Re: [RFC] pte equality tests

[Rik van Riel]

On Sun, 2 Jan 2000, Brian Parris wrote:
> On Fri, 31 Dec 1999, David S. Miller wrote:
> >    Date:   Fri, 31 Dec 1999 20:34:31 -0500
> >    From: Jeff Dike <[e-mail]>
> > 
> >    So, would it be reasonable to replace the equality tests with a test like
> > 
> > Your analysis is correct (it can/does happen), however I don't
> > think it happens often enough to worry about it.
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> i hope the above attitude never becomes popular in linux as it
> seems to be better suited to many MS products

You seem to be overlooking one thing.

If the tested version takes 5% longer than the original
and the exception only occurs in 1% of the cases (saving
100% overhead there), then you'll have more overhead
than you have by not doing that special case.

Sometimes a situation is so rare that special-casing it
has the opposite effect then expected.

regards,

Rik
--
The Internet is not a network of computers. It is a network
of people. That is its real strength.