Re: reduce size of pv_pte
Masao Uebayashi <uebayasi <at> gmail.com>
2011-05-12 03:46:51 GMT
In the very long run, we will be able to cache a reference to a
vm_physseg on the fault handler stack, then pass (paddr_t, vm_page) to
pmap. I agree that PHYS_TO_VM_PAGE is a bad idea in general.
On Thu, May 12, 2011 at 6:25 AM, Lars Heidieker <lars <at> heidieker.de> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> On 04/26/11 00:40, YAMAMOTO Takashi wrote:
>> the following patch reduces the size of pv_pte, thus pv_entry and vm_page.
>> YAMAMOTO Takashi
> Hi, that's interesting. It is cutting a pv_entry from 40 bytes down to
> 32 bytes for 64 bit.
> I've a concern about the runtime requirements because of the
> PHYS_TO_VM_PAGE lookup which might be significant.
> Do all amd64 machines have only a few physical memory segments?
>> Index: include/pmap_pv.h
>> RCS file: /cvsroot/src/sys/arch/x86/include/pmap_pv.h,v