[Qemu-devel] OS with only segmentation - will it be faster?

[Qemu-devel] OS with only segmentation - will it be faster?

[C K Kashyap]

[-- Attachment #1: Type: text/plain, Size: 495 bytes --]

Hi,
This is not strictly qemu related but I think people who have a good idea
about it must be on this list.
I was wondering if I had an app that requires a fixed quantity of memory -
sufficiently less than the available physical memory. Would it benefit from
getting rid of the paging mechanism in the OS/hardware?
As in, since the number of tasks are also fixed - we'd use only segmentation
to partition the VM area? Would eliminating the paging layer give good
returns?

-- 
Regards,
Kashyap

[-- Attachment #2: Type: text/html, Size: 539 bytes --]

Re: [Qemu-devel] OS with only segmentation - will it be faster?

[malc]

On Fri, 13 Aug 2010, C K Kashyap wrote:

> Hi,
> This is not strictly qemu related but I think people who have a good idea
> about it must be on this list.
> I was wondering if I had an app that requires a fixed quantity of memory -
> sufficiently less than the available physical memory. Would it benefit from
> getting rid of the paging mechanism in the OS/hardware?
> As in, since the number of tasks are also fixed - we'd use only segmentation
> to partition the VM area? Would eliminating the paging layer give good
> returns?
> 

Microsoft researchers working on Singularity claimed[1] that it does
provide significant speed improvements, and recent (few days ago)
discussion on comp.arch suggested as much (no need to go though TLB).

[1] http://channel9.msdn.com/shows/Going+Deep/Singularity-A-research-OS-written-in-C/
    (maybe the claim was made in some other video, in any case it should
     be there on Channel 9)

-- 
mailto:av1474@comtv.ru

Re: [Qemu-devel] OS with only segmentation - will it be faster?

[C K Kashyap]

[-- Attachment #1: Type: text/plain, Size: 1175 bytes --]

Thanks Malc .. I'll check out the video ... and perhaps ping you off the
list.


On Fri, Aug 13, 2010 at 10:15 AM, malc <av1474@comtv.ru> wrote:

> On Fri, 13 Aug 2010, C K Kashyap wrote:
>
> > Hi,
> > This is not strictly qemu related but I think people who have a good idea
> > about it must be on this list.
> > I was wondering if I had an app that requires a fixed quantity of memory
> -
> > sufficiently less than the available physical memory. Would it benefit
> from
> > getting rid of the paging mechanism in the OS/hardware?
> > As in, since the number of tasks are also fixed - we'd use only
> segmentation
> > to partition the VM area? Would eliminating the paging layer give good
> > returns?
> >
>
> Microsoft researchers working on Singularity claimed[1] that it does
> provide significant speed improvements, and recent (few days ago)
> discussion on comp.arch suggested as much (no need to go though TLB).
>
> [1]
> http://channel9.msdn.com/shows/Going+Deep/Singularity-A-research-OS-written-in-C/
>    (maybe the claim was made in some other video, in any case it should
>     be there on Channel 9)
>
> --
> mailto:av1474@comtv.ru
>



-- 
Regards,
Kashyap

[-- Attachment #2: Type: text/html, Size: 1784 bytes --]