On Sunday 02 November 2003 08:20 am, YAMAMOTO Takashi wrote:
> hi,
>
> attached patch shrinks pv_entry by eliminating pv_pmap member.
> while this increases cost to get pmap pointer a little,
> i think size of pv_entry is more important.
>
> any comments?
>
> YAMAMOTO Takashi

Is there even a reason to have pv_compare() look at these values?  Why not 
"sort" strictly based on the pv_entry pointers themselves, or pv->pv_pte?