We've been lining up some really nice performance improvements in the OpenJDK recently.
One of my favorites in the last few months is https://bugs.openjdk.org/browse/JDK-8297933 by Roland Westrelin (Red Hat). This has been seen to improve scores on entire benchmarks by double digits or even integer factors. Seems to be in a reasonably good state since 21 EA build 5.
We can improve various String methods such as startsWith, endsWith and regionMatches by leveraging the intrinsified mismatch methods in ArraysSupport. Progress Change must be properly reviewed ...
@cl4es 21.EA.9 seems to be about 3.5% faster for me than 17.0.6.
I have seen one data set that shows even 7% better performance. I'm not sure about the origin, of course.
@reneschwietzke nice! we also seem to be on a good trajectory since 11 here.
The difference between 17.0.6 and 21 might be smaller than most think, seeing that it's "only" a bit in excess of 1.5 years of development so far -- and several of the performance improvements implemented in that time frame have been backported.
@cl4es Understood. It is also important to note that this example is computational heavy. A typical microservice implementation might not see a thing.
In a previous example (same test and machine type) I have seen 19 and 20 being slower than 17.
cl4es
Mark Derricutt (talios)
René Schwietzke