Date: 2024-09-01-14:09:24
I just finished listening to a podcast by, actually I don’t know who the host is, but the name of it is I believe Performance Engineering with Andrew Hunter it’s a Jane Street podcast November 2023 at some point and just want to take some notes on some of the things that I found interesting.
I think one of the key takeaways for me was that not to use sampling profilers as often or maybe not, I shouldn’t say not to use them, I shouldn’t reach for them as often until you really know in particular what you’re looking at.
One of the tools that Andrew mentioned was to use pprof and you more want to look at like a DAG-like structure of the nodes being executed and in particular he mentioned using this for malloc and I think I can understand maybe some general ideas behind it but I like to know a little bit more concretely why that is.
Some other things I thought that were interesting because they kind of lined up with how I feel is to be in this field you kind of have to viscerally just like or sorry you have to have a visceral reaction to having unperformed code or not fast code and the way I look at that is like this is just mostly very self-driven you have to be fairly interested in this until and it’s not a very interesting thing for other people until you keep looking deeper and you’re in the right rooms where other people are also very interested in this.
I think another thing that was kind of interesting but it was mentioned more as just a passing comment was how much discipline it takes because you’re constantly being you’re constantly looking at code and thinking this could be faster this could be faster but or that this shouldn’t be doing that and realistically that is not where you should be spending your time.
Another thing that was kind of mentioned here which I never really thought about was like potentially not even having construction the CPU at all and just using the using the bus I don’t really know what that means but I’d like to explore a bit more into that.
Thank you.