Often I’ll ditch a piece of software because it doesn’t fit my needs. Or because it’s not open source and a reasonable alternative comes along. Or because it frustrates the hell out of me. A year ago I developed all my unstructured meshes in Gmsh, which is a great little language for developing everything from very simple to very complex finite element geometries, which can then be imported (with some difficulty sometimes) into your PDE solver of choice. Or, as I’m doing in my chemical engineering class right now, you can write your own solver. Writing your own is solver is a wonderful exercise and very important, but insofar as you can use unstructured meshes, there’s really no reason to reinvent the wheel when Gmsh is so much better now than it was. There are obvious limitations to unstructured meshes for certain problems but they work very well for the types of problems I’m working on, especially because I often have absolutely no idea what the solution is going to look like for my dynamic systems. There is a nice little community around the software as well. It’s always great to find that software you initially dismissed over a year ago has totally reinvented itself and fixed all the issues you had with it previously. Regardless, check out Gmsh, even if you don’t do finite element calculations or calculations at all. It’s pretty fun to play around with.

Oh and it does a fantastic job of optimizing elements. This was done in 30 seconds on a single processor in a ~3-4 million element tetmesh, and basically completely eliminated the bottom 2 quadrants of mesh quality (the range in “mesh quality” here (I won’t explain the details), is 0 -> 1). If you can’t see it (someone who knows please explain to me why firefox on linux screws up image scaling so bad … I’m sure there is a simple solution but I definitely don’t know it), a blown up version is here.