Hi Blake,

Thanks for the support. The problem is that I am really new to this
architecture area and so to Gem5. So I am hoping to have your support in
future also.

regards,
Rohith

On Wed, Nov 4, 2015 at 10:29 PM, Geoffrey Blake <[email protected]> wrote:

> Hi Rohith,
>
> It is possible to add the code you are suggesting in your message.  You
> will also have to modify the simulator atomic CPU models to do the proper
> things to not leak memory accesses,  and perform rollback of the
> transactional state in case of conflicts in addition to adding your
> database based conflict detection and versioning.  If you intend to extend
> this to timing mode and the caches you will also have to dig into the cache
> coherence and cache models as well unless faking it via the database is
> sufficient for your needs.
>
> In the past I had a version of gem5 that had LogTM support for Alpha, and
> it appears to still exist at http://www.eecs.umich.edu/~blakeg/ at the
> bottom of the page if you want to look at a potential starting point.
>
> Thanks,
> Geoff Blake
>
> On Wed, Nov 4, 2015 at 6:00 AM, rohith mathew <[email protected]>
> wrote:
>
> > Hi all,
> >
> > I'm trying to implement a syatem with a Hardware Transactional
> Memory(HTM)
> > in gem5 with x86. I just want to know whether it is possible to have a
> > database in the main memory for the operation of HTM. If possible can
> > anyone give me some suggestions to do that.
> >
> > Thanks in advance.
> >
> > regards,
> > Rohith
> > _______________________________________________
> > gem5-dev mailing list
> > [email protected]
> > http://m5sim.org/mailman/listinfo/gem5-dev
> >
> _______________________________________________
> gem5-dev mailing list
> [email protected]
> http://m5sim.org/mailman/listinfo/gem5-dev
>
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to