How to make good use of the parallel capabilities of GPUs, how to allocate and schedule various GPU memory resources, how to migrate various CPU algorithms to GPUs for acceleration... All these issues are inseparable from an understanding of GPU architecture. Given the continuously evolving nature of GPU architectures, focusing on just one generation of architecture is like capturing only a single frame from a masterpiece, making it difficult to fully grasp its essence and mysteries.
Physically-Based Rendering! Implementation of Disney's BRDF principles. My level is quite average, and I spent a considerable amount of time on understanding and coding the BRDF. However, it's time to summarize. I have previously implemented ray tracing on the GPU using OpenGL multiple times, achieving impressive (not really) results. Unfortunately, our code only supported diffuse reflection materials. To simulate more materials, we have to continue learning.