Johannes Bechberger
@mostlynerdless.de
630 followers 96 following 140 posts
OpenJDK hacker at SapMachine, works on hello-ebpf for fun
Posts Media Videos Starter Packs
mostlynerdless.de
Oh, I'll take a look this afternoon/tomorrow afternoon
Reposted by Johannes Bechberger
scylladb.com
💪 Monday motivation: We are officially one month away from #P99CONF. eBPF has always been a popular topic, and this year is no exception. Don't miss @tanelpoder.com, @ddelnano.bsky.social, Jake Hillion, and @mostlynerdless.de on October 22-23: www.p99conf.io?latest_sfdc_...

#ScyllaDB
mostlynerdless.de
I had a great recording session yesterday with Jake, we hope you like our chaotic talk
mostlynerdless.de
Turns out this is not possible.
mostlynerdless.de
The new CPU-time profiler is truly powerful. You can even write a configurable wall-clock profiler with it on Linux (guess what I'll show in a follow-up blog post).
mostlynerdless.de
The rationale was: Almost nobody uses a different platform in production (and test environments) anyway, so it's not worth the hassle to spend years implementing it on other platforms.
mostlynerdless.de
With Java 25, there is finally a CPU-time profiler in the OpenJDK.

You can read more about this new profiler and some implementation details in my blog series starting with mostlynerdless.de/blog/2025/06... #java25
Java 25's new CPU-Time Profiler (1) - Mostly nerdless
Learn all about Java 25's new CPU-time profiler and why it matters in this weeks blog post from the creator himself.
mostlynerdless.de
mostlynerdless.de
Learn about the redundant synchronisation in the request queue of Java 25's new CPU time sampler and what we do about it in this week's blog post: mostlynerdless.de/blog/2025/09...

P.S.: It also answers the really important question of whether the sampler queue is even a queue.
Java 25’s new CPU-Time Profiler: Removing Redundant Synchronization (4) - Mostly nerdless
Learn about the redundant synchronisation in the request queue of Java 25's new CPU time sampler and what we do about it in this blog post.
mostlynerdless.de
mostlynerdless.de
I got your mail, just didn't get around to replying. I'll take a look next week.

But in general, I would say that it's probably fast enough. There are other areas in the JVM that have probably more potential for optimization (cc @shipilev.bsky.social)
mostlynerdless.de
One of my colleagues thought it was giving people too many reasons not to switch to JDK 25 :D
mostlynerdless.de
Sneak peek into my upcoming blog post:
Plot of sample loss and memory consumption vs queue size in the new CPU-time sampler of JDK 25
mostlynerdless.de
Thanks for the reminder :)
mostlynerdless.de
Neither did I. Hope so see someplace else.
Reposted by Johannes Bechberger
scylladb.com
To get you ready for @mostlynerdless.de and @jakehillion.me’s #P99CONF technical session on the Linux scheduler, we're sharing a deep dive from Johannes' hello-ebpf series that looks at concurrency testing using Custom Linux schedulers. www.p99conf.io/2025/08/06/h...

#ScyllaDB
Hello eBPF: Concurrency Testing using Custom Linux Schedulers
How anyone can write a basic Linux scheduler and use it, for example, to fuzz for concurrency bugs or optimize for specific workloads.
www.p99conf.io
mostlynerdless.de
So more time for other pastimes this autumn.
mostlynerdless.de
Seems like I won't be part of the great Devoxx Belgium conference this year.
mostlynerdless.de
TIL: the k3s uninstall script changes the hostname