Interpreting output of WP34s M.LU (likely bug)

[supernumero]

(11-10-2019 09:50 PM)Paul Dale Wrote:  I think you'll find the pivots are zero based not one based, they are also not necessarily a permutation. An unpivoted matrix would then end up with pivots 012 -> 12. You 112 is saying to use the second row first and then the rest of the updated matrix in order (no more pivoting).

You are correct about the forward and back substitution process to solve systems of equations.
Unlike the 15C, the 34S has no space to remember that a matrix is in LU form (the 15C stored pivots in the matrix descriptor I believe).

Pauli

Thanks, Pauli. This clarifies the output of LINEQS, which I was misinterpreting as a bug. I read this article that explains that 3 bits of the "sign nibble" in diagonal entries M_ii of a matrix M in LU form are used to store row-reordering information in the 15c, but that article does not explain how to interpret the bits either. (In any event, these 3 bits of the sign nibble are hidden to the user of the 15c; they can only be accessed using synthetic tricks.)

SN