On Mar 3, 2011, at 2:36 PM, Christian Thalinger wrote: > On Mar 3, 2011, at 2:25 PM, Rémi Forax wrote: >> On 03/03/2011 02:10 PM, Christian Thalinger wrote: >>> On Mar 3, 2011, at 1:52 PM, Christian Thalinger wrote: >>>> On Mar 3, 2011, at 1:12 PM, Christian Thalinger wrote: >>>>> On Mar 3, 2011, at 9:47 AM, Rémi Forax wrote: >>>>>> Yes. I use an agent so it can have trigger a deoptimization >>>>>> even if I don't understand why ? >>>>> Here is what I get with a debug build: >>>> <snip> >>>> >>>>> The assert looks like: >>>>> >>>>> 7016279: JSR292 Exception when creating CallSites from multiple threads >>>>> >>>>> Is your source code creating CallSites from multiple threads (massively >>>>> parallel)? >>>> >>>> Scratch that. I think I found the problem: the constant pool cache entry >>>> is not properly updated after a GC thus resulting in a bogus value for the >>>> BSM. >>> UseSerialGC seems to do proper pointer updating. -- Christian >> >> Yes, it's a workaround. > > Thanks for that testcase! I supposed we still have a problem with parallel > GC but I didn't have a failing testcase to prove it.
I filed: 7024561: JSR 292 constant pool cache _f1 field doesn't get properly updated with parallel GC -- Christian _______________________________________________ mlvm-dev mailing list mlvm-dev@openjdk.java.net http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
