[Adeos-main] ADEOS NIOS II port - definitions in include/asm-nios2/ipipe.h

[Adeos-main] ADEOS NIOS II port - definitions in include/asm-nios2/ipipe.h

[Martin Elshuber]

Hello,

I faced a problem with Xenomai xntimer when using the
http://download.gna.org/adeos/patches/v2.6/nios2/adeos-ipipe-2.6.30-nios2-1.2-00.patch
patch.

xntimer uses the ipipe abstractions for hrclock and hrtimer and relies 
on the constants

#define __ipipe_hrtimer_freq    1000000 /* 1 us */
#define __ipipe_hrclock_freq    nasys_clock_freq

defined in include/asm-nios2/ipipe.h to do a final adjustment.

The two definitions above will fail, if either the hrtimer is not 
clocked with 1Mhz, or
the hrclock is clocked with another frequency than nasys_clock_freq.
I am quite new to ADEOS, but guess that

#define __ipipe_hrtimer_freq    na_hrtimer_clock_freq
#define __ipipe_hrclock_freq    na_hrclock_clock_freq

is a better choice. At least for me this solves the problem.

Martin Elshuber

Re: [Adeos-main] ADEOS NIOS II port - definitions in include/asm-nios2/ipipe.h

[Philippe Gerum]

On Fri, 2011-04-08 at 17:21 +0200, Martin Elshuber wrote:
> Hello,
> 
> I faced a problem with Xenomai xntimer when using the
> http://download.gna.org/adeos/patches/v2.6/nios2/adeos-ipipe-2.6.30-nios2-1.2-00.patch
> patch.
> 
> xntimer uses the ipipe abstractions for hrclock and hrtimer and relies 
> on the constants
> 
> #define __ipipe_hrtimer_freq    1000000 /* 1 us */
> #define __ipipe_hrclock_freq    nasys_clock_freq
> 
> defined in include/asm-nios2/ipipe.h to do a final adjustment.
> 
> The two definitions above will fail, if either the hrtimer is not 
> clocked with 1Mhz, or
> the hrclock is clocked with another frequency than nasys_clock_freq.
> I am quite new to ADEOS, but guess that
> 
> #define __ipipe_hrtimer_freq    na_hrtimer_clock_freq
> #define __ipipe_hrclock_freq    na_hrclock_clock_freq
> 
> is a better choice. At least for me this solves the problem.

Agreed, this was sloppy, only fitting the DE2 I used for porting Xenomai
over nios2 mmu-less, and I was just about to propagate this bug to the
next patch...

Yet another set of macro is involved (instead of na_*) with the new
design I've been using lately for the nio2mmu port over 2.6.35, but at
least they are consistent this time, wrt hrtimer/hrclock settings.
Thanks for the heads up.

> 
> Martin Elshuber
> 
> 
> _______________________________________________
> Adeos-main mailing list
> Adeos-main@domain.hid
> https://mail.gna.org/listinfo/adeos-main

-- 
Philippe.