Weihang,
It is safe to comment out any instance of makePrefetch(). These do not
modify the architectural state of the processor model.
Luke
On Tue, 30 May 2006, Weihang Jiang wrote:
> Thanks a lot, Luke. If I want to isolate the effect of prefetch, is it safe
> to just comment these two places?
>
> On 5/30/06, Luke Yen <lyen@xxxxxxxxxxx> wrote:
> >
> > Weihang:
> >
> > makePrefetch() is called by Opal for store permission prefetching as
> > well as for prefetch instructions. It is called by both store and atomic
> > instructions for the former as well as prefetch instructions for the
> > latter.
> >
> > For the store permission prefetching the request is issued as soon as
> > the address is calculated, but the store operand is not yet ready.
> > Obviously its effectiveness is limited by the
> > latency between address calculation and store operand calculation. Unless
> > you have a workload that performs a lot of stores it shouldn't make a
> > difference
> > whether this feature is turned on or off.
> >
> > The other prefetch case is for SPARC prefetch instructions. Currently
> > we treat all prefetch instructions as loads. In contrast, SPARC defines
> > other types of prefetching such as TLB entry prefetching and store
> > prefetching.
> >
> >
> > Luke
> >
> >
> > On Mon, 29 May 2006, Weihang Jiang wrote:
> >
> > > I grepped through the opal, and found out that ruby->prefetch() is
> > called in
> > >
> > > store_inst_t::Execute() and prefetch_inst_t::accessCache( ). It is easy
> > to
> > > understand the second case, but why ruby prefetch is called in
> > > store_inst_t::Execute() ? Any reply is really appreciated.
> > >
> > > On 5/29/06, Weihang Jiang <weihang.jiang@xxxxxxxxx> wrote:
> > > >
> > > > Hi,
> > > > I see that opal module calls ruby module through two interfaces,
> > > > makePrefetch() and makeRequest(). Can anyone explain when
> > > > makePrefetch() should be called? How is it different from
> > > > makeRequest()? Any documents about it? Thanks
> > > >
> > > > Regards,
> > > >
> > > > --
> > > > Weihang Jiang
> > > >
> > >
> > >
> > >
> > > --
> > > Weihang Jiang
> > >
> > _______________________________________________
> > Gems-users mailing list
> > Gems-users@xxxxxxxxxxx
> > https://lists.cs.wisc.edu/mailman/listinfo/gems-users
> >
>
>
>
> --
> Weihang Jiang
>
|