# Math related sum distance problem with TSP content

I hava an interesting problem related with GIS / MATH / TSP.

If I calculate the distance between A and B on a x/y-style coordinate system it is obvious to use the following formula:

``````sqrt((Ax-Bx)^2+(Ay-By)^2) = Distance
``````

If I compare now 2 distances AB and CD it's obvious to do this with the formula on top:

``````sqrt((Ax-Bx)^2+(Ay-By)^2) = Distance
AND
sqrt((Cx-Dx)^2+(Cy-Dy)^2) = Distance
``````

Now if you need to do a lot of distance calculation, each SQRT() operation eats up valuable CPU time. And as we know 2<3 and 2^2<3^2 you can leave the sqrt() command away - just to enhance performance! SQRT is very expensive!

And now where it starts to get complicated for me. WHY doesn't this work with ABCDEF paths? Here is a classical TSP 3-opt optimization on these points.(example ABCDEF, ABCDFE, ABCFED, etc. permutations...). It's obvious that:

``````sqrt((Ax-Bx)^2+(Ay-By)^2) + sqrt((Cx-Dx)^2+(Cy-Dy)^2)
``````

is not the same

``````(Ax-Bx)^2+(Ay-By)^2 + (Cx-Dx)^2+(Cy-Dy)^2
``````

is obvious, but it should be proportional, or is this a wrong assumption?

I have the following log which shows that my theory is wrong!

``````Distanz without SQRT(): 2423932.0 - Distanz SQRT(): 24131.530843013228
Distanz without SQRT(): 2421100.0 - Distanz SQRT(): 24124.90131901531
Distanz without SQRT(): 2419674.0 - Distanz SQRT(): 24126.12515189262
Distanz without SQRT(): 2414714.0 - Distanz SQRT(): 24117.38148735822
Distanz without SQRT(): 2414304.0 - Distanz SQRT(): 24112.91523442742
Distanz without SQRT(): 2409924.0 - Distanz SQRT(): 24117.550376378276
Distanz without SQRT(): 2403676.0 - Distanz SQRT(): 24114.96478009344
``````

`````` sqrt((Ax-Bx)^2+(Ay-By)^2) + sqrt((Cx-Dx)^2+(Cy-Dy)^2)
`````` (Ax-Bx)^2+(Ay-By)^2 + (Cx-Dx)^2+(Cy-Dy)^2
Let's simplify the problem, let α be `(Ax-Bx)^2+(Ay-By)^2`, and β be `(Cx-Dx)^2+(Cy-Dy)^2` , then you are saying `sqrt(α) + sqrt(β)` is proportional to `α + β`, which of course a wrong assumption.