[Xenomai-core] FPU-Problems on a system without FPU?

[Xenomai-core] FPU-Problems on a system without FPU?

[Niklaus Giger]

Hi

I studied a little bit my buildbot reports on a PPC405 (without FPU) regarding 
the FPU problem. Grepping through the saved reports on the server enabled me 
to pinpoint the changes quickly. I found the following situation:

Last good check with the quick build was build 58 1604
http://ngiger.dyndns.org/buildbot/hcu3_q/builds/58/step-xenotest/0

First bad check with the quick build was build 59 with revision 1605 
http://ngiger.dyndns.org/buildbot/hcu3_q/builds/59/step-xenotest/0

But the full build 13 with revision 1604 already had the failure!
http://ngiger.dyndns.org/buildbot/hcu3_f/builds/13/step-xenotest/0
The full build 12 had revision 1599 and was good.
http://ngiger.dyndns.org/buildbot/hcu3_f/builds/12/step-xenotest/0

May be this helps to pinpoint the FPU problem a bit.

Question: Is reporting 1 != 2 always the expected output of the FPU check?

A bad check looks like:
== Testing FPU cXenomai: suspending kernel thread c1ba8ea0 ('rtk5/0') at 
0xc30303ec after exception #1792
heck routines...
r0: 1 != 2
<..>
r31: 1 != 2
== FPU check routines: OK.
== Threads: sleeper_ufps-0 rtk-1 rtk-2 rtk_fp-3 rtk_fp-4 rtk_fp_ufpp-5 
rtk_fp_ufpp-6 rtup-7 rtup-8 rtup_ufpp-9 rtup_ufpp-10 rtus-11 rtus-12 
rtus_ufps-13 rtus_ufps-14 rtuo-15 rtuo-16 rtuo_ufpp-17 rtuo_ufpp-18 
rtuo_ufps-19 rtuo_ufps-20 rtuo_ufpp_ufps-21 rtuo_ufpp_ufps-22
RTT|  00:01:59
RTH|ctx switches|-------total
RTD|           5|           5


Unfortunately I still do not have a x86-box at home where I could run buildbot 
including xeno-test.

Best regards

-- 
Niklaus Giger

Re: [Xenomai-core] FPU-Problems on a system without FPU?

[Wolfgang Grandegger]

Niklaus Giger wrote:
> Hi
> 
> I studied a little bit my buildbot reports on a PPC405 (without FPU) regarding 
> the FPU problem. Grepping through the saved reports on the server enabled me 
> to pinpoint the changes quickly. I found the following situation:
> 
> Last good check with the quick build was build 58 1604
> http://ngiger.dyndns.org/buildbot/hcu3_q/builds/58/step-xenotest/0
> 
> First bad check with the quick build was build 59 with revision 1605 
> http://ngiger.dyndns.org/buildbot/hcu3_q/builds/59/step-xenotest/0
> 
> But the full build 13 with revision 1604 already had the failure!
> http://ngiger.dyndns.org/buildbot/hcu3_f/builds/13/step-xenotest/0
> The full build 12 had revision 1599 and was good.
> http://ngiger.dyndns.org/buildbot/hcu3_f/builds/12/step-xenotest/0
> 
> May be this helps to pinpoint the FPU problem a bit.
> 
> Question: Is reporting 1 != 2 always the expected output of the FPU check?
> 
> A bad check looks like:
> == Testing FPU cXenomai: suspending kernel thread c1ba8ea0 ('rtk5/0') at 
> 0xc30303ec after exception #1792
> heck routines...
> r0: 1 != 2
> <..>
> r31: 1 != 2
> == FPU check routines: OK.
> == Threads: sleeper_ufps-0 rtk-1 rtk-2 rtk_fp-3 rtk_fp-4 rtk_fp_ufpp-5 
> rtk_fp_ufpp-6 rtup-7 rtup-8 rtup_ufpp-9 rtup_ufpp-10 rtus-11 rtus-12 
> rtus_ufps-13 rtus_ufps-14 rtuo-15 rtuo-16 rtuo_ufpp-17 rtuo_ufpp-18 
> rtuo_ufps-19 rtuo_ufps-20 rtuo_ufpp_ufps-21 rtuo_ufpp_ufps-22
> RTT|  00:01:59
> RTH|ctx switches|-------total
> RTD|           5|           5
> 
> 
> Unfortunately I still do not have a x86-box at home where I could run buildbot 
> including xeno-test.

On a system without FPU you have to run this test with the option "-n". 
Gilles fixed a bug around that recently. You need commit 1603, at least. 
(look for the thread "switchtest on systems without FPU").

Wolfgang.
x