All of lore.kernel.org
 help / color / mirror / Atom feed
* [Xenomai-help] Xenomai-PPC: One problem
@ 2005-12-07 14:51 smannori
  2005-12-07 15:43 ` Heikki Lindholm
  2005-12-07 15:57 ` Heikki Lindholm
  0 siblings, 2 replies; 4+ messages in thread
From: smannori @ 2005-12-07 14:51 UTC (permalink / raw)
  To: xenomai

Hi everybody,

thanks to your help now the interrupt shield option compile OK :)

Unfortunately, I must to leave the Xenomai FPU option disabled.

With this option active my PowerBook "Ubuntu-Xenomai" machine go crazy after
few seconds after the launch of the the "latency" test.

Gilles suspects that there is a conflict with the Altivec FPU units inside
the G4 CPU, because ADEOS does not save the Altivec FPU-exended register.

I think that I can live and prosper without FPU support in the kernel, but I
suppose also that someone is interested in this - possibile - limitation e/o
potential bug.

Simone Mannori






^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [Xenomai-help] Xenomai-PPC: One problem
  2005-12-07 14:51 [Xenomai-help] Xenomai-PPC: One problem smannori
@ 2005-12-07 15:43 ` Heikki Lindholm
  2005-12-07 15:57 ` Heikki Lindholm
  1 sibling, 0 replies; 4+ messages in thread
From: Heikki Lindholm @ 2005-12-07 15:43 UTC (permalink / raw)
  To: smannori; +Cc: xenomai

smannori@domain.hid kirjoitti:
> Hi everybody,
> 
> thanks to your help now the interrupt shield option compile OK :)
> 
> Unfortunately, I must to leave the Xenomai FPU option disabled.
> 
> With this option active my PowerBook "Ubuntu-Xenomai" machine go crazy after
> few seconds after the launch of the the "latency" test.
> 
> Gilles suspects that there is a conflict with the Altivec FPU units inside
> the G4 CPU, because ADEOS does not save the Altivec FPU-exended register.

I find this very unlikely. Unlike Intel SSE stuff, ppc altivec is a 
separate unit from the fpu. If your compiler generates altivec code 
(instead of normal fpu code) from math functions in the xenomai 
executable, then it might be the issue. AFAIK gcc uses altivec only 
after release 4.0. In case you have gcc 4 (ubuntu 5.10), you might be 
able to turn altivec code generation off by, say, "-mno-altivec".

-- Heikki Lindholm



^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [Xenomai-help] Xenomai-PPC: One problem
  2005-12-07 14:51 [Xenomai-help] Xenomai-PPC: One problem smannori
  2005-12-07 15:43 ` Heikki Lindholm
@ 2005-12-07 15:57 ` Heikki Lindholm
  2005-12-07 17:09   ` [Xenomai-help] Xenomai-PPC: One problem: data smannori
  1 sibling, 1 reply; 4+ messages in thread
From: Heikki Lindholm @ 2005-12-07 15:57 UTC (permalink / raw)
  To: smannori; +Cc: xenomai

smannori@domain.hid kirjoitti:
> Hi everybody,
> 
> thanks to your help now the interrupt shield option compile OK :)
> 
> Unfortunately, I must to leave the Xenomai FPU option disabled.
> 
> With this option active my PowerBook "Ubuntu-Xenomai" machine go crazy after
> few seconds after the launch of the the "latency" test.
> 
> Gilles suspects that there is a conflict with the Altivec FPU units inside
> the G4 CPU, because ADEOS does not save the Altivec FPU-exended register.
> 
> I think that I can live and prosper without FPU support in the kernel, but I
> suppose also that someone is interested in this - possibile - limitation e/o
> potential bug.

Be a bit more specific, please:
- compiler version,
- kernel version,
- i-pipe version,
- xenomai version,
- environment where the bug happens (what else is running, like 
X11/glQuake...)
I could take a look at this, if I can reproduce it.

-- Heikki Lindholm


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [Xenomai-help] Xenomai-PPC: One problem: data
  2005-12-07 15:57 ` Heikki Lindholm
@ 2005-12-07 17:09   ` smannori
  0 siblings, 0 replies; 4+ messages in thread
From: smannori @ 2005-12-07 17:09 UTC (permalink / raw)
  To: xenomai

Heikki :

> - compiler version,

simone@domain.hid:~$ gcc -v
Reading specs from /usr/lib/gcc/powerpc-linux-gnu/3.4.5/specs
Configured with: ../src/configure -v
--enable-languages=c,c++,f77,pascal,objc,ada --prefix=/usr
--libexecdir=/usr/lib --with-gxx-include-dir=/usr/include/c++/3.4
--enable-shared --with-system-zlib --enable-nls --without-included-gettext
--program-suffix=-3.4 --enable-__cxa_atexit --enable-libstdcxx-allocator=mt
--enable-clocale=gnu --enable-libstdcxx-debug --disable-softfloat
--enable-targets=powerpc-linux,powerpc64-linux --with-cpu=default32
--disable-softfloat powerpc-linux-gnu
Thread model: posix
gcc version 3.4.5 20050809 (prerelease) (Ubuntu 3.4.4-6ubuntu8)
simone@domain.hid:~$

> - kernel version,

simone@domain.hid:~$ uname -a
Linux xeno 2.6.14.3-ipipe #4 PREEMPT Wed Dec 7 15:20:25 CET 2005 ppc
GNU/Linuxsimone@domain.hid:~$

> - i-pipe version,
-rw-r--r--  1 nobody nogroup 122733 2005-11-29 15:16
adeos-ipipe-2.6.14-ppc-1.0-07.patch

> - xenomai version,
 cat /proc/xenomai/version
2.1

> - environment where the bug happens (what else is running, like
> X11/glQuake...)

1) Patch the kernel : OK
2) kernel compiling : OK
3) Boot OK
4) Xenomai config, make and make install: OK
5) Without xenomai process all is OK
6) launch /usr/xenomai/testsuite/latency (run)
7) the program start OK and runs
8) after few seconds the machine become instable... then freeze :(

Disabling FPU Xenomai support the machine is perfect (I have 2:30 uptime
without a single glitch).

Unfortunately the Radeon R350 (M10) inside the Powerbook is not officially
sopported by X a/o in the the kernel. There is a project

http://r300.sourceforge.net/R300.php

that I tried to compile and install BUT without success (working BUT NOT DRI
support :( ). Then I reinstalled from scratch a brand new Ubuntu 5.10.

> I could take a look at this, if I can reproduce it.

I think that you need an Apple machine similar to my Powerbook or something
with a G4 processor.

Many tanks in advance for your help.

Simone Mannori - www.inria.fr - Paris - Rocquencourt






^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2005-12-07 17:09 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-12-07 14:51 [Xenomai-help] Xenomai-PPC: One problem smannori
2005-12-07 15:43 ` Heikki Lindholm
2005-12-07 15:57 ` Heikki Lindholm
2005-12-07 17:09   ` [Xenomai-help] Xenomai-PPC: One problem: data smannori

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.