From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040207 Firefox/0.8 Description of problem: The document given by the URL has a complete description of the problem. (I will also try to attach this file to the bug report.) Briefly, a particular CPU intensive Java program that ran fine under Redhat 9 using a variety of Java versions exhibits *massive* slowdowns under RHEL. Also, when using Sun JDKs, there is tons of disk activity when the program runs. (It should be almost entirely CPU bound, and is CPU bound on Redhat 9.) I have tried setting the LD_ASSUME_KERNEL environment variable to 2.4.1, and it does not help. The program in question is single threaded. I listed the IBM JDK in this bug report, but the program doesn't run correctly on RHEL with any JVM I have tried. The same program runs perfectly well on other architectures (such as Mac OS X and Solaris), so I do not believe that it is a bug in the program itself. The problem seems to be partially affected by SMP - on a two processor system, the program makes almost no progress, while on a uniprocessor system running the same version of RHEL, the slowdown is about 3x, but it eventually completes. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: Steps to reproduce are given in the problem description file referred to by the URL. Actual Results: Massive slowdown, tons of disk activity, interactive responsiveness of other processes greatly reduced, process extremely slow to respond to interrupt signal from tty. Expected Results: The program should be CPU bound, and complete in a few minutes (given the run parameters and input described). Additional info:
Created attachment 98883 [details] Complete description of the problem, and how to reproduce.
Sometime between March 26th (when I posted this bug originally) and June 22nd, the problem went away. I'm pretty sure one of the kernel updates must have fixed it, although I have no idea which one. Java programs execute fine now.
OK, closing.