Slashdot Mirror


Eclipse Now Runs On Jikes Research Virtual Machine

jscribner writes "IBM reached another key milestone in open source: Eclipse, a completely open source platform, now runs on the Jikes Research Virtual Machine (RVM) improving its teaching/research potential because it provides the community with a significant open source Java benchmark that runs on top of a flexible open testbed (Jikes RVM). The testbed runs on Linux and uses the GNU Classpath implementation of Java libraries (read: complete open source solution). Although Jikes RVM was developed by IBM researchers at the IBM TJ Watson Research Center, it was donated to the community in October 2001 and now has a steering committee and core team that include both IBMers and other university researchers."

2 of 20 comments (clear)

  1. A selection of memory managers by saugart · · Score: 5, Informative
    Even better than just having a modified reference counting scheme, it has a choice of memory managers you can plug in. You can write your own, too. The API for it is JMTk (the Java Memory management Toolkit).

    There's more info on the memory management system in the Jikes RVM user's guide.

    The memory managers supported include:

    • SemiSpace (copying),
    • MarkSweep (non-copying),
    • CopyMS (non-generational copy/mark-sweep hybrid)
    • GenCopy (classic copying generational),
    • GenMS (generational with mark-sweep mature space), or
    • RefCount (a reference counting collector with non-concurrent cycle collection)
    • NoGC (allocation only, no garbage collection)
  2. Re:why is this great by jkauzlar · · Score: 2, Informative
    Jikes is reportedly faster than Sun's JVM and may have better memory management. Many have said they were disappointed with their application's performance, but when they ran it under Jikes, it was like a breath of fresh air. I've heard no portability advantages since Sun does a good job developing for just about every popularly used OS. Open source code could be ported anywhere there's a compiler, of course.

    My hope is that an open source solution would allow for further experimentation with the JIT compiler (compiling byte code to native code on the fly) could actually result in faster JVM code than C code (disregarding load times of course). Experimenters may also find tricks to speed up and improve the GUI as well, which Java seriously needs.