Aleksey Shipilev wrote:
Please see the updated patch [1]. I had solved the NPE issue by
checking currentThread() result - that's odd but can be considered a
workaround.
LUNI tests are ok, DRLVM build is ok, 8x speedup on MT/ThreadLocalBench.
Looking at it now. I'm going to rearrange the files a bit to put
ThreadLocal into the DRLVM's kernel classes so that the IBM VME keeps
working (as I mentioned earlier).
Bob:
I see this comment in the ThreadLocal.java file:
/* Thanks to Josh Bloch and Doug Lea for code reviews and impl advice. */
am I right in thinking this was just design level discussion? If Josh
or Doug contributed directly to the code we'll have to get their
approval to have it Apache Licensed.
Regards,
Tim
