On Thu, 2008-10-02 at 06:40 -0700, Victor Gallardo wrote:
> > Victor, while at it, can you do a fixup patch on top of it that guards
> > the new feature with a Kconfig option like some of the other ones so
> > that the code for it doesn't get compiled in when building, for example.
> > for 405GP only ?
> >
> > The trick is to have the option not be part of the possible mask, so
> > that the compiler optimises out the feature tests as if (0) (gcc
> > nowadays is supposedly smart enough to rip off the code when it finds
> > such constructs).
>
> Hi Ben,
>
> Can you give an example of what you are asking for? I am not sure if I
> understand your request.
Well, if you look at the way emac_has_feature is implemented:
static inline int emac_has_feature(struct emac_instance *dev,
unsigned long feature)
{
return (EMAC_FTRS_ALWAYS & feature) ||
(EMAC_FTRS_POSSIBLE & dev->features & feature);
}
And now, if you look a few lines up, you see that various CONFIG_*
options define what is in EMAC_FTRS_POSSIBLE.
The trick is, you can thus make your new option only be part of
EMAC_FTRS_POSSIBLE if support for a 460EX based board has been enabled
or even better, one that uses a GPCS PHY. You do that by creating a new
Kconfig option such as CONFIG_EMAC_SUPPORTS_GPCS for example that gets
select'ed by the boards that need it.
That way, when compiling a kernel for a board that does -not- need it,
the feature bit will be absent from EMAC_FTRS_POSSIBLE. That will allow
the compiler to figure out that when emac_has_feature() is called for
that option, the result will always be 0. Thus the compiler gets to
optimize out all the code relative to that option.
Cheers,
Ben.
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-dev
