PathFinder status

[NiLSPACE]

I know, but in the bottom left corner you can see how long it took.

[LogicParrot]

I know, but in the bottom left corner you can see how long it took.

Calculating the Jump Points (What you see when "visualize recursion" is ticked) takes time. But in reality we could reuse those Jump Points for multiple path queries and even for multiple mobs. Also, there's "JPS+" which is meant to identify the Jump Points faster. But I don't know how it works and what are its weaknesses yet.

---

Also, I believe it should be much faster than A* in the simulation if you try it inside an enclosed space such as a maze.

[LogicParrot]

https://archive.org/details/GDC2015Rabin

[xoft]

The question is how they measure their preformance - whether by the number of recursions, or how many grid cells are evaluated etc. When there are different costs for these operations, the algorithms may behave differently performance-wise.

I'm not saying we shouldn't do it, I'm just saying we should think carefully about our circumstances.

[LogicParrot]

I totally agree. This is why it's very low on my priority list. Our performance doesn't seem to be terrible right now.

[LogicParrot]

Althought not feature-complete, the PathFinder is pretty decent. Once I deal with water, fences, and doors, I think I'll pause working on the PF and focus on the Mob ToDo list for a while.

[NiLSPACE]

The path finder is integrated in Cuberite, but perhaps Safwat still has the original code left.

[LogicParrot]

Where I can download the Pathfinder and how work it?

1. The PathFinder is part of Cuberite, the mobs use it to navigate. Cuberite version is here: https://github.com/cuberite/cuberite/tre...r/src/Mobs , search for Path.cpp and Path.h

2. A compiler flag makes the code work without Cuberite at all, and makes it use Irrlicht for rendering. But this version is currently broken, I'll fix it and upload it later. You'll have to stick with 1 for now.

[LogicParrot]

Recalc rewrite in progress

[LogicParrot]

Pathfinder recalculation logic was seperated from mob logic in #2668. This made monster.cpp cleaner, but it could use some more refactoring.