VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Wed Aug 13 2014 - 13:16:23 CDT

Hi,
  I'm not sure what Intel article you ran across, but I know that they have
made their own modified versions of Tachyon in the past (as far back as
the late 90s and early 2000s) for benchmarking and performance
testing of TBB and other tools. That said, I do have a Tachyon
build target for Xeon Phi in the current source code, but it is not
yet optimized for Xeon Phi in a very meaningful way yet. If your goal
is to render very large scenes rapidly, then the most straightforward
approach is to build one of the standard MPI versions of Tachyon,
and run it across a largish number of nodes. If you need help
compiling on Stampede, let me know.

Cheers,
  John Stone
  vmd_at_ks.uiuc.edu

On Wed, Aug 13, 2014 at 03:33:28PM +0000, Anne Bowen wrote:
> Hi John,
>
> I have another questions regarding compiling Tachyon on an intel cluster w. xenon phis (stampede here at tacc). I have been using the default linux-mpi make but am wondering if there are other options to optimize Tachyon (either the parallel performance or to use the phis) that I should be using. I stumbled on an Intel article regarding Tachyon on the Xeon Phi?s so I thought I should check!
>
>
> Thanks!
> Anne
>
> On Sep 4, 2013, at 11:46 AM, John Stone <johns_at_ks.uiuc.edu> wrote:
>
> > Anne,
> > Yes, you need to download Tachyon separately and compile from the
> > source on my Tachyon page (the one you listed below).
> > Let me know if you have questions about getting it compiled.
> >
> > Cheers,
> > John Stone
> > vmd_at_ks.uiuc.edu
> >
> > On Wed, Sep 04, 2013 at 11:17:58AM -0500, Anne Dara Bowen wrote:
> >> Hi John,
> >> I have compiled vmd from source but didn't see any options enable mpi for
> >> Tachyon. Do you mean I need to download the tachyon source separately
> >> (from here? http://jedi.ks.uiuc.edu/~johns/raytracer/) and compile that
> >> version?
> >> Thank you!
> >> Anne
> >> On Sep 4, 2013, at 11:06 AM, John Stone <johns_at_ks.uiuc.edu> wrote:
> >>
> >> Hi,
> >> In order to run Tachyon with multi-node parallelism via MPI, you will
> >> have to re-compile Tachyon from source code, against your local MPI
> >> installation. Unlike POSIX threads or other parallel programming
> >> interfaces, MPI doesn't provide binary compatibility, so in practice,
> >> one must always recompile MPI applications from source code for every
> >> different cluster or supercomputer MPI installation you run on.
> >> The Tachyon binaries that we ship with VMD are not compiled with MPI
> >> enabled due to these issues, they only use threads by default. It
> >> should
> >> be very straightforward to compile an MPI-enabled Tachyon on your
> >> machine,
> >> but you'll have to download the source code, edit the Make-config file
> >> as necessary, and compile it using your local MPI on your machine.
> >>
> >> Cheers,
> >> John Stone
> >> vmd_at_ks.uiuc.edu
> >>
> >> On Wed, Sep 04, 2013 at 10:51:22AM -0500, Anne Dara Bowen wrote:
> >>
> >> Hello,
> >>
> >> I have been trying to submit a parallel batch render (via sge on a
> >> linux cluster) using tachyon using a command like:
> >>
> >> /share/apps/vmd/1.9/lib/tachyon_LINUXAMD64 -trans_vmd +V -fullshade
> >> -aasamples 12 -rescale_lights 0.3 -add_skylight 1.0
> >> /fullpath/test.00000.dat -o /fullpath/test.00000.tga
> >>
> >> I am specifying the number of nodes/cores in the qsub script but each
> >> time only one node is used (output using the verbose +V option...CPU
> >> Information: Node 0: 8 CPUs, CPU Speed 1.00, Node Speed 8.00
> >> Name: c206-110.longhorn Total CPUs: 8 Total Speed: 8.000000)
> >> regardless of how many nodes I submit it to?and will still specify 8
> >> CPUs even when I only submit the job using 1,2 or 4 cores.
> >>
> >> Is there something I am doing incorrectly or that I need to add?
> >>
> >> (I have also tried to submit the render with the flavor of mpirun the
> >> machine I am using likes: e.g. mpirun
> >> /share/apps/vmd/1.9/lib/tachyon_LINUXAMD64 therestoftheoptionsthesame
> >> and also playing with the -numthreads option but this seems to have no
> >> effect)
> >>
> >> Thank you,
> >> Anne
> >>
> >> --
> >> NIH Center for Macromolecular Modeling and Bioinformatics
> >> Beckman Institute for Advanced Science and Technology
> >> University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> >> http://www.ks.uiuc.edu/~johns/ Phone: 217-244-3349
> >> http://www.ks.uiuc.edu/Research/vmd/
> >
> > --
> > NIH Center for Macromolecular Modeling and Bioinformatics
> > Beckman Institute for Advanced Science and Technology
> > University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> > http://www.ks.uiuc.edu/~johns/ Phone: 217-244-3349
> > http://www.ks.uiuc.edu/Research/vmd/
> 

-- 
NIH Center for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
http://www.ks.uiuc.edu/~johns/           Phone: 217-244-3349
http://www.ks.uiuc.edu/Research/vmd/