Francesco Nigro @forked_franz Twitter profile

Last Seen Profiles

@Ryanscott4200

@Custombeardco

@PekkaRinne35

@Way2BlessedMom

@Romantikteyze

@PatrickCannon22

@Crystal62287697

@smith45dora

@rhikww

@Profdilipmandal

@Intra4Good

@saikrishna12377

@DOC2001SL

@JeffPorter01778

@rebecca_iiv

@NoahBrooks99

@AgyeiSa66387658

@avesL3monad3

@Haza_books

@McClain4Freedom

@LaceyEllis65998

@NatanaelSo89424

@dwins2023

@AJG82

@ThePar3ntalUnit

@FizzPop87916406

@tentcityhall

@EpichitoMet

@AprilSmith28033

@CdPlatformu07

@LastBearStandng

@westendholland

@mari_tyan_r18

@joseper70284545

@ajabri

@EvaBowie59607

Francesco Nigro

@forked_franz

11 months

cloudflare always got good stuff!! #iouring

0

31

114

Francesco Nigro

@forked_franz

4 months

ok - WOW

3

26

114

Francesco Nigro

@forked_franz

6 months

Many thanks for this 😭 When I've started years ago my journey in the #java #performance land, I found @mjpt777 @PeterLawrey and @nitsanw blog posts, conferences and code...discovering they were all Java Champions (1/n)

Java Champions

@Java_Champions

6 months

Congratulations to new #Java Champion @forked_franz

54

44

268

9

6

103

Francesco Nigro

@forked_franz

1 year

#java #performance Blog Post Time! Say hi to , first post on @QuarkusIO @netty_project and moar :P

Journey to a N-fold improvement due to a stealthy scalability issue

Learn about fixing a severe JIT scalability issue

redhatperf.github.io

2

38

82

Francesco Nigro

@forked_franz

5 months

For JVM startup profiling I strongly suggest checking and for startup allocation profiling (in case performance isn't a concern) check @AndreiPangin nice suggestion at

does legth of bar in memory graph by async profiler means call times or memory size? · Issue #583 ·...

I'm using async profiler to track large memory allocating behavior. AFAIK, memory profiling is implemented by hooking the TLAB allocation event. However, I'm not sure whether the call stack...

github.com

Francesco Nigro

@forked_franz

5 months

Think about it: profiling sub-second startup of programs isn't very sampling-friendly and requires some mental shift vs runtime profiling. Love profiling won't negate understanding how useful is low-cost (TBD) tracing

0

1

7

1

21

76

Francesco Nigro

@forked_franz

4 months

@seinecle @gunnarmorling from @richardstartin from @pshufb And for mmap, just look at

Linux Page Cache for SRE

SRE deep dive into Linux Page Cache # In this series of articles, I would like to talk about Linux Page Cache. I believe that the following knowledge of the theory and tools is essential and crucial...

biriukov.dev

3

15

69

Francesco Nigro

@forked_franz

6 years

@jugmilano @venkat_s Don't have enough space on my body, but that really worth a tatoo :)

1

25

63

Francesco Nigro

@forked_franz

3 years

Thanks @ScyllaDB for

What We’ve Learned after 6 Years of IO Scheduling - ScyllaDB

Learn the design decisions and considerations we've made over the past six years to improve IO scheduling, then sign up for #P99CONF.

www.scylladb.com

0

13

56

Francesco Nigro

@forked_franz

1 year

Unpopular opinion (can still change my mind): we should reject PRs performing thousands of stupid/obvious changes if not coming by the repo core committers. Stop using (the wrong way) OSS as a LinkedIn score multiplier

5

3

51

Francesco Nigro

@forked_franz

11 months

@vertx_project t the nerverending efforts to optimize the stack has paid off on @TFBenchmarks ! See

1

19

51

Francesco Nigro

@forked_franz

2 years

I'm super happy about my new role as a performance (SW) engineer and so happy to have a crazy amazing team to learn from (1/2)

