Open markcmiller86 opened 5 years ago
@JustinPrivitera there is a note in the opening comment of this issue regarding performance for Curve plugin (ascii data) for curves with millions of points. Its slow for sure. That just seems like bizarre behavior given VisIt supposedly "scales well".
I do think there are a large number of ascii formats that are trivially decomposeable (e.g. that don't require the hammer of Conduit/Blueprint export) into a .visit
file to handle correctly. But, we could probably just "do the right thing" reading such a file into a parallel VisIt.
I am in the planning stages of that exact kind of behavior for very large STL files (on both read and write). It might be worth comparing notes and/or combining efforts to identify all the common/shared functionality for such cases.
@markcmiller86 Sure I'm happy to chat. Turning it into Blueprint and partitioning into multidomain is fairly simple.
We have several plugins supporting ASCII input; Curve2D, XYZ, PlainText probably others. They all have to deal with similar issues, bad chars, overflow, creation of mesh or curve objects, read options, etc.
We should consoidate all this functionality to a single plugin with various options
We don't really handle big curve data too well. We take forever to read it. We don't scale with it because it is treated as single block (though multi-block curves are now supported), we store it in memory of mdserver and engine, and viewer (as its renderable form). It can be several minutes to get an image for a curve of 100 million points. We ought to be able to handle that case ok.