Russell's Blog

I tracked down the problem that had been mucking up the unity-beta results from Ophidian. Either Pierre or I accidentally checked in a version of CUBE into the repository that wrote the output file data tables in the output in a different order. So, all this time I had thought it was a logic bug, or worse, a math error, turns out to be the result of importing the wrong table.

The weird thing is that it somehow passed my integrity tests. It was the right length, it had the right precision, it had the right boundary conditions, and it had an inflection point.

Er, it had lots of inflection points. I guess I should add a test for the entropy before I pass imported data to the rest of my code. These are all supposed to be nice, smooth, differentiable functions. I don't even use the data in this table -- I think it has something to do with benchmarking the convergence cycle. Maybe it's a poorly calculated magnetic shear. Maybe it's some sort of error distribution. All I know about it is that the values came out of a vector burried somewhere in the solver, and that it's labled "s".

That's the trouble with computers. You leave one thing out of place, and before you know it, they go stomping all over the place. Sooner or later, you will find yourself the victim of your own Techno Trousers. Anyway, here my plot from last week, but without the garbage.

No, I'm not going to explain what it means. You'll just have to wait for the paper.