【作业1】flocking behavior

[SIGKILL]

作业1

作业描述

实现了简单的flocking behavior 模拟。本想用quadtree来优化性能的，但是不会写orz。大佬们可以提供点思路吗？

效果展示

https://github.com/SIGUSR97/taichi-flocking/blob/master/data/5000_boids.gif?raw=true(图像大于 4 MB)

代码链接

Code

[TiantianLiu]

Quadtree需要使用Advanced data layout来编写，单单用ti.field()会比较困难，今天(10/12)上课我会讲：）

也欢迎check太极图形课第03节课的课件。

[archibate]

Not sure if this helps:

HW 1/2: Gravity simulation with tree method in O(N log N), 2D/3D

[video] Backgrounds Newton’s gravity law: a = \frac M{r^2} The traditional approach for simulating a N-body system is, for each particle i and j, we evaluate once this formula. This requires O(N^2) complicity. But if two objects are very far from me, then I may consider these two as a single one. That is, in formula: a = \frac{M_1}{{r_1}^2} + \frac{M_2}{{r_2}^2} \approx \frac{M_1 + M_2}{{r_{CoM}}^2} where r_{CoM} = \frac{M_1r_1 + M_2r_2}{M_1 + M_2}. But how far is far? Well, we define…

Using linear octree might accel this up, but it’s currently single-threaded due to the fact that Taichi can’t allocate thread-local array.

[YuPeng]

太极之前版本的example里面有quadtree的例子，你可以参考: here