Re: [rvm-research] JikesRVM as a static compiler
JOE <[email protected]> Sun, 22 Jan 2017 02:13:47 +0000
| Newsgroups | gmane.comp.java.jikes.rvm.devel |
|---|---|
| Message-ID | <CAKNw+yhtyJMdktzOGEkbYz+Chff3t_ZpMzky1f2Wzs6eV2FhtQ@mail.gmail.com> |
Hi Erik, Thanks for the pointers. I have been using the SemiSpace collector. Are there any known issues with that? It seems to be working for me. I have seen the debug information printed when it crashes and I find it quite impressive and very helpful in finding my problems. Compliance with the standard Java library is not important since that library is OS dependent. Operating system concepts are over 40+ years old and its abstractions and interfaces are dated and stale. I believe the OS gets in the way of running an object oriented language because it has no relevant support for it. Take look at the JikesRVM’s OS and JNI code. It is very complicated and involved. I am really impressed how the Jikes team was able get that code and interface to work. I just want thank you for helping make JikesRVM such an incredible tool and VM. With best regards, Joe On Sat, Jan 21, 2017 at 4:33 AM Erik Brangs <[email protected]> wrote: > Hi, > > On 19.01.2017 16:49, JOE wrote: > > Has there been any research or projects using JikesRVM as a static > compiler to run a Java program natively on a computer without an OS? The > static compilation would supply Java objects that provide equivalent OS > services and provide access to the computer hardware devices. > I'm not aware of any such research. However, this doesn't mean much. I'm > not a researcher and consequently don't follow research papers closely. > > If you wanted to try to use Jikes RVM in such a way, you'd likely run into > several problems. > > Stability would likely be a big problem. For example, the default > collector (GenImmix) likely has a space leak and is prone to crashing on > SPECjbb2005. Moreover, there's generally not much effort in the code base > to recover from errors. The VM just crashes, possibly printing debug > information. > > It's also likely that you'd have to invest in improving the compilers if > you want to have any kind of reasonable comparison with other ahead-of-time > compilers. For example, the current opt compiler doesn't have SSA. It also > doesn't support important optimizations like array bound check elimination > or partial redudancy elimination. > > Other problems might include incomplete coverage of the standard library > (due to use of GNU Classpath instead of OpenJDK), non-compliance with the > JVM spec (e.g. JMM) and lack of concurrent compacting garbage collectors in > mainline Jikes RVM. > > If you're interested in a Java-based OS, there's the JNode project ( > https://github.com/jnode/jnode / http://www.jnode.org/ ). > > If you wanted to have a native Java implementation without underlying OS > for productive use (as opposed to research), your best bet would probably > be to contact a company that sells realtime JVMs and have them evaluate > whether this is feasible. I suppose there are several companies that do > this but the only one that I can think of off the top of my head is aicas, > which sells the JamaicaVM. > > > Kind regards, > > Erik Brangs > > > ------------------------------------------------------------------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, SlashDot.org! http://sdm.link/slashdot > _______________________________________________ > Jikesrvm-researchers mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers > ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot _______________________________________________ Jikesrvm-researchers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers