Another way to do it is to stream the data down a socket to dx, which is
something we have implemented for the cactus code - details at

http://www.cactuscode.org

There's a howto there, plus a dx module which reads streamed hdf5 data.

Tom


On Tue, 20 Mar 2001, Lloyd A Treinish wrote:

>
> sure is.  I do it almost every day.  A paper that discusses some of the
> ideas is at http://www.research.ibm.com/weather/vis/vis_design.htm or
> http://www.research.ibm.com/weather/vis/w_vis.pdf.  Typical static results
> are viewable at http://www.research.ibm.com/weather/NY/NY.html
>
> One thing that you need to determine is how tightly coupled you want the
> simulation and visualization.  In my case, we are generating files from the
> model in a specific format to drive a particular type of visualization
> suited toward qualitative assessment (e.g., browsing), feature detection,
> etc.  The idea is to quickly detect problems in the run or provide a heads
> up for post-processing analysis.  The files containing multiple variables
> are flat binary, one set per time step.  (There's more than one per time
> step because the model is nested on a multigrid).  I have a set of filter
> programs (shell scripts and C programs) invoked as ! filters in general
> array Import (and playing with caching) to assess the status of the model
> run, what's been computed, calculate statistics on what's been completed,
> read the data, etc.  All of this is under the covers of an application,
> where the user just selects which model run is of interest.   The
> simulation is running on an SP with shared filesystems with workstations
> (AIX, NT, Linux) on 100Mb or gigabit ethernet.  This approach allows a user
> to also work with runs already completed and resident on disk with the same
> application.  I also use multiple instances of the same application in
> script mode running independently on SMP nodes of the SP to generate fixed
> products in parallel such as those on the sample web site.  If you want
> something more tightly coupled, then you could have the simulation
> initiated from DX (e.g., ! filter), have the simulation invoke DX (e.g.,
> via dxlink) or set up more peer-to-peer communication.
>
> --------------------------
> Lloyd A. Treinish
> Deep Computing Institute
> IBM Thomas J. Watson Research Center
> P. O. Box 218
> Yorktown Heights, NY 10598
> 914-945-2770 (voice)
> 914-945-3434 (facsimile)
> [EMAIL PROTECTED]
> http://www.research.ibm.com/people/l/lloydt/
> http://www.research.ibm.com/weather
>
>
> "Frederick R. Phelan Jr." <[EMAIL PROTECTED]>@opendx.watson.ibm.com
> on 03/20/2001 10:05:46 AM
>
> Please respond to [email protected]
>
> Sent by:  [EMAIL PROTECTED]
>
>
> To:   [email protected]
> cc:
> Subject:  [opendx-users] Simultaneously Run Simulation and View Results in
>       DX. How?
>
>
>
> Is there any way to simultaneously run a simulation and view the results in
> dx as the numbers are being crunched?
>
> The simulation is a transient fluid flow program, in which results are
> continually being calculated at new time steps. Probably we would probably
> want to update the "live" picture, every 100 time steps or so.
>
> Thanks for any tips.
> Fred Phelan
>
> Frederick R. Phelan Jr., Ph.D.
> Multiphase Materials Group
> Polymers Division
> National Institute of Standards and Technology
>
> NIST, Bldg. 224/Rm. B108
> 100 Bureau Dr., STOP 8543
> Gaithersburg, MD 20899-8543
> 301.975.6761 (VOX)
> 301.975.4932 (FAX)
> [EMAIL PROTECTED]
> http://www.nist.gov/frederick_phelan
>
>
>
>

Reply via email to