Re: [CapROS-devel] Big Pages

[Norman H. <no...@ca...>]

It is an entirely different point. It was not meant to address any  
point currently active on this list.
I like GPTs. Maybe I would use them if I were starting over.
My point was only that the kernel can exploit the large TLB entries  
provided in some architecture without changing kernel semantics and  
you can also provide larger disk pages with current hardware by  
changing kernel semantics.
Interactions with GPT are even more fruitful.

On 2007 Aug 17, at 7:03 AM, Jonathan S. Shapiro wrote:

> I think this note misses the point. The goal is not to have a single
> larger page size. The goal is to have an architecture in which  
> arbitrary
> page sizes (power of 2, realistically >= 512 bytes) can be supported.
> KeyKOS cannot do this, which is one reason that Coyotos (and  
> apparently
> now CapROS) has moved to GPTs.
>