[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [bgl-discuss] Using TAU
Jeff,
The first routine psi189 takes 26 microseconds per call and the
second xsi189 takes 15. So, they are lightweight but not insignificant. If
we have to choose, I would mask the instrumentation in the second
subroutine xsi189 and enable the instrumentation in the first one. You may
use a selective instrumentation file to do this.
Thanks,
- Sameer
On Tue, 28 Mar 2006, Jeff Nucciarone wrote:
> Sameer Shende wrote:
> > Jeff,
> > I am a developer of TAU and I'll be happy to assist you with
> > TAU related issues. Yes, you could use TAU to evaluate the performance of
> > your application. We have installed TAU in
> > /soft/tools/tau/tau-2.15.1 on the ANL BGL.
> > Thanks,
> > - Sameer
>
> Sameer,
>
> Thanks.
>
> basically I have two routines in my code that burn the majority of the
> cpu time. Neither of these routines is very large but the sheer number
> calls is what takes the time.
>
> Here's a snippet of the profile:
>
> > Flat profile:
> >
> > Each sample counts as 0.01 seconds.
> > % cumulative self self total
> > time seconds seconds calls Ks/call Ks/call name
> > 39.21 29465.17 29465.17 1093998608 0.00 0.00 psi189
> > 20.49 44860.82 15395.65 991396136 0.00 0.00 xsi189
> > 5.41 48925.87 4065.05 32 0.13 1.99 main
> > 5.31 52917.05 3991.18 _pxldmod
> > 4.01 55932.73 3015.68 BGLML_Messager_tree_advance
> > 3.92 58880.76 2948.03 991396136 0.00 0.00 xpline
> > 3.09 61201.20 2320.44 2187997216 0.00 0.00 aaint
> > 1.94 62655.48 1454.28 processTreeSendQueue
>
>
> As you can see, routines psi189 and xsi189 are called enough times to
> choke a healthy horse. Would instrumenting just these two routines cause
> any problems or should throttling be applied?
>
> I'm primarily interested in fpu and cache reuse in these two routines.
>
> Thanks,
>
> --Jeff
>
>
> > ----------------------------------------------------------------------------
> > Sameer Shende,
> > Postdoctoral Research Associate,
> > Department of Computer and Information Science,
> > 1202 University of Oregon,
> > Eugene, OR 97403.
> > ph : (541) 346-0850 fax : (541) 346-5373.
> > ----------------------------------------------------------------------------
> >
> > On Mon, 27 Mar 2006, Ed Jedlicka wrote:
> >
> >
> >>Here is the TAU home page -
> >><http://www.cs.uoregon.edu/research/tau/home.php>http://www.cs.uoregon.edu/research/tau/home.php
> >>there is documentation, publications and contact information -
> >>
> >>At 03:38 PM 3/27/2006, Jeff Nucciarone wrote:
> >>
> >>>Greetings.
> >>>
> >>>I have an application on BG that I don't think is performing up to snuff.
> >>>I'd like to use a tool such as Tau to look for opportunities to tune the code.
> >>>
> >>>My first question would be to ask if Tau is indeed the correct tool for this.
> >>>
> >>>If so, is the a quick guide or a primer available for getting started?
> >>>
> >>>The application I have is relatively simple and is contained in a single
> >>>file. I usually compile by hand and don't bother with a makefile since it
> >>>is only a single file.
> >>>
> >>>tnx
> >>>
> >>>--Jeff
> >>>
> >>>--
> >>>Jeff Nucciarone nucci@xxxxxxx http://www.personal.psu.edu/nucci
> >>> Senior Research Programmer, High Performance Computing Group, ITS/ASET
> >>> The Pennsylvania State University
> >>> "Don't just do it........ do it right."
> >>>
> >>>
> >>
> >>- --------------------------------------------------------------------
> >>To add or remove yourself from this mailing list, use the 'notifyme'
> >>command on any BGL machine. To remove: notifyme -n, to add: notifyme -y.
> >>
> >
> >
> >
>
>
> --
> Jeff Nucciarone nucci@xxxxxxx http://www.personal.psu.edu/nucci
> Senior Research Programmer, High Performance Computing Group, ITS/ASET
> The Pennsylvania State University
> "Don't just do it........ do it right."
>
- --------------------------------------------------------------------
To add or remove yourself from this mailing list, use the 'notifyme'
command on any BGL machine. To remove: notifyme -n, to add: notifyme -y.