselect Module version information
>>>>>- MTRR (Memory Type Range Register) support
>>>>>- Local APIC support on uniprocessors
>>>>>- IO-APIC support on uniprocessors
>>>>>- Deselect Unsynced TSC support
>>>>>
>>>>>- Adeos support
>>>>>- Deselect Power Management support
>>>>>- Deselect ACPI Support - Deselect Sound card support
>>>>>
>>>>>- AGP support
>>>>>- DRI support
>>>>>
>>>>>The problem: The calibration test works fine, but the latency 
>>>>>example crashes. The application starts, but I cannot stop it 
>>>>>via CTRL-C. The machine hangs until I to a hard reset. I founds 
>>>>>some hints on the web to try to compile the kernel with K6 
>>>>>architecture, but it makes no difference.
>>>>>
>>>>>Then I tried RTAI 3.1, same kernel, same konfiguration. Here, I 
>>>>>have no problems with the latency test. Has anyone made the 
>>>>>same experience or an idea what to do next to get RTAI 3.2 
>>>>>running?
>>>>>
>>>>>Thanks a lot and regards,
>>>>
>>>>Which latency, user or kernel space. If is in user have you tried to
>>>>stop it by simply pressing the enter key?
>>>>
>>>>Paolo.
>>>>
>>>>
>>>>_______________________________________________
>>>>RTAI mailing list
>>>>RTAI at rtai.org
>>>>https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai
>>>>
>>>>Hallo,
>>>>
>>>>just to be sure I repeated the whole installation procedure, 
>>>>this time with a 2.4.29 kernel. I found the following results:
>>>>
>>>>RTAI 3.1 with kernel 2.4.27:
>>>>RTAI 3.1-testsuite works both with kernel- and user-space , 
>>>>calibration works
>>>>
>>>>RTAI 3.2 with kernel 2.4.27 and 2.4.29:
>>>>RTAI 3.2-testsuite works for the switcher example. Also the 
>>>>calibration seems to be work. Latency test and Preempt test 
>>>>crash, the screen freezes. Only hard reboot helps - CTRL-C or 
>>>>ENTER have no effect. This happens both for kernel- and 
>>>>user-space.
>>>>
>>>>I also tried the 3.1-testsuite with RTAI 3.2 - somehow the old 
>>>>testsuite seems to work. I will do some testing tomorrow.
>>>>
>>>>Somewhere I read that the testsuite doesn't work with an 
>>>>Ahtlon/K7-specific kernel (well, this is the processor type I 
>>>>chose) - could this be the reason? I will try to test an 
>>>>K6-kernel, but I don't want to loose performance using a wrong 
>>>>setting here...
>>>>
>>>>I will also try to find some output in the log files..
>>>>
>>>>Any ideas what could be wrong?
>>>>
>>>>Regards,
>>>>
>>>>Tobias
>>>>
>>>>--
>>>>___________________________________________________
>>>>Play 100s of games for FREE! http://games.mail.com/
>>>>
>>>>
>>>>_______________________________________________
>>>>RTAI mailing list
>>>>RTAI at rtai.org
>>>>https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai
>>>
>>>
>>>By the way: Does RTAI need kernel-support for APIC on a UP system 
>>>- or is it recommended? If not, I will try to build without - 
>>>maybe this helps. At captain.at was an hint in this direction...
>>>
>>
>>Provided your machine has an APIC it is reccomended, as in such a 
>>case the Linux timer (8254) is disturbed the least. Moroever when 
>>oneshot mode is used there will be not added jitterr due to 
>>accessing the ISA bus to program the timer + the fact that APIC 
>>timers programming is a lot faster.
>>
>>If you upgraded to 3.3 and could be happy with a periodic mode 
>>there would be even the possibility of not touching any of the 
>>timers used by Linux despite not having any APIC available.
>>
>>Paolo.
>>
>>
>>_______________________________________________
>>RTAI mailing list
>>RTAI at rtai.org
>>https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai
> 
> 
> Hi,
> 
> I tested some combinations of K6/K7 architecture and with/without kernel APIC support.
> 

Can you try with a K6 kernel even if you have a K7? UP APIc works fine 
on i386s and I don see why it should not on K6. When I've time and will 
be allowed to use non mine machines I'll have a try myself.

> I found out that the kernel APIC support crashes all examples on my machine. I disabled kernel APIC support and some examples from testsuite I tested made no problem. So it's all APIC-related.
> 
> Is this a problem of a bad APIC controller or a buggy BIOS? (Which means using another mainboard could help!)
> Or is it some kernel option, which disturbs RTAI?
> 
> Since my application should run with 1ms loop time (accurary within 5% would be enough) - do I have to care about the additional jitter when NOT using APIC support, or is it just some microseconds? Upgrading to 3.3 is (for the time being) not possible - I use MCA2 which supports RTAI-3.2 as the latest version...

No problem with that rithm, worst case latency will not change too much 
with respect to APIC and if you work in periodic mode there will be no 
difference.

Paolo.





More information about the Rtai mailing list