Drawing of conics is absurdly inexact when bigger numbers are involved

Rulatir shared this problem 2 weeks ago

Not a Problem

When drawing a simple conic x*y = N where N is on the order of 10^12, the graph is about 0.5 off from the graph of y=x/N around x = sqrt(N).

It is probably relevant that if I place a point on the conic, it actually appears off the drawing of the conic and exactly on the function graph.

Files: inexact-conics.ggb

The same problem

Comments (3)

ccambre ● 2 weeks ago

Well, you can test the boundaries of math software using big numbers, that's true. But to come to which conclusion?The 'simpleness' of the conicness is not the issue.

URL

Rulatir ● 2 weeks ago

Please read bug reports carefully to the end:

"It is probably relevant that if I place a point on the conic, it actually appears off the drawing of the conic and exactly on the function graph."

This proves that GeoGebra is actually capable of computing exact points on the conic for numbers of this size, and we are still well within the 15-digit accuracy of the floating point type used. It is probably some over-eager optimization in the algorithm that does the drawing of the conic that is responsible for the inaccurate graph

URL