13

0

43

Francesco Nigro

@forked_franz

2 years

After so many years I found @mjpt777 to be one of the best engineering inspirational talk ever

Engineering You

Martin Thompson talks about the characteristics of a good software engineer and explores the individual practices and techniques that can help bring out the engineer in everybody.

www.infoq.com

2

7

44

Francesco Nigro

@forked_franz

2 years

@marcan42 If you live in a cold place and wanna save some money by letting ThinkPad to heat it I believe you are missing an opportunity really

1

2

42

Francesco Nigro

@forked_franz

1 year

#DevoxxUK talk from me and @SanneGrinovero is out! Check it out

Cracking the scalability wall by Francesco Nigro & Sanne Grinovero

For more info on the next Devoxx UK event 👉 www.devoxx.co.ukSmall things matter: sometimes the smallest detail can have an impact far beyond your imaginatio...

www.youtube.com

2

17

41

Francesco Nigro

@forked_franz

4 months

FYi I've looked with @royvanrijn the awesome article from @richardstartin at to see how to improve the hashcode using 31, and...with Graal JIT no need - it's just autovectorized :O, although (1/2)

Vectorised Polynomial Hash Codes

richardstartin.github.io

Francesco Nigro

@forked_franz

4 months

@thomaswue @graalvm I am very impressed by autovectorization of Graal JIT

0

17

1

5

39

Francesco Nigro

@forked_franz

6 months

And I think that the http Vertx/Netty/improvements should be rightly give to @julienviet and @richardstartin (check ) - small things placed in the right place can make HUGE difference:)

Finding Bytes in Arrays

richardstartin.github.io

tivrfoa

@tivrfoa

6 months

@helidon_project Excellent results! @vertx_project is doing great too! #Java

1

7

18

1

8

38

Francesco Nigro

@forked_franz

4 years

Just read the nice article of @richardstartin and - wanna search again for gold on Hacker's Delight - really wanna improve the search API of #netty

Finding Bytes in Arrays

richardstartin.github.io

1

6

36

Francesco Nigro

@forked_franz

3 months

This is a huge one! Would be nice to see if the huge unmap pause at #1BRC could be spotted with it, in JVM mode

Release Binary launcher and AsyncGetCallTrace replacement · async-profiler/async-profiler

v3.0 Features #724: Binary launcher asprof #751: Profile non-Java processes #795: AsyncGetCallTrace replacement #719: Classify execution samples into categories in JFR converter #855: ctimer mode ...

github.com

0

5

35

Francesco Nigro

@forked_franz

1 year

@PR0GRAMMERHUM0R Performance is never absolute, but relative.

1

0

31

Francesco Nigro

@forked_franz

4 months

Sometime folks don't realize what is the big deal with @QuarkusIO bytecode generation to avoid reflective calls: Reflection has been optimized for years, but prior to JFK 18, it heavily relies on inflation, to perform decently. (1/2)

Java: what is JITC's reflection inflation?

I recently came across this interesting term and searched on Net to know more about it. However the info I found is sketchy. Could someone pl. give me a somewhat detailed explanation of what this i...

stackoverflow.com

2

5

34

Francesco Nigro

@forked_franz

7 months

This is what I love about working closely with @OpenJDK @RedHatJava team; you find an issue and/or performance puzzle and they come out with a super cool diagnostic tool to be used to simplify its investigation next time

JDK-8317683: Add JIT memory statistics by tstuefe · Pull Request #16076 · openjdk/jdk

This proposal adds a way to monitor C1/C2 memory usage per compilation. We are flying a bit blind there: NMT is okay for initial triaging but no help when digging deeper into compiler footprint. NM...

github.com

1

7

32

Francesco Nigro

@forked_franz

1 year

Recently found how many devs seems to not care about profiling as a common practice while valuating A vs B. Not mentioning how flamegraphs seems just unknown to many :"( Is it your same experience?

10

1

32

Francesco Nigro

@forked_franz

3 years

Yesterday fallen into for #aeron @mjpt777 is really well done, complete and readable! nice one!

