Topic: AutoLisp collapse of Speed?

[vlisp]

I am working on a task that requires significant computational resources. The first 5 iterations AutoLisp works significantly fast, then slows down significantly. No objects are drawn in AutoCAD while the calculations are being made. Net calculations - rotation and translation of points. Any ideas what this could be?

Thanks!
 

[jtoverka]

The best I can offer you is to look at the computational complexity of your algorithm also known as the Big O.

https://en.wikipedia.org/wiki/Big_O_notation#/media/File:Comparison_computational_complexity.svg

Otherwise, we can't offer any other advice without seeing the code.

[vlisp]

The complexity of the algorithm is not great, because the first iterations calculate them very quickly. Then at one node (vertex) of polygon makes 15 rotations for about 20 minutes! The the rotation polygon there is 5 vertices. 

[dgorsman]

To reiterate one of the above points: without code, or even a program flow/outline, we're just taking wild stabs in the dark.  My jab: you're performing unnecessary repetitive calculations rather than organizing flow for an early exit.

[Lee Mac]

Are you perhaps building lists within the loop which should be cleared with every iteration, but instead grow exponentially?

[vlisp]

Hi Lee.
No. I use while every time and the lists decrease with each iteration. I've already done that. It's not about that. The list of points of polygon are constant and each iteration I am cutting one point.

[jtoverka]

To reiterate one of the above points: without code, or even a program flow/outline, we're just taking wild stabs in the dark.  My jab: you're performing unnecessary repetitive calculations rather than organizing flow for an early exit.

Hi Lee.
No. I use while every time and the lists decrease with each iteration. I've already done that. It's not about that.

We solve code problems, not riddles 

[BIGAL]

Need lots more info like what are you actually doing the task ? Understand needs iteration but there are ways of speeding that up sometimes.

[vlisp]

To reiterate one of the above points: without code, or even a program flow/outline, we're just taking wild stabs in the dark.  My jab: you're performing unnecessary repetitive calculations rather than organizing flow for an early exit.

The code is very long. If the lean part of the code will not be able to get an idea of ​​the problem. Yes there really are repetitive polygons as a result of the calculations. But this is the more economical option than comparing the newly obtained polygon with the already received polygons.

We solve code problems, not riddles 

[ronjonp]

Why don't you post your code? You will most likely get more help on this ... and as a bonus by some of the smartest minds out there.