Hi,
When I was studying the PBD system in Content Examples(Niagara_Advance), there are some custom code got my attention, I replicated the code in my own project, here is the bug(?) I found.
The Custom Hlsl block returns an identity matrix, and is connected to the RLocal input of Custom Hlsl 002 block.
my first thought is to remove this useless multiplication(since M * I), like 001 block under. However, I got different result.
I tried to use Mul(Matrix * Matrix) node with multiple input node, I got the same wrong result.
Then I tried to use multiple Mul nodes with two inputs each chained together, I got the correct result this time.
Liked the MatNodeNoLocalR in the picture ( The green arrow is pointed to the wrong output, it should be MatNoLocalR )
I suspect that there is some issue within the VM? I have no idea how System Update Stage works. Can I get some direction about the cpp files about this problem?
大家好，
在研究Content Examples的Niagara_Advance中PBD时候，发现使用custom hlsl或者 mul matrix*matrix节点进行多矩阵连乘时，会出现计算结果错误的情况，我用自己的工程复现了System Update中 初始化NeighbourGrid模块中的这部分代码，如图
350685-
由于这个RLocal是一个单位矩阵，按理001和002的计算结果应该一致，但读取的Attribute中，结果并不相同，
我尝试使用了 自带的Mul节点，进行多个input输入，结果也是错的。
但是，我用多个Mul节点连接，计算结果就正确了。
由于不太明白System Update的更新机制(似乎是用虚拟机进行计算?)，不知道这是不是一个VM的Bug。。想了解下如何解决这个问题，或者有没有大佬可以科普下 Niagara System Update的C++代码在哪个位置，想深入了解下Niagara的更新机制。。 谢谢~