Thu Sep 17 11:21:52 CEST 2009
Seems to be working. The only problem is the staging of pivoting,
since this has an influence on the data flow.
Ok, what I want is not echelon form, but gauss or gauss-jordan
elimination. The inner loop does this:
. | .... . / 0 | ....
P | .... -> P / 1 | ....
. | .... 0 | ....
There are two parameters: wheter to scale the pivot (P/1) and whether
to eliminate the top rows to zero (gauss-jordan) or not (gauss).
Ok. moved to gauss-jordan elimination.