Comments on Java Code Geeks: Java 7: How to write really fast Java code

tag:blogger.com,1999:blog-8574118689743302986.post2096689243072590033..comments2012-01-25T06:29:47.600+02:00Comments on Java Code Geeks: Java 7: How to write really fast Java codeByron Kiourtzoglouhttp://www.blogger.com/profile/00834323472598416198noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-8574118689743302986.post-47615046519556960372012-01-25T06:29:47.601+02:002012-01-25T06:29:47.601+02:00Dunno how you would test it but I suspect the majo...Dunno how you would test it but I suspect the majority of the performance difference is due to the write barriers for AtomicLong as it abides by the visibility requirements of the JVM Memory Model. The Memory Model implies (though impls may differ) main memory (not cache) is made consistent thus for each AtomicLong compareAndSet its a much longer bus ride than simply updating L3 cache (as in the thread local RNG)Stoflet Darrylnoreply@blogger.com