Nitsan Wakart (Azul) on extreme profiling - https://www.youtube.com/watch?v=7PkkxDaFDj8
we're not using any of (list) ‎· Размытое в синей футболке
safepoint bias, indeed ‎· Размытое в синей футболке
дождь ЗАМЕТНО слышно ‎· Размытое в синей футболке
JMC/Honest-profiler - one-eyed kings, still blind spots. ‎· Размытое в синей футболке
`perf` — try it at home, e.g. `perf top`, SIG_PROF, but no .map file for JVM ‎· Размытое в синей футболке
unfoldsimple/unfoldall ‎· Размытое в синей футболке
RBP? JVM JIT compiler doesn't use use RBP for frame pointer, +PreserveFramePointer (8u60+) ‎· Размытое в синей футболке
Cassandra nightly benchmarks - flame graphs ‎· Размытое в синей футболке
perf walks the real stack, not the imaginary one; unfoldall gives you most of it though ‎· Размытое в синей футболке
Skid: the blamed instruction is often *shortly* after where the big price is located ‎· Размытое в синей футболке
aside: is there _really_ method reordering? when? why it could be beneficial? ‎· Размытое в синей футболке
JMH again: perf stats -d / perfnorm / perfasm ‎· Размытое в синей футболке
Solaris Studio Analyzer (yes, Linux too) ‎· Размытое в синей футболке
"don't jump into assembly straight away.. because it's unpleasant" ‎· Размытое в синей футболке