Aeron Cookbook

Guide to Aeron Transport, Aeron Archive and Aeron Cluster with samples and usage guidance

aeroncookbook.com

2

10

33

Francesco Nigro

@forked_franz

1 year

This is a wonderful and awesome tool from my mate @clebertsuconic : Is super useful and effective, so, check it out!!

GitHub - check-leak/check-leak: Find and Prevent memory leaks in your applications

Find and Prevent memory leaks in your applications - check-leak/check-leak

github.com

0

15

32

Francesco Nigro

@forked_franz

5 months

@felixge fyi Well done @AndreiPangin that's a great addition!!

Accurate CPU sampling without perf_events · Issue #855 · async-profiler/async-profiler

Background Async-profiler has two options for CPU profiling: -e cpu and -e itimer. -e cpu (default) relies on perf_events, while -e timer is based on setitimer. Perf events typically yield more acc...

github.com

2

15

33

Francesco Nigro

@forked_franz

3 years

For who believe #loom is going to save the day 2 devs: async libs often expose backpressure, but what happen if the "natural" concurrency level of blocking code become unbounded? Little law say...latency skyrocket. Backpressure MUST be handled somewhere and OOM isn't the answer

4

8

30

Francesco Nigro

@forked_franz

4 months

Ok wow 😍 @alibaba_cloud engineering! I feel dumb I've never wrote such a simple test to verify it, cool!

Regarding Java 8, the accuracy issue of AllocTracer in --total mode · Issue #874 · async-profiler...

hi Andrei The same reproducer as #873。 In this code, the number of calls to testSmallAlloc and testLargeAlloc is different, but their total memory allocations in bytes are close, which can also be ...

github.com

1

6

30

Francesco Nigro

@forked_franz

2 years

When I think to both know about Linux networking layers and #netty then remind me that I am not yet there uff

How Is Netty Used to Write a High-Performance Distributed Service Framework?

This article gives an overview of Netty and how it works with the distributed service framework.

www.alibabacloud.com

1

10

30

Francesco Nigro

@forked_franz

1 year

Awesome talk on #performance queuing theory

LISA17 - Queueing Theory in Practice: Performance Modeling for the...

Eben Freeman, Honeycomb.io@_emfree_ Cloud! Autoscaling! Kubernetes! Etc! In theory, it's easier than ever to scale a service based on variable demand. In pra...

www.youtube.com

0

2

29

Francesco Nigro

@forked_franz

3 years

I know a great perf eng (nudge nudge @PeterVeentjer ) that years ago during an interview told me: "If you have an app that relies on I/O, looks there to find the low hanging fruits to get the most with less changes" And in the years after it turns to be a wise advice indeed :)

Gunnar Morling 🌍

@gunnarmorling

3 years

"Any improvements made anywhere besides the bottleneck are an illusion". Lots of great advice on #Java performance optimization in @holly_cummins ' session at @AccentoDev !

4

23

128

0

6

28

Francesco Nigro

@forked_franz

7 months

I have strangely missed this nice article from @bellsoftware around #java #jfr #profiling Which help to understand the limits of the approach used by jfr for "CPU" profiling (better "method" profiling)

Hunting down code hotspots with JDK Flight Recorder | BellSoft Java

Hunting down code hotspots is probably the most common task for Java profilers.

bell-sw.com

0

8

28

Francesco Nigro

@forked_franz

3 years

Interesting changes on #asyncprofiler to make possible profile in containers

Profile low-privileged processes with perf_events by Jongy · Pull Request #411 · async-profiler/a...

