Hi Mike,

On Sat, Apr 21, 2012 at 1:04 PM, Mike Frysinger <[email protected]>wrote:

> On Thu, Apr 19, 2012 at 23:22, Bob Liu <[email protected]> wrote:
> > --- a/arch/blackfin/mach-common/pm.c
> > +++ b/arch/blackfin/mach-common/pm.c
> >
> >  #if defined(CONFIG_BFIN_EXTMEM_WRITEBACK) ||
> > defined(CONFIG_BFIN_L2_WRITEBACK)
> > +# ifdef CONFIG_BF60x
> > +__attribute__((l1_text))
> > +# endif
> >  static void flushinv_all_dcache(void)
>
> so add a Kconfig knob like DCACHE_FLUSH_L1 so any Blackfin cpu can use it ?
>
It maybe an anomaly on bf609, it requires FLUSH instruction in L1.

>
>
>  {
> > -     u32 way, bank, subbank, set;
> > -     u32 status, addr;
> > +     register u32 way, bank, subbank, set;
> > +     register u32 status, addr;
>
> you sure this is necessary ?  the "register" keyword tends to be a
> hint to the compiler, not a requirement.
>
I think it would be better to use register local variables when performing
flushinv_dcache_all to avoid potential dcache coherent issues.
Although without these "register" keyword, our compiler will probably use
register to store local variable too.


> -mike
> _______________________________________________
> Linux-kernel-commits mailing list
> [email protected]
> https://blackfin.uclinux.org/mailman/listinfo/linux-kernel-commits
>
_______________________________________________
Linux-kernel-commits mailing list
[email protected]
https://blackfin.uclinux.org/mailman/listinfo/linux-kernel-commits

Reply via email to