@Hoshinova I would say it is ok to commit in this state. We can change it if the issue arise. Also we are going to raise the minimum requirement for 4.0 so it might just work. 🤞
#98190 was the original bug that made me worried that this kind of optimization scheme was not viable. Even with const keyword are not enough for these compilers. But maybe that was because of…
This looks good to me. I think the approach is good as the tracking distortion doesn't usually change when previewing animation which is the primary goal.
Not sure I get it. Why does it need to be cleared if the material write to it?
In order to avoid this index computation use uint4 on both C++ and GLSL side. The memory waste is not really important.
When you want to limit something to C++ implementation, use #ifndef GPU_SHADER. See :
This is not needed. A padding member will never be anything else than int or float. int4 is already 16 bytes.