I was expecting something about NUMA performance, temporal memory access instructions, shared versus global versus register memory on GPUs, SRAM, and so on. The

I was expecting something about NUMA performance, temporal memory access instructions, shared versus global versus register memory on GPUs, SRAM, and so on. There's an article about all these things waiting to be written. This article is instead about memory-hierarchy-aware access pattern optimization, which is important, but not the whole story.