Reinventing rendering one shader at a time The other day I ran into a perfect example of exactly why GPU programming is so foreign and weird. In this post I will explain why, because it's a microcosm of the issues that lead me to build Use.GPU, a WebGPU rendering meta-framework. What's particularly fun about this post is that I'm pretty sure some seasoned GPU programmers will consider it pure heresy. Not all though. That's how I know it's good. GLTF model , rendered with Use.GPU GLTF A Big Blob of Code The problem I ran into was pretty standard. I have an image at size WxH, and I need to make a stack of smaller copies, each half the size of the previous (aka MIP maps). This sort of thing is what GPUs were explicitly designed to do, so you'd think it would be straight-forward. If this was on a CPU, then likely you would just make a function downScaleImageBy2 of type Image => Image .…