Re: [Gems-users] Page Fault in Ruby

[Kevin Moore <kmoore@xxxxxxxxxxx>]

Mrinmoy,
	Yes, I think you're right.  Ruby operates on physical addresses after 
address translation.  For normal memory operations, a TLB miss handler 
or page fault handler will run, if necessary, before ruby sees the 
memory request.
--Kevin

On Jun 8, 2006, at 11:33 AM, Mrinmoy Ghosh wrote:

> Hi Dan,
>              Thanks for your answer. I was enquiring about page faults
> in the target or the simlated virtual machine.
> I have one small question. Since you say here that Ruby is not aware of
> Paging, I think you mean the following. Please correct me  if I am  
> wrong.
> It is the responsibility of the TLB fault handler to  check the page
> table in memory and check for page faults. If a page table entry is
> nulll the page fault handler is invoked and the page is brought in from
> the disk. As far as Ruby is concerned, it assumes that any request sent
> to it from the caches will always be present, as it would have been
> handled by the page fault handler before.
>
> Please let me know if I am clear in my understanding of this.
>
> Mrinmoy
> Dan Gibson wrote:
>
> > Hi Mrinmoy,
> >
> > Would you be referring to page faults on the host machine, or page
> > faults on the target machine?
> >
> > For the former:
> > Ruby queries its host machine when stats are dumped for the number of
> > page faults encoutered during simulation, but it is not "aware" of a
> > page fault on its host as it occurs.
> >
> > For the latter:
> > Ruby is not directly aware of paging activity on its target. Since
> > paging is a system-level phenomenon, you might be able to caputre 
> > paging
> > with a Simics hap callback (see the Simics Reference Manual and the
> > Simics Programming Guide for details), or possibly by observing the
> > dynamic instruction stream and signaling when a parcicular physical PC
> > is reached.
> >
> > Regards,
> > Dan Gibson
> >
> > Mrinmoy Ghosh wrote:
> >
> >
> >
> > > Hi,
> > >    I was interested to know, if ruby memory module signals Page
> > > Faults, and if yes, some pointers in the code as to how or where 
> > > this is
> > > done.
> > >
> > > Mrinmoy
> > >
> > > _______________________________________________
> > > Gems-users mailing list
> > > Gems-users@xxxxxxxxxxx
> > > https://lists.cs.wisc.edu/mailman/listinfo/gems-users
> > _______________________________________________
> > Gems-users mailing list
> > Gems-users@xxxxxxxxxxx
> > https://lists.cs.wisc.edu/mailman/listinfo/gems-users
> _______________________________________________
> Gems-users mailing list
> Gems-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/gems-users