We concluded the latest live session yesterday. Unfortunately, I forgot to start recording until we were already 10-15 minutes into the session, so the initial part is missing. But nothing significant was lost. The recording starts just at the right time when were doing an overview of the CPU microarchitecture.
We covered how things work at the hardware and kernel level in the context of the following optimizations used in 1BRC:
Read vs mmap performance
Static vs dynamic allocation of work in a multithreaded design
Thrashing of the L1 cache when dealing with shared data between threads
Design of a cache friendly hash table
Due to lack of time, we could not cover discussion of branch prediction related optimizations, but we did talk about them in the previous session, catch it at the following link
PS: This was the first time I was using my iPad to draw things on the Zoom whiteboard and for some reason it kept freezing while drawing. I will try to find better alternatives to do this for the future sessions.