Implements what's been discussed in #397 . General flow The start operation now accepts a flag fdtransfer, which initializes the FdTransfer class (creates a UDS and waits for a connection). Co...

github.com

1

9

27

Francesco Nigro

@forked_franz

1 year

Just added a short guide on how to troubleshoot at it contains a JMH bench and a reproducer too, to have fun :)

Journey to a N-fold improvement due to a stealthy scalability issue

Learn about fixing a severe JIT scalability issue

redhatperf.github.io

0

16

28

Francesco Nigro

@forked_franz

6 months

Getting recognised in this has been a milestone to me, proving that it is working and the community is receiving something back, and that I should keep on pushing. Thanks everyone!!! (n/n)

1

27

Francesco Nigro

@forked_franz

4 years

Thanks @RedHat 4 #rechargeday I have enjoyed the life with my family and especially my dad

0

1

26

Francesco Nigro

@forked_franz

9 months

Given that I'm on vacation I can reiterate this msg: watch #devoxxuk Especially if you are maintainers of products/libs which care about scalability. I've recently get notification of another known product affected by it (cannot disclose it yet, sorry)

Cracking the scalability wall by Francesco Nigro & Sanne Grinovero

For more info on the next Devoxx UK event 👉 www.devoxx.co.ukSmall things matter: sometimes the smallest detail can have an impact far beyond your imaginatio...

www.youtube.com

1

7

24

Francesco Nigro

@forked_franz

3 years

@mariofusco So guitar pedals are based on 5G alien technology

2

0

24

Francesco Nigro

@forked_franz

10 months

Happy that the work done on Undertow was referenced at in the register pressure paragraph!

Decreasing the Number of Memory Accesses: The Compiler's Secret Life 2/2 - Johnny's Software Lab

We investigate memory loads and stores that the compiler inserts for us without our knowledge: "the compiler's secret life". We show that these loads and stores, although necessary for the compiler...

johnnysswlab.com

1

2

24

Francesco Nigro

@forked_franz

1 year

Not many knows...after reading from @helidon_project I've decided to dig (& fix) @netty_project http 1.1 perf, hope now a new version of the same blog post will be posted, maybe comparing with @vertx_project (look at ) (1/2)

TechEmpower Framework Benchmarks

Performance comparison of web application frameworks using community-contributed test implementations.

www.techempower.com

2

0

24

Francesco Nigro

@forked_franz

8 months

Highly recommended! It contains some interesting updates and new stuff related this infamous scalability #perf issue!

Rafael Winterhalter

@rafaelcodes

8 months

Interesting (and very digestible) talk about an inner working of JVM performance.

1

33

122

0

5

25

Francesco Nigro

@forked_franz

7 months

+100000

Arnaldo Carvalho de Melo

@acmel

7 months

Willy asks for help in removing all linked lists from the kernel.

4

15

67

3

0

25

Francesco Nigro

@forked_franz

1 year

Ok I can say it. I love my job 💟

0

24

Francesco Nigro

@forked_franz

2 years

I am genuinely curious to see the new numbers of #netty Vs #helidon here Especially with the yet to be merged change addressing checkcast scalability:)

Helidon Níma — Helidon on Virtual Threads

Helidon 4.0.0-ALPHA1 is now released with our brand new Helidon Níma, providing a virtual threads-based web server. This is an early…

medium.com

2

4

24

Francesco Nigro

@forked_franz

5 months

I cannot really thanks enough @AndreiPangin for #asyncprofiler candies he put into it, for example

Show targets of vtable/itable calls · Issue #736 · async-profiler/async-profiler

In some applications, a signiticant amount of CPU time is spent on dispatching megamorphic virtual/interface calls. This appears in the profile as vtable stub or itable stub. However, this does not...

github.com

2

8

23

Francesco Nigro

@forked_franz

6 years

@jugmilano

1

12

23

Francesco Nigro

@forked_franz

5 years

After many years unable to have proper holidays for several reasons, I will go in honeymoon for 1 month in Japan...I really don't know yet how to handle this....without my laptop with me!!!! 😮

3

0

21

Francesco Nigro

@forked_franz

1 year

My dad (RIP beloved father) used to say: "cry if something hurt you, be happy if things works great..but don't let any of these feelings to last more then 5 minutes and determine your future"

0

22

Francesco Nigro

@forked_franz

4 months

@penberg If you need (when you already have a basic understanding of how things work) to catch up with the architectural part, this blog post is great (random article, you can find many in, including for CPU archs)

AMD’s CDNA 3 Compute Architecture

AMD has a long history of vying for GPU compute market share. Ever since Nvidia got first dibs with their Tesla architecture, AMD has been playing catch up. Terascale 3 moved from VLIW5 to VLIW4 to…

chipsandcheese.com

0

1

22

Francesco Nigro

@forked_franz

9 months

That's why no microbench can skip perfasm or assembly inspection, to draw the right conclusions!

