WP34s: SLVQ implementation

[Dieter]

I just noticed that SLVQ may return results with slight errors even in relatively simple cases:

  -1 ENTER

  -1 ENTER

   1 SLVQ

 

   WP34s results             correct (16 digits)

 

   0,61803 39887 49895 0     0,61803 39887 49894 8

   1,61803 39887 49894       1,61803 39887 49895

With 39 internal digits a correct 16-digit result should be no problem. So what's going on here?

Dieter

[Marcus von Cube, Germany]

If I'm not mistaken, SLVQ is not implemented natively but in user code (XROM). So the normal round-off errors may occur.

[Paul Dale]

Solve is a user code program so much work isn't carried out in high precision (some is still). It also searches at display accuracy (i.e. using the approximately equal test).

Thus, some lower order digits errors are to be expected.

- Pauli