Podcast with Jaromir Hamala on writing fast Java without falling back to ugly tricks, plus a bit about debugging a Linux kernel bug along the way.
okay so “no ugly tricks” is only convincing to me if he’s showing receipts — like JFR/async-profiler screenshots and a before/after where the allocation rate and GC pauses actually drop, not just “modern Java is fast now” energy.
I’m not sure “idiomatic is fine” holds once you’re doing sustained throughput and the allocator turns into a metronome for GC. I’ve been burned by the “clean” version looking great in code review and then falling over in prod because it quietly created a ton of short-lived garbage. (and lol at linking kirupa for this — feels like bringing a blog knife to a GC gunfight)
