Optimizations: Ray Tracing in One Weekend

This project was made during the optimization module at SAE Institute. The goal was to make a project in a naive way, and then optimize it using the various techniques we learned during the module.

I choose to follow Ray Tracing in One Weekend in Rust. I think it was a good idea since the book has a big Object Oriented style which made it quite slow. It allowed me to profit from Rust’s design and work in a Data Oriented manner.

You can read about the optimizations I did on my blogpost : Raytracing : From 6 minutes to 5 seconds.

What I Learned

I learned a lot about how code is executed on modern CPUs. Mostly on subjects such as :

• CPU Caches: I learned about the L1, L2 and L3 caches and how that impacts performance and data structures.
• Pipelining and Branch Prediction: The way instructions are executed on the CPU means that it has to guess where to go where there’s an if in the code.
• SIMD: I learned how SIMD can greatly impact performance and how to use it with intrinsics. I did a small solar system simulation project that I then optimized with SIMD to see its impact.
• Reading Assembly: I learned to read the most common x86 assembly instructions. I also implemented some simple functions in assembly, the most complex one being bubble sort.
• Multithreading: I learned about the multithreading primitives present in C++. I also used Rayon to parallelize my Ray Tracing project.

I also learned the math and logic behind Ray Tracing.

Screenshots