[visit-developers] databases with per time step data and per simulation data

Miller, Mark C. miller86 at llnl.gov
Fri Feb 12 11:41:37 EST 2016


Hi Allen,

I think I am wrong here on one point.

You *could* opt to publish curve or mesh objects in a PopMD call that are actually defined with one of the axes being time. So, if you had just some global values in each file in a time series of files and you wanted those values to be 'plottable' as a curve object, you could opt to publish curve objects in PopMD. The only swizzle would be that when VisIt called GetCurve in your plugin for one of these curve objects, your plugin would have to do the work of iterating over all the files in the time series and teasing the appropriate global values out of each time state to construct the curve object to return in the call. I suppose same would be true for a GetMesh call too.

Does that make sense?

Mark


From: "Miller, Mark C." <miller86 at llnl.gov<mailto:miller86 at llnl.gov>>
Reply-To: "visit-developers at elist.ornl.gov<mailto:visit-developers at elist.ornl.gov>" <visit-developers at elist.ornl.gov<mailto:visit-developers at elist.ornl.gov>>
Date: Thursday, February 11, 2016 6:10 PM
To: "visit-developers at elist.ornl.gov<mailto:visit-developers at elist.ornl.gov>" <visit-developers at elist.ornl.gov<mailto:visit-developers at elist.ornl.gov>>
Subject: Re: [visit-developers] databases with per time step data and per simulation data

Hi Allen,

I think I am a bit confused by your question.

It sounds like you are asking if there is a way for meshes to have one of their dimensions be time rather than a spatial dimension. For example, can a curve object have an x-axis that represents time? Or, could we maybe have a 2D mesh in which one dimension is a spatial dimension, say x, and the other is a time dimension? And, in this context you mean 'time' in the way that databases know about time. So, in order for VisIt to assemble all the 'stuff' that goes into such an object, it would have to request (some) data from each of the states of a time-varying database.

We can do that in a somewhat limited way using the QueryOverTime feature. That allows VisIt to march over all the states of a database teasing a teeny tiny bit of data out of each one (a node or a zone and variables thereon) and then plot this as a curve. However, in order for that to work, the data has to be associated with the nodes or zones of a "mesh+variable" object defined over all time. Given current way VisIt works, you could have some global values in each time-state and affect such a QueryOverTime from those values.

What many of our codes doe in these circumstances is write what they call 'time histories'. The codes themselves assemble values over time to be output and then, at end of the run or perhaps periodically throughout the run, the output a whole curve object (typically to a Silo file), for which the x-axis represents time. These curve objects are sometimes stored in a file wholly separate from the main database. Sometimes, they are stored in a master file. Using this same approach, they can of course wind up creating even 2 and 3D mesh objects in which one of the dimensions is time. But, the data producer becomse responsible for creating such objects and the a user needs to know where to go to find them apart from the main database time series.'

Does that make sense?

In some way, I think your question relates to some generalization we could afford to do in the way VisIt thinks about what a mesh object is. Some thoughts are written here, http://www.visitusers.org/index.php?title=Ensemble_Navigation. I'd be curious to get your feedback on that.

Mark

From: Allen Sanderson <allen at sci.utah.edu<mailto:allen at sci.utah.edu>>
Reply-To: "visit-developers at elist.ornl.gov<mailto:visit-developers at elist.ornl.gov>" <visit-developers at elist.ornl.gov<mailto:visit-developers at elist.ornl.gov>>
Date: Thursday, February 11, 2016 4:41 PM
To: "visit-developers at elist.ornl.gov<mailto:visit-developers at elist.ornl.gov>" <visit-developers at elist.ornl.gov<mailto:visit-developers at elist.ornl.gov>>
Subject: [visit-developers] databases with per time step data and per simulation data

Hello,

I was wondering if any of the current readers have the ability to read databases with per time step data and per simulation data.

Say I have total energy at each time step for the whole simulation (single value per time step) and would like to draw a 2d plot energy vs time. Then I want a plot of the energy on the mesh at a particular time step.  All would be in the same database.

I can add this global var to the meta data as a curve and the reader would know to slurp it up over all time steps. Of the plot would not change with a change in the time slider but that would be the user responsibility.



Cheers,


Allen



Allen Sanderson

SCI Institute

University of Utah

www.sci.utah.edu<http://www.sci.utah.edu>



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://elist.ornl.gov/pipermail/visit-developers/attachments/20160212/03e8de1b/attachment.html>


More information about the visit-developers mailing list