Optimize method HttpStatusClass.valueOf(int) by laosijikaichele · Pull Request #13543 · netty/netty

Motivation: The method HttpStatusClass.valueOf(int) can be optimized by replacing 'if branches' with 'switch cases'. The 'switch cases' can be optimized by JIT to run faster...

github.com

2

4

22

Francesco Nigro

@forked_franz

3 months

Wow @ClickHouseDB JIT at

0

8

21

Francesco Nigro

@forked_franz

3 months

This is the second scary part of the bug explained at and !!

Cracking the scalability wall by Francesco Nigro & Sanne Grinovero

For more info on the next Devoxx UK event 👉 www.devoxx.co.ukSmall things matter: sometimes the smallest detail can have an impact far beyond your imaginatio...

www.youtube.com

Luca Molteni

@volothamp

3 months

I spoke about the performance of Drools and Optaplanner at the latest @VoxxedTicino , addressing a challenging JVM bug Take a look at the video and let me know your thoughts! 😀

3

14

31

0

6

22

Francesco Nigro

@forked_franz

5 years

Any feedback about using Java streams and JIT friendlyness? Seems that most Stream methods are just too big to be inlined :( Even @graalvm seems unable to avoid weird intermediate allocations due to the size of some stream classes...

3

16

21

Francesco Nigro

@forked_franz

2 years

Very nice wrap-up of how #netty works under the hood for NIO It would be nice to see something like that for #io_uring @josefgrieb to show how much it simplifies things compared to NIO :P

A Tour of Netty

Introduction

medium.com

1

5

21

Francesco Nigro

@forked_franz

3 years

#concurrent programming is hard. #distributed algorithms, maybe more. But what about concurrent bugs on distributed algorithm libraries? Please people, if something is that hard, just make it single threaded.

1

3

22

Francesco Nigro

@forked_franz

3 years

What people often forget about event driven programming based on some state machine is that if events come in some unexpected order some actions has to be taken. Knowledge is a power and from great powers...

6

9

20

Francesco Nigro

@forked_franz

1 year

Classify option for #asyncprofiler nice one :)

Automatically classify execution samples into categories · Issue #719 · async-profiler/async-prof...

When analyzing a flame graph, one often asks questions like How many time in total was spent on class loading? What percentage of time Java code was running in the interpreter and in C1/C2 compiled...

github.com

0

6

20

Francesco Nigro

@forked_franz

5 years

#jctools #concurrent #ComputerScience Awesome discovery on scalability while benchmarking a novel concurrent queue algorithm

XADD Chunked Array Queue by franz1981 · Pull Request #230 · JCTools/JCTools

It is a chunked array q designed to scale with the number of producers.

github.com

1

6

20

Francesco Nigro

@forked_franz

6 months

Thanks to their efforts to educate others, I've fallen in love with this discipline, getting a giant first step into it. I've promised myself I would have done the same for every bit of knowledge gained, never retaining anything for me, to give others the same chance I had (2/n)

1

2

21

Francesco Nigro

@forked_franz

2 years

It's so easy to get impressed by the super nice articles of folks like @shipilev , @richardstartin , @nitsanw and others... The more I read them the more they will awake my thirst of knowledge Just thanks to have taken so much time and effort to write them!!

