1-AUG-2019 SolveSys - Equation Library and Nonlinear Systems Solver

[Han]

(02-12-2017 11:19 PM)akmon Wrote:  Now that SVD is included in the core of the app, it´s not only more conformtable for updating, but it avoids errors from "SVD not found". Well done. Buy the way, remember to update your first post.

I haved noticed that slightly less accuracy in our famous system (A= -3,4985E-10, B=8,98E-11), I have also discorvered that if you are very accurate on your guess you´ll reach the exact solution, but it is too much to suppose and this system of equations are very unlikely to appear in general terms. What still concern me is what causes that fatal crash on VC just only choosing from one guess to other.

This is exactly what I was referring to regarding the loss in accuracy. Using the CAS version, you get just a slight advantage in accuracy. (see http://www.hpmuseum.org/forum/thread-2953.html)

But you are right, such a system tends to be atypical due to the singularity at B=0 (assuming (((A+B)/B)^2\) ), or A=-B

Right now, you should only need to update the SVD program if there's a bug fix. However, you will still have to deal with the CAS corruption (esp. on the emulator). On the hardware, it should not occur unless you either 1) recompiled a program or 2) experience a warmstart. My posting the code as source code probably doesn't help with (1).

The crash is due to the firmware's inability to handle undefined values in the Home view. The app is programmed as a "non-CAS" program. However, it still makes a few calls into the CAS. I hope that they will fix this in the next firmware.

Quote:About icons, three examples. The first, represents the graphical purpose of this app. The second, a wink to the firs Solvesys, and the last, another wink to the first EQlib from HP48.

Ahh, I do like the nod to either SolveSys or EQLIB. And compsystems' idea about using the old solve card as an icon is also a nice idea.