|
|
|
Back to newsletter 257 contents
A reader asked me if the Xprof profiler was still of any use. The answer is no, it's old tech that was removed from the JVM a few years ago (JDK 10). For those of you unaware of it (I first wrote about it nearly 20 years ago), the HotSpot JVM had the Xprof profiler embedded in it, and it was used by the JVM to identify which methods to compile to native code. Subsequently other techniques are used and because Xprof caused a root scanning overhead, it was removed.
What is still of interest to us are the techniques used to make it the lowest possible overhead profiler: infrequent sampling, no stack walking, no thread matching or analysis, minimal processing of data (just a counter per method signature). These are techniques that you can apply to your applications to find things that can be analyzed later for improvements; they are also how modern object creation profiling is done in JFR to keep overhead low, where samples are only taken rarely (when the TLAB gets full), the sampled stack size is configurable and the events processed elsewhere.
Now on to all the usual newsletter list of links, tips, tools, news and articles, and as usual I've extracted all the tips into this month's tips page
Java performance tuning related news
Java performance tuning related tools
Back to newsletter 257 contents