Hacker News Reader

(Author here) It's not about Java's ability to do reflection, many languages expose the ability to get stack-traces. It's usually used for logging and exceptions. But I could certainly write a blog post on how stack-walking is implemented for the likes of async-profiler, as I'm currently working on a new stack-walking API for Java (JEP 435).

View on HN

> It's usually used for logging and exceptions.

That's why I was querying it's suitability for profiling. The time to get a lock for a global "safepoint" and then release it must be significant no?

It's not as significant as one might think (the JVM uses it internally for GC and other things). The main problem is the bias it leads to in profiling and that the time-to-safepoint can sporadically be in the tens of milliseconds with is not great.

Reply on news.ycombinator.com