LAM/MPI logo

LAM/MPI General User's Mailing List Archives

  |   Home   |   Download   |   Documentation   |   FAQ   |   all just in this list

From: Apurva Bhatnagar (apurvanarayan_at_[hidden])
Date: 2007-11-08 02:11:06


hey i am not sending a zero byte message but when my arrow is negavtive i
get a return message size as zero..

On 11/7/07, Anthony Chan <chan_at_[hidden]> wrote:
>
>
> The clocks ain't synchronized in most clusters.
> MPE logging does clock alignment to align the timestamps based on a
> send/recv based synchronization algorithm whose accuracy is limited by
> latency of the network (which could be noise, hence even less accurate).
> If your cluster has a global clock, you could disable the sync code by
> setting the environment variable MPE_CLOCKS_SYNC=false, default is true.
> If your run involes more than 16 processes, the sync algorithm may not be
> the most accurate. In that case you can choose a better one (See
> mpe2/README). In general, it does not make much difference.
>
> BTW, you shouldn't send zero byte message, :)
>
> A.Chan
>
> On Wed, 7 Nov 2007, Apurva Bhatnagar wrote:
>
> > Hey are u talking about system clocks , which are different in
> heterogeneous
> > clusters...
> >
> > On 11/7/07, Apurva Bhatnagar <apurvanarayan_at_[hidden]> wrote:
> >>
> >> But can this be solved. i am using 7.1.4 i think...
> >>
> >> On 11/7/07, Anthony Chan <chan_at_[hidden]> wrote:
> >>>
> >>>
> >>> What version of MPE are you doing ?
> >>>
> >>> Do you mean your arrow is pointing backward when message size is zero
> ?
> >>> If this is the case, this is probably due to the clock synchronization
> >>> algorithm isn't accurate enough for the given MPI_Wtime resolution.
> >>> If this is the only case that you see backward arrow, I don't think
> >>> you need to worry about it.
> >>>
> >>> A.Chan
> >>>
> >>>
> >>> On Wed, 7 Nov 2007, Apurva Bhatnagar wrote:
> >>>
> >>>> Dear Sir,
> >>>> I am running LAM /MPI on my code on DE algorithm which requires
> string
> >>>> evaluations which i do it on multiple computers . I am facing a
> >>> problem with
> >>>> Jumpshot results.When i am using a uniform (homogeneous cluster) my
> >>> results
> >>>> seem ok but when i use a heterogeneous cluster some time i face
> >>> problem that
> >>>> my message send time becomes negative and message size becomes
> >>> ZERO..what
> >>>> does it signify,
> >>>> Shall i send in the printscreen of the jumpshot if its needed.
> >>>> Regards
> >>>> Apurva
> >>>>
> >>>>
> >>>>
> >>>> --
> >>>> " Life is a BICYCLE the more you work the faster you GO!! "
> >>>>
> >>>> Apurva Narayan
> >>>> Final Year Electrical Engg,
> >>>> Faculty of Engineering,
> >>>> Dayalbagh Educational Institute,
> >>>> Agra - 5
> >>>> Ph: +91-9412605116
> >>>> Email : apurvanarayan_at_[hidden]
> >>>> Web : http://appu.netfirms.com
> >>>>
> >>> _______________________________________________
> >>> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
> >>>
> >>
> >>
> >>
> >> --
> >> " Life is a BICYCLE the more you work the faster you GO!! "
> >>
> >> Apurva Narayan
> >> Final Year Electrical Engg,
> >> Faculty of Engineering,
> >> Dayalbagh Educational Institute,
> >> Agra - 5
> >> Ph: +91-9412605116
> >> Email : apurvanarayan_at_[hidden]
> >> Web : http://appu.netfirms.com
> >>
> >
> >
> >
> > --
> > " Life is a BICYCLE the more you work the faster you GO!! "
> >
> > Apurva Narayan
> > Final Year Electrical Engg,
> > Faculty of Engineering,
> > Dayalbagh Educational Institute,
> > Agra - 5
> > Ph: +91-9412605116
> > Email : apurvanarayan_at_[hidden]
> > Web : http://appu.netfirms.com
> >
>

-- 
" Life is a BICYCLE the more you work the faster you GO!! "
Apurva Narayan
Final Year Electrical Engg,
Faculty of Engineering,
Dayalbagh Educational Institute,
Agra - 5
Ph: +91-9412605116
Email : apurvanarayan_at_[hidden]
Web : http://appu.netfirms.com