Love your notes explaining some of the GPU thread-indexing concepts!

Shameless plug for my own little post going a bit into the performance benefits of "vectorized" sorting, even vs. programmable L1 cache: https://winwang.blog/posts/bitonic-sort/