0

3

19

Francesco Nigro

@forked_franz

3 years

Why Java ReentrantLock isn't exposing an API to limit the size of the wait queue while acquiring the lock? Back-pressure everything, my new motto

2

21

Francesco Nigro

@forked_franz

4 months

@gunnarmorling @PeterLawrey in @netty I've used it to implement a generic indexOf(byte) which was key to improve \r\n http parsing, can be used with 8 bytes strides or vs the mmapped ByteBuffer

Finding Bytes in Arrays

richardstartin.github.io

1

4

19

Francesco Nigro

@forked_franz

1 year

@brendangregg I think the Netflix article should mentioned few other work on it (already mentioned elsewhere, in other twwets), but as well that explain (thanks @AndreiPangin ) why AGCT wasn't able to detect it on flamegraphs, while perf yep

Unknown Java Frame on JDK-8180450 reproducer · Issue #673 · async-profiler/async-profiler

MInimal (benchmark) code to reproduce it using # JMH version: 1.35 # VM version: JDK 11.0.16.1, OpenJDK 64-Bit Server VM, 11.0.16.1+1 import org.openjdk.jmh.annotations.*; import org.openjdk.jmh.in...

github.com

0

2

19

Francesco Nigro

@forked_franz

4 months

A christmas present for the "Advent of InstanceOf" :P

Http 2 type check slow path detected · Issue #13745 · netty/netty

This is the MISS report from a load test running against a HTTP 2 server written in Netty (basically an hello world-like case, really): -------------------------- Miss: -------------------------- 1...

github.com

2

4

20

Francesco Nigro

@forked_franz

2 months

When you speedup a core method by 30% just by saving an array ref a local variable...ok, I don't feel cool 😔

2

0

19

Francesco Nigro

@forked_franz

3 months

Many thanks @richardstartin and @BachorikJ for this nice presentation around #profiling tools at @Jfokus !!! And thanks again to @AndreiPangin to be the enabler of some of the coolest things happening in the @java profiling land in the last years!

Jean-Philippe Bempel @[email protected]

@jpbempel

3 months

Future-proofing JVM profiling by @BachorikJ and @richardstartin at @Jfokus @datadoghq

2

29

74

0

2

19

Francesco Nigro

@forked_franz

3 years

Nice work of @chvest on #netty 5 yes yes and yes

0

6

18

Francesco Nigro

@forked_franz

4 months

#1BRC i would be interested if anyone has profiled how much CPU cycles are consumed by C2 compilation in their solutions..:P

5

3

19

Francesco Nigro

@forked_franz

9 months

Just wow for @qconlondon by @halvarflake

Adventures in Performance: Efficiency Analysis of Large-scale Compute

Thomas Dullien discusses how language design choices impact performance, how Google's monorepo culture and Amazon's two-pizza-team culture impact code efficiency, and why statistical variance is an...

www.infoq.com

0

5

19

Francesco Nigro

@forked_franz

2 years

Today helped a friend at @RedHatJava to solve a perf issue around ReentrantReadWriteLock and...gosh - being reentrant means *always* keeping a per-thread state. Do you know what happen on release? You gotta cleanup the thread local state to save mem leaks of Thread...:O

3

2

18

Francesco Nigro

@forked_franz

2 years

inlining/interpreted/compiled/native frames are landing on @AndreiPangin #asyncprofiler Bye bye perf map...we've been good friends for many years...

0

3

17

Francesco Nigro

@forked_franz

9 months

@debasishg That's a very nice one indeed! Mine is Dijksra one

2

4

18

Francesco Nigro

@forked_franz

6 months

@tivrfoa @helidon_project @vertx_project Yep! A summary of our efforts are at It's important to remember that Quarkus results are making uses a full-fat ORM (in its latestish version), which, for CPU bound tests like this with simple queries, has impacts

Framework Benchmarks Round 22

Now in its tenth year, this project measures the high-water mark performance of server side web application frameworks and platforms using community-contributed test implementations.

www.techempower.com

1

8

18

Francesco Nigro

@forked_franz

4 months

@thomaswue @graalvm I am very impressed by autovectorization of Graal JIT

0

17

Francesco Nigro

@forked_franz

5 months

This 👇 thanks for this @TheDonRaab !!

Donald Raab

@TheDonRaab

5 months

I finally blogged about "My Journey to and as a Java Champion." Looking back five years and recalling things I had done before July 2018 and some of the things I have done since has been fun. Enjoy! 🙏

0

30

162

1

2

17

Francesco Nigro

@forked_franz

11 months

Loved the #performance #book recommendations from @medawsonjr at

My Top 7 Performance Books for Engineers

The most frequent request I get from engineers looking to branch into this niche is for performance book recommendations. These are my Top 7.

www.jabperf.com

0

4

18

Francesco Nigro

@forked_franz

2 years

Ehi, that's me!!!

Chris Vest

@chvest

2 years

Franz is on another level with his latest performance investigations:

1

7

40

0

3

17

Francesco Nigro

@forked_franz

5 months

Having 2 days national/regional holiday and bring my 2yo child to watch the very first time the "city" is one of the most awesome experience I ever had ❤️

0

18

Francesco Nigro

@forked_franz

2 years

#asyncprofiler finally get in! Yeee The official, lightweight and unbiased (on large objects) allocation profiling (sampling)

0

5

18

Francesco Nigro

@forked_franz

2 years

The Type Pollution Agent has moved to under the Red Hat Perf Team umbrella ;) Old repo will be read-only soon

GitHub - RedHatPerf/type-pollution-agent

Contribute to RedHatPerf/type-pollution-agent development by creating an account on GitHub.

github.com

0

9

17

Francesco Nigro

@forked_franz

3 months

Finally this week placed some "Focus time" on the calendar to keep on writing "real" code, and I'm super excited and scared at the same time 🤣 Writing performance fixes is not really the same as coding something new!

1

0

17

Francesco Nigro

@forked_franz

2 years

Thanks to this article I wanted to refresh my knowledge on the Linux CFS scheduler and found this nice article too

To sched_yield() Or Not To sched_yield()?

Is Linux' sched_yield() useful while polling, or merely "sound and fury signifying nothing"?

cpufun.substack.com

Gunnar Morling 🌍

@gunnarmorling

2 years

📢 Blogged: " #Loom and Thread Fairness" Taking @java 19's Project Loom for a spin, I learned about an interesting aspect to the scheduling of CPU-bound workloads on virtual threads.

18

60

277

0

5

17

Francesco Nigro

@forked_franz

3 years

It's very important to NOT be a linked list developer

Yogini Bende

@hey_yogini

3 years

What is a stack developer? 👀

213

41

808

5

0

17

Francesco Nigro

@forked_franz

2 years

I'm often surprised how horrible code can be incredibly optimized by JIT - I would rather prefer it to break and be slow instead!

2

0

17

Francesco Nigro

@forked_franz

1 year

And now @mariofusco is rocking on the stage of @DevoxxGR ! 🎸 🎸 🎸

Mario Fusco 🇪🇺🇺🇦

@mariofusco

1 year

If you are at @DevoxxGR and eager to learn when and how to profile your Java code come to @forked_franz and mine deep dive tomorrow moring Even better if you want to follow our workshop and code with us clone our project here:

1

2

7

0

1

17

Francesco Nigro

@forked_franz

5 years

Pushed a PR to the awesome #jctools for a POC of a XADD multi-producer wait-free queue...fingers crossed!!!!! p.s. The logic is totally mine, but I'm sure a paper with it must exists...it is too simple!!!!Multo-consumer should be easy too

1

2

16

Francesco Nigro

@forked_franz

5 months

This is a game changer -> #asm4everyone

1

4

15

Francesco Nigro

@forked_franz

4 years

If you expect the JVM to be smart enough to save duplicated bound/range checks over arrays, look here:

Faster decodeHexNibble by franz1981 · Pull Request #9896 · netty/netty

Motivation: decodeHexNibble can be a lot faster using a lookup table Modifications: decodeHexNibble is made faster by using a lookup table Result: decodeHexNibble is faster

github.com

0

6

16

Francesco Nigro

@forked_franz

7 months

Cannot wait to be at the @graalvm community event with @maxandersen , @mariofusco and other @RedHat mates!

2

1

16

Francesco Nigro

@forked_franz

5 months

As usual, the most powerful optimization is to not doing anything. Or reduce useless things to get in the hot path

1

16

Francesco Nigro

@forked_franz

4 years

I'm not a company man but ehi... @RedHat is changing me: love the great humanity of you, thanks 😍

2

0

14

Francesco Nigro

@forked_franz

1 year

🤔 I have waited past 1st April to give some bad news... #2ndapriltruth

Quarkus/Vertx/Netty HTTP performance · Issue #1 · FgForrest/HttpServerEvaluationTest

Netty HTTP performance has changed A LOT recently hence I suggest to: bump up from 4.1.75.Final to latest available and use epoll native (or io uring through the incubator -> https://github.com/...

github.com

0

4

16

Francesco Nigro

@forked_franz

1 year

Very proud to see our #quarkusinsights to make folks interested into how @QuarkusIO works and perform under the hood unveiling some OS interesting behaviours too eg

Techempower benchmark spends 7.7% of time in LockSupport.unpark()? · quarkusio quarkus · Discussion...

Hi, out of curiosity, I profiled the Quarkus Techempower benchmark with JProfiler 11.1.4. The biggest "hotspot" that it reports is in LockSupport.unpark(), where allegedly 7.7% of all CPU...

github.com

1

16

Francesco Nigro

@forked_franz

4 months

It's nice to see how #1BRC polarize and activate different languages fans; it is indeed a good example to see how much dev effort is required to hit the same memory wall, if any

1

2

15

Francesco Nigro

@forked_franz

1 year

shit moment: when u spend N months improving a LOT something vs a chosen competitor and found (right at the end) it was cheating. :"( And I was complaining to have improved smth just ~30-40% @normanmaurer I need a friendly song here

7

0

15

Francesco Nigro

@forked_franz

4 months

I've spoken with few @Java_Champions friends recently which didn't knew about @DevoxxUK nor @javazone from me and @SanneGrinovero : if u're a @OpenJDK or @openj9 user is highly recommended!!

JavaZone 2023 presentation: Cracking the scalability wall - by Sanne Grinovero

In this talk, we’ll discuss a severe JVM performance issue, the methodology leading to its discovery, how it affects the whole Java ecosystem, and what you could do to avoid it in your code. This...

2023.javazone.no

2

3

15

Francesco Nigro

@forked_franz

4 years

@mariofusco @gunnarmorling Sure; I've loved this talk Just not sure if is for not I/O nerds, but @axboe has done a great job to make it digestible IMO :) If you need the TLDR one these are few around too:

0

6

14

Francesco Nigro

@forked_franz

2 years

After many years possessing intel CPUs I'll switch to an AMD Ryzen 5950X! I've never read AMD perf/dev tech docs and I'm a bit scary re Linux perf tooling current state (to not mention that toplev won't work with it afaik), any #perf friend experience?

5

3

15

Francesco Nigro

@forked_franz

3 years

I'm a man of honor :) I'm still investigating and will prepare some CI job to run of real decent HW too, but has been quite interesting... #io_uring is impressive as expected and @josefgrieb and @normanmaurer have done an awesome job in #netty for it :)

ARTEMIS-3163 Experimental support for Netty IO_URING incubator by franz1981 · Pull Request #3479 ·...

https://issues.apache.org/jira/browse/ARTEMIS-3163

github.com

Matt Fleming

@fleming_matt

3 years

@forked_franz @ApacheActiveMQ Blog post/tweets or it didn’t happen 😉

1

0

1

2

5

14

Francesco Nigro

@forked_franz

4 years

FYI by @DOOM369 alternative link

Micro optimizations in Java. Good, nice and slow Enum

Enums are crucial part of every application larger than “Hello World“. We use them everywhere. They are actually very useful: they…

link.medium.com

2

4

15

Francesco Nigro

@forked_franz

8 months

This year at @DevoxxBE me and @mariofusco will entertain you with the fabulous world of uncertainty and mistery (and misery?) ie JVM profiling of @QuarkusIO apps: If you find any #perf issue worthy of opening an issue, win t-shirt! (1/2)

4

3

15