__switch_to test-and-branch ALTIVEC specific?

__switch_to test-and-branch ALTIVEC specific?

[Marcelo Tosatti]

Hi,

This test&branch looks AltiVec specific, any reason for not #ifdef'ing it
out?

--- arch/ppc/kernel/process.c.orig      2005-09-13 14:43:13.000000000 -0300
+++ arch/ppc/kernel/process.c   2005-09-13 14:43:51.000000000 -0300
@@ -287,11 +287,13 @@ struct task_struct *__switch_to(struct t
 #endif /* CONFIG_SPE */
 #endif /* CONFIG_SMP */

+#ifdef CONFIG_ALTIVEC
        /* Avoid the trap.  On smp this this never happens since
         * we don't set last_task_used_altivec -- Cort
         */
        if (new->thread.regs && last_task_used_altivec == new)
                new->thread.regs->msr |= MSR_VEC;
+#endif
 #ifdef CONFIG_SPE
        /* Avoid the trap.  On smp this this never happens since
         * we don't set last_task_used_spe

Re: __switch_to test-and-branch ALTIVEC specific?

[Kumar Gala]

This probably has to due with what happens on a G4 system with a  
kernel not build with Altivec.. However, I dont remember exactly what  
behavior is desired.

- kumar

On Sep 13, 2005, at 12:50 PM, Marcelo Tosatti wrote:

> Hi,
>
> This test&branch looks AltiVec specific, any reason for not #ifdef'ing
> it
> out?
>
> --- arch/ppc/kernel/process.c.orig      2005-09-13 14:43:13.000000000
> -0300
> +++ arch/ppc/kernel/process.c   2005-09-13 14:43:51.000000000 -0300
> @@ -287,11 +287,13 @@ struct task_struct *__switch_to(struct t
>  #endif /* CONFIG_SPE */
>  #endif /* CONFIG_SMP */
>
> +#ifdef CONFIG_ALTIVEC
>         /* Avoid the trap.  On smp this this never happens since
>          * we don't set last_task_used_altivec -- Cort
>          */
>         if (new->thread.regs && last_task_used_altivec == new)
>                 new->thread.regs->msr |= MSR_VEC;
> +#endif
>  #ifdef CONFIG_SPE
>         /* Avoid the trap.  On smp this this never happens since
>          * we don't set last_task_used_spe
>
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev@ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc-dev
>

Re: __switch_to test-and-branch ALTIVEC specific?

[Marcelo Tosatti]

On Tue, Sep 13, 2005 at 02:05:57PM -0500, Kumar Gala wrote:
> This probably has to due with what happens on a G4 system with a  
> kernel not build with Altivec.. However, I dont remember exactly what  
> behavior is desired.

What would be the appropriate set of definitions to use, then ? 

#if defined (CONFIG_ALTIVEC || CONFIG_6xx) ?

Re: __switch_to test-and-branch ALTIVEC specific?

[Kumar Gala]

On Sep 13, 2005, at 3:35 PM, Marcelo Tosatti wrote:

> On Tue, Sep 13, 2005 at 02:05:57PM -0500, Kumar Gala wrote:
>
>> This probably has to due with what happens on a G4 system with a
>> kernel not build with Altivec.. However, I dont remember exactly what
>>
>
>
>> behavior is desired.
>>
>
> What would be the appropriate set of definitions to use, then ?
>
> #if defined (CONFIG_ALTIVEC || CONFIG_6xx) ?

That's probably better.  Hopefully Ben will have some reason if what  
you suggest is not sufficient.

- k

-dev and -embedded cross-posting

[Eugene Surovegin]

Guys,

let's not cross-post -dev posts here. In general, -embedded isn't 
a totally separate list, it just covers embedded specific stuff, which 
is not interesting to several loud -dev readers :). Things which 
aren't embedded specific should go -dev.

I think, everybody *interested* in general PPC32 development issues 
here, are also subscribed to -dev list.

Thanks.

-- 
Eugene

Re: __switch_to test-and-branch ALTIVEC specific?

[Segher Boessenkool]

>> This probably has to due with what happens on a G4 system with a
>> kernel not build with Altivec.. However, I dont remember exactly what
>> behavior is desired.

A kernel not build with AltiVec support will just
SIGILL any user process that tries to use it (and
MSR[VEC] = 0).

The code here can safely be #ifdef'ed out (but note
I only looked at the code, I didn't test it ;-) )


Segher

Re: __switch_to test-and-branch ALTIVEC specific?

[Paul Mackerras]

Marcelo Tosatti writes:

> This test&branch looks AltiVec specific, any reason for not #ifdef'ing it
> out?

No, your patch looks fine.  (/me wishes patchwork had a "Queued"
state.)

Paul.

Re: __switch_to test-and-branch ALTIVEC specific?

[Benjamin Herrenschmidt]

On Tue, 2005-09-13 at 14:05 -0500, Kumar Gala wrote:
> This probably has to due with what happens on a G4 system with a  
> kernel not build with Altivec.. However, I dont remember exactly what  
> behavior is desired.

No, we should never set MSR_VEC on a kernel not built with
CONFIG_ALTIVEC.

Ben.

Re: __switch_to test-and-branch ALTIVEC specific?

[Benjamin Herrenschmidt]

On Tue, 2005-09-13 at 16:43 -0500, Kumar Gala wrote:
> On Sep 13, 2005, at 3:35 PM, Marcelo Tosatti wrote:
> 
> > On Tue, Sep 13, 2005 at 02:05:57PM -0500, Kumar Gala wrote:
> >
> >> This probably has to due with what happens on a G4 system with a
> >> kernel not build with Altivec.. However, I dont remember exactly what
> >>
> >
> >
> >> behavior is desired.
> >>
> >
> > What would be the appropriate set of definitions to use, then ?
> >
> > #if defined (CONFIG_ALTIVEC || CONFIG_6xx) ?
> 
> That's probably better.  Hopefully Ben will have some reason if what  
> you suggest is not sufficient.

Just #ifdef CONFIG_ALTIVEC is fine

Ben.