RFR(M): 8181211: C2: Use profiling data to optimize on/off heap unsafe accesses

Vladimir Kozlov vladimir.kozlov at oracle.com
Tue Jun 6 14:47:47 UTC 2017


On 6/6/17 4:13 AM, Roland Westrelin wrote:
>
> Hi Vladimir,
>
> Thanks for the test results.
>
>> SA jtreg tests failed (tests ran in 2 modes: -Xmixed and -Xcomp):
>>
>> compiler/ciReplay/TestSAServer.java
>> serviceability/sa/TestPrintMdo.java
>
> There's a missing entry in the vmStructs. Once fixed, I verified both
> tests pass now.

New webrev or updated webrev.00?

>
>> Some tests failed because you moved RM:
>>
>> http://cr.openjdk.java.net/~roland/8181211/webrev.00/src/share/vm/oops/method.cpp.udiff.html
>
> That RM can lead to a dandling pointer if the outputStream argument is
> not tty and the buffer that is passed grows under the RM. Anyway, it may
> be better to leave that RM in in that change and take care of the bug
> separately?

Yes, file a separate bug.

Vladimir

>
> Roland.
>


More information about the hotspot-compiler-dev mailing list