Interesting that Steve E. is asking about mixed (FE-type) elements, and I'm glad for Lloyd's responses, since I'm currently working with Fluent to try to work around the inherent limited set of elements supported by DX, so that Fluent can update their Fluent2DX filter.
In particular, Fluent's people hit on the same idea as Steve, which is to degenerate cubes rather than decompose to tets. Just a few days ago, I provided them with apparently working "windings" to describe wedges and pyramids (and even tets themselves) as degenerate cubes. I'm waiting for some "real" data sets to come back with these filters applied so we can proceed to test the assertion that modules that do interpolation will either work or not work. Lloyd suggests they may have problems, and frankly that's what I believe as well, but the best test is to really try it and I don't know if anyone really has put this to the test before. Interestingly, DX can render these little degenerates, though of course, if you do something silly like display face normals, you'll find a few oddballs in there (pyramid has a "point" normal" and wedge has a "point" and an "edge" normal; the edge normal is parallel to the edge which is even odder). I was able to drop a MapToPlane and an Isosurface and a Streamline through a small test field, but all tests were slightly suspect. The Streamline seemed to terminate very quickly, but the test field was constant velocity, so I'm waiting for a more realistic data set to try again. The Isosurface appeared continuous but seemed to be lit peculiarly, possibly the result of the wacky normals being encountered during rendering. The MapToPlane was probably usable, but we won't know til I see "real" data to check continuity. The arguments for degenerating are: (a) it's the simplest solution (b) preserves the positions and element counts (same size arrays) (c) preserves the original connectivity for ShowConnections, i.e., you'd see hex boundaries where there were hexes, instead of tets as you would get after decomposing Argument against is (A) it may not work! At this time, neither Fluent nor I are volunteering to add new element types to DX, but maybe the time has come for some eager beaver(s) to consider this. Or at least, someone might consider what's involved in extending the definition of MultiGrid to permit mixed topology between members. I'll keep you posted on our experiment. Chris Pelkie Vice President/Scientific Visualization Producer Conceptual Reality Presentations, Inc. 30 West Meadow Drive Ithaca, NY 14850 [EMAIL PROTECTED]
