RE: e1000 performance hack for ppc64 (Power4)

[Feldman, Scott]

> So in summary I think this is mainly an alignment issue and 
> it is likely to help other platforms as well so should be 
> considered for the driver.

But we're applying the alignment constraint in the wrong direction.  The
e1000 h/w doesn't have an alignment constraint, it's your arch.  Dave
Hansen's suggestion is in the right direction; let's see what Anton's
response is.

-scott