[Ktechlab-devel] The engine must be fixed. =|

[Alan Grimes]

Right now the engine consists of my own matrix and vector classes
(QuickMatrix and QuickVector). They're fairly robust and a fairly good
balance between performance and features, ie no dynamic rescaling, etc...)

The class Matrix is a much refactored version of what I found when I
started hacking on on ktechlab.

I've written a function called "validateLU" and it seems to be showing
that there ARE errors in the LU factorization algorithm. The current
engine does not do any row-permutations because the current system tries
to optimize things by not completely redoing a matrix that has only had
minor changes, that's one potential source of errors.

At this point I think we should move to using Eigen and completely junk
my idea of a "libktechmath"...

I submitted a patch that will agressively abort the program if the
engine doesn't produce a valid LU. This problem must be fixed before
this branch of ktechlab moves forward. the bug is that severe! =|

If this doesn't work, Ktechlab can't simulate jack. =(

-- 
New president: Here we go again...
Chemistry.com: A total rip-off.
Powers are not rights.