Development blog where I share my knowledge and projects
Tuning the renderer for up to 100x speedup
Optimizing memory access by combining the kernels into a single one
Move more rendering logic to the gpu
In previous posts we made the decision to copy all rays to the device even if they are done, But if a large proportion of rays is done, then we are wasting time copying them to the device.
I’ve been working on a highmap voxel renderer that generates the following renders: