tag:blogger.com,1999:blog-8405107760807432973.post7626460736810359038..comments2024-01-04T09:55:32.459-08:00Comments on Java Concurrency (&c): Why Many Profilers have Serious Problems (More on Profiling with Signals)Jeremy Mansonhttp://www.blogger.com/profile/04241094734813086257noreply@blogger.comBlogger16125tag:blogger.com,1999:blog-8405107760807432973.post-5258949786867273972013-07-24T14:21:39.972-07:002013-07-24T14:21:39.972-07:00I think it does a good job of summarizing some of ...I think it does a good job of summarizing some of the problems of the existing profilers, though, which is why I reference it.Jeremy Mansonhttps://www.blogger.com/profile/04241094734813086257noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-30896426666216737532013-07-24T12:42:33.573-07:002013-07-24T12:42:33.573-07:00The Mytkowicz and Diwan paper really bothers me. F...The Mytkowicz and Diwan paper really bothers me. For example, it's definition of "hotness" seems to mean "self time percent". The whole reason gprof was invented 30 years ago was that self time is an inadequate diagnostic. Another way it bothers me is that it concentrates on methods, rather than lines of code. Yet another way is its shallow understanding of sampling Mike Dunlaveyhttps://www.blogger.com/profile/16657552696382525525noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-30796549180736517092011-11-28T23:08:58.803-08:002011-11-28T23:08:58.803-08:00I don't currently use Oracle Studio, although ...I don't currently use Oracle Studio, although I have tried it in the past.<br /><br />I don't know of any publicly available tools, other than Intel's, to do hardware level profiling on a per-thread basis.<br /><br />I guess that's not so helpful. Sorry!Jeremy Mansonhttps://www.blogger.com/profile/04241094734813086257noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-26713553296205649312011-11-27T22:51:34.242-08:002011-11-27T22:51:34.242-08:00Do you use Oracle studio analyzer ? I think you me...Do you use Oracle studio analyzer ? I think you mention it in your other blog post.<br /><br />I am trying to understand if I can run a false sharing Java program and still look at the thread id's, cache lines to demonstrate false sharing. Are you aware of any other tools apart from intel's ?Mohan Radhakrishnanhttps://www.blogger.com/profile/08457140016320542845noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-16979202950060247852010-09-12T22:10:16.564-07:002010-09-12T22:10:16.564-07:00@BlueDavy - Yup, we use profiler that works intern...@BlueDavy - Yup, we use profiler that works internally with our pprof / Google perftools framework. Screenshots are pretty pointless, since they are just as ugly as the perftools stuff. However, no one is doing the updates on the public perftools framework to make it work with Java.Jeremy Mansonhttps://www.blogger.com/profile/04241094734813086257noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-49109930900844089112010-09-12T22:02:12.315-07:002010-09-12T22:02:12.315-07:00@kontiky - That looks like a pretty poorly written...@kontiky - That looks like a pretty poorly written microbenchmark. I would suggest you look at Cliff Click and Brian Goetz's various talks on how to write microbenchmarks in Java - it is kind of a black art.<br /><br />@anonymous - we're working on it (it's a background task). We do suffer the problem of regular sampling, because our sampler samples once every 10ms without adding a Jeremy Mansonhttps://www.blogger.com/profile/04241094734813086257noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-33136659898756031452010-09-11T08:57:08.791-07:002010-09-11T08:57:08.791-07:00Jeremy: I want to second the plea for you guys to ...Jeremy: I want to second the plea for you guys to open source your profiler. Please do this even if the Ui is not a slick GUI--I need the functionality now.<br /><br />I was really impressed with that Mytkowicz paper when it came it some months ago. It was brilliant in more ways than you mentioned. In particular, the whole way that they define actionable profiles is spot on. Also, how they Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-5266830776681401302010-09-01T23:28:03.743-07:002010-09-01T23:28:03.743-07:00does u mean google wrote a profiler for java just ...does u mean google wrote a profiler for java just like google perf-tools,it can be used for online app,and do cpu profile and memory profile?<br />If yes,can u say more about the cpu & memory profile how to realize and give some screenshots?bluedavyhttps://www.blogger.com/profile/06260502985181568028noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-46539182648915755142010-09-01T21:20:08.210-07:002010-09-01T21:20:08.210-07:00@Palo - VisualVM uses the same, documented methods...@Palo - VisualVM uses the same, documented methods, and has the same limitations (Sorry about my earlier post - I misread your comments).Jeremy Mansonhttps://www.blogger.com/profile/04241094734813086257noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-60954867284030715552010-09-01T21:19:03.102-07:002010-09-01T21:19:03.102-07:00This comment has been removed by the author.Jeremy Mansonhttps://www.blogger.com/profile/04241094734813086257noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-30704757799865994262010-08-10T01:51:28.923-07:002010-08-10T01:51:28.923-07:00Hi Jeremy,
what about VisualVM Sampler plugin? As ...Hi Jeremy,<br />what about VisualVM Sampler plugin? As I understand, it uses some sort of sampling, both stacktraces and heap dumps.<br />Did not find official page, but here's an announcement from author: <a rel="nofollow">http://java.dzone.com/announcements/visualvm-12-great-java</a>Unknownhttps://www.blogger.com/profile/06108545412582601582noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-22756990516792017212010-08-05T11:54:12.998-07:002010-08-05T11:54:12.998-07:00Hello, Jeremy!
We've found strange problem in ...Hello, Jeremy!<br />We've found strange problem in our real system<br />http://forums.sun.com/thread.jspa?threadID=5443540<br />Could you, please, explane this strange performance degradation and advise problem solution?kontikyhttps://www.blogger.com/profile/00848709732204078869noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-89791091695948368752010-08-02T15:57:50.489-07:002010-08-02T15:57:50.489-07:00Hi Peter -
My understanding is that the hardware ...Hi Peter -<br /><br />My understanding is that the hardware performance counters for VTune are associated with a particular method, not a stack trace / call graph. AFAICT - I'm neither a VTune developer or user - VTune's stack trace / call graph tracking mechanism uses JVMTI to be informed when a method is invoked, and then VTune keeps track of that information internally (that approachJeremy Mansonhttps://www.blogger.com/profile/04241094734813086257noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-89747787891801403142010-08-02T15:21:28.708-07:002010-08-02T15:21:28.708-07:00Hi Jeremy,
what is your opinion about profilers t...Hi Jeremy,<br /><br />what is your opinion about profilers that are able to integrate into the cpu to read out performance monitors available there. E.g. like Intel VTune.pveentjerhttps://www.blogger.com/profile/17847641595368096163noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-86990567438452703962010-08-02T12:00:00.231-07:002010-08-02T12:00:00.231-07:00We're considering it. The problem is that our...We're considering it. The problem is that our team is very good at figuring out JIT issues and POSIX semantics, but not so great with the user interfaces. :)Jeremy Mansonhttps://www.blogger.com/profile/04241094734813086257noreply@blogger.comtag:blogger.com,1999:blog-8405107760807432973.post-26000887423703890202010-08-02T01:26:51.132-07:002010-08-02T01:26:51.132-07:00Any chance this profiler work will show up in an o...Any chance this profiler work will show up in an open-source project?Noel Grandinhttps://www.blogger.com/profile/00203932250096476916noreply@blogger.com