On Mon, 9 Oct 2006, Holger Berger wrote:
On 10/9/06, Paul Durrant <[EMAIL PROTECTED]> wrote:
On 10/9/06, Holger Berger <[EMAIL PROTECTED]> wrote:
> Project proposal: 64k kernel project
>
> I propose a 64k kernel project. This project should contribute and
> support on an on-going basis:
>
> - A SPARC kernel which uses 64k pages instead of 8k pages ("dwarf
> pages") as default page size. The system should no longer use or
> support the usage of 8k pages
How would this work with network I/O? Do the IOMMUs support 64k pages?
AFAIK yes
I remember I/O (both the ingraindness of MAXBSIZE in segmap and the
exposure of PAGESIZE via VOP_GETPAGE) was a big killer argument against
the 64kB pagesize project back in its time. Even these days, I/O sizes are
often "few kB", especially on networks even with jumbo frames, but also
still on disks/filesystems :(
Do you really want to expose a 64k chunk of memory to IO just to DMA
in a single ethernet packet?
These are static buffers. We know that *some* memory gets wasted but
the amount should be limited
Depends a lot on your average "I/O chunk size".
FrankH.
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
