Marcelo Fukushima wrote: > On Wed, Jul 15, 2009 at 11:29 AM, Raffaello > Giulietti<[email protected]> wrote: >> Christian Thalinger wrote: >>> Raffaello Giulietti wrote: >>>> Hello, >>>> >>>> I'd like to start playing with da Vinci, in particular with the java.dyn >>>> classes (InvokeDynamic, CallSite, MethodHandle, etc.) >>>> >>>> They seem to be included in the more recent "Java Platform, Standard >>>> Edition 7 Binary Snapshot Releases" found at >>>> http://download.java.net/jdk7/binaries/. So I'm wondering if they are >>>> complete enough and usable or if I shall build Da Vinci from the >>>> Mercurial repos. >>> Yes, current builds have JSR 292 support, although it's not complete. >> I just successfully compiled and ran a tiny program with one >> InvokeDynamic call site and the bootstrap method. I did the build in >> NetBeans but had to execute the program on the command line. Running >> inside NetBeans doesn't work right, probably because NB sees confusing >> errors in the source. >> >>> One major point is that there is no compiler support and you have to run >>> your programs with the interpreter (-Xint). This, plus some other >>> fixes, has been fixed in the mlvm repository. >>> >> I guess you mean JIT compiler support, not javac. >> >> Seems to work even without -Xint (but, of course, not without >> -XX:+EnableInvokeDynamic). What are the problems that -Xint is supposed >> to circumvent? > > i think he meant that invoke dynamic (also refered to 'indy' on this > list) does not work on compiled mode (JIT compiled that is) >
Marcelo, as I explained, I can run my example even *without* -Xint, i.e., with the JIT enabled. But perhaps the example is so small that the JIT doesn't even start its duties. Anyway, good to know that -Xint exists as a workaround for possible crashes of the JIT. _______________________________________________ mlvm-dev mailing list [email protected] http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
