Great use of bpftrace! eBPF-based tracing is really useful for understanding multithreaded performance in the kernel. E.g. if you're running into kernel-level l

Great use of bpftrace! eBPF-based tracing is really useful for understanding multithreaded performance in the kernel. E.g. if you're running into kernel-level lock contention, the preemption time won't show up in the perf report (except for the few moments of optimistic spin locking). I was trying to understand a bottleneck related to concurrent page faulting, and being able to just hook into the relevant event is nice.