Re: [rvm-research] Regarding Address of Object Reference in Field section of Object

Robin,

I realize that when I download a new version of repo and build it, it
builds and the poisoned BaseBase works perfectly. I have to carefully see
the changes in my code. But the diff  is potentially showing no changes
that affects the poisoned plan. It is strange. I will get back to you after
I find what is the issue. Thank you very much Robin for your help.

Cheers
Hari Krishnan


On Tue, Jan 14, 2014 at 5:12 PM, Robin Garner <[email protected]>wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> The Poisoned collector in current hg tip works for me in BaseBase,
> FullAdaptive and FastAdaptive configurations, at least for the handful
> of dacapo 2006 benchmarks I looked at.
>
> The Poisoned heap collector is also included in the nightly sanity
> checks, and gets 98%+ success, so odds are its either a stray change
> in your code base or something to do with your environment.
>
> Which version of JikesRVM are you building ?  What environment ?  What
> benchmark are you running ? Can you double-check that there are no
> changes to the code base, and do a fresh build ?
>
> cheers
> Robin
>
> On 14/01/14 18:18, Hari Krishnan wrote:
> > Rifat,
> >
> > The seg fault is thrown for poisoned implementation also. Not just
> > for my implementation. I hope you can let me know why  this is
> > happening. Thank you.
> >
> >
> > On Tue, Jan 14, 2014 at 12:55 AM, Hari Krishnan
> > <[email protected] <mailto:[email protected]>> wrote:
> >
> > Rifat,
> >
> > When I add in the constraints class to return true for the
> > specified method, the rvm when executing is giving a seg fault.
> >
> >
> > Cheers Hari
> >
> >
> > On Tue, Jan 14, 2014 at 12:16 AM, Rifat Shahriyar
> > <[email protected] <mailto:[email protected]>>
> > wrote:
> >
> > You need to set the constraint needsObjectReferenceReadBarrier() to
> > true. Look at PoisonedConstraints.
> >
> > Regards - Rifat
> >
> >
> > On 14/01/2014 5:11 pm, Hari Krishnan wrote:
> >> Robin,
> >>
> >> Thank you. I am looking at that code.  When I override
> >> objectReferenceRead it seems like it is never called. I am trying
> >> to understand how it works. Is there anything I have to do other
> >> than overriding that in my mutator class?
> >>
> >> Cheers Hari Krishnan.
> >>
> >>
> >> On Mon, Jan 13, 2014 at 11:47 PM, Robin Garner
> >> <[email protected] <mailto:[email protected]>> wrote:
> >>
> > The MMTk plan "poisoned" unconditionally sets and clears the
> > low-order bit of all object references - this is the regression
> > test for object read and write barriers.
> >
> > cheers
> >
> > On 14/01/14 15:53, Hari Krishnan wrote:
> >> Robin,
> >
> >> The process is basically to mark the references as I
> > trace the
> >> objects. Since I want to mark the instance field
> > references in each
> >> object, I want to mutate a single bit indicating the
> > reference is
> >> processed.  So yes to question (you want to add a single
> > bit of
> >> metadata to each object reference ?).Thank you for your
> > help. I
> >> want to try Read Barriers. I appreciate your quick
> > response. Is
> >> there any code that  already overrides read barrier?
> >
> >> Cheers Hari Krishnan.
> >
> >
> >> On Mon, Jan 13, 2014 at 10:45 PM, Robin Garner
> >> <[email protected] <mailto:[email protected]>
> > <mailto:[email protected] <mailto:[email protected]>>>
> > wrote:
> >
> >> Hari,
> >
> >> Just to clarify, you want to add a single bit of
> > metadata to each
> >> object reference ?
> >
> >> The "simplest"[1] solution is to use one of the two
> > low-order bits
> >> of each object reference (objects are always (at least)
> > 4-byte
> >> aligned). In order to do this you'll need a read barrier
> > that masks
> >> out this bit so that the mutator always sees aligned
> > pointers, and
> >> a write barrier that reinstates the metadata.
> >
> >> Alternatively, if this is too expensive, I'd suggest adding
> >> additional space to the object via the classloader - how
> > easy this
> >> is depends on how you want to use the data, how much
> > space you're
> >> prepared to use etc.
> >
> >> cheers
> >
> >> [1] Read barriers are never simple :-/
> >
> >> On 14/01/14 12:01, Hari Krishnan wrote:
> >>> Hi ,
> >
> >>> I want to set a bit, toggle a bit and test a bit in the
> > section
> >>> where each address of the instance field (object
> > references)
> >>> saved in objects. Is there any bit that I can use in
> > each object
> >>> fields for this purpose?
> >
> >>> Cheers Hari Krishnan
> >
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.11 (GNU/Linux)
> Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
>
> iQEcBAEBAgAGBQJS1cRaAAoJEHm9xBvoMhvaZN0H/RjYw3R4p/XINKzecexnS6jm
> F1018qNS5G51aE/FQ9uiLewbU+pwCVw4q81V+cYkGm2K4r0YDCOprpSfGI6aSTb2
> UtA6RYd2JJZXg1KvjIQ9+DJs6OCn11one7ffjB0X+rQSqRjIGwp2XzHFss7l7toU
> qIt4uTC3CCLxYKXdOKJtFOM3UgbgJo7a3Pis1ZFPVJLkE2VbGmQcbtEfuB5iSKd1
> DmSbjXZw+L9vfdqvMjufYwLbltBDsedb8+IzeClKqnIUsEDJEJTD4vwaU/weVLgO
> 62RF+xMvQrS7fF6ErBBl4gtpYOjJLgLWxiYA6xNZwWv1dThnKUEd5NwiNN1Ee+A=
> =pE7f
> -----END PGP SIGNATURE-----
>
>
> ------------------------------------------------------------------------------
> CenturyLink Cloud: The Leader in Enterprise Cloud Services.
> Learn Why More Businesses Are Choosing CenturyLink Cloud For
> Critical Workloads, Development Environments & Everything In Between.
> Get a Quote or Start a Free Trial Today.
>
> http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk
> _______________________________________________
> Jikesrvm-researchers mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers
>

------------------------------------------------------------------------------
CenturyLink Cloud: The Leader in Enterprise Cloud Services.
Learn Why More Businesses Are Choosing CenturyLink Cloud For
Critical Workloads, Development Environments & Everything In Between.
Get a Quote or Start a Free Trial Today. 
http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk

_______________________________________________
Jikesrvm-researchers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers