Do a proof of concept of marching the first steps in the vertex shader #8
Description
The idea is the following: all the steps that can be done conservatively in the vertex shader, are a cost saved many folds in the fragment shader.
Considering this vertex shader marching as a low resolution pass, I suspect that there is a tradeoff between that resolution, how many steps can be done, and how much cost is saved. I also suspect that this tradeoff will likely be optimal somewhere between 16x16 and 64x64 pixels.
If this approach works, there will be non trivial problems to solve, including the correct duplication of the shader.
To know if it's worth it, a first step is simply to test doing such a 2 stage ray marching.