[9] RFR(S): 8035328: closed/compiler/6595044/Main.java failed with timeout

[Tobias Hartmann]

Thank you, Chris.

Best,
Tobias

On 21.06.2014 00:13, Christian Thalinger wrote:
> Since the comments talk about “waste of time” this looks good to me.
>
> On Jun 19, 2014, at 12:15 AM, Tobias Hartmann 
> <tobias.hartmann at oracle.com <mailto:tobias.hartmann at oracle.com>> wrote:
>
>> Hi,
>>
>> please review the following patch for 8035328.
>>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8035328
>> Webrev: http://cr.openjdk.java.net/~thartmann/8035328/webrev.00
>>
>> *Problem**
>> *This bug is not a test bug. The test executes hundreds of small 
>> methods consisting of only a return statement multiple times. This 
>> was intended to test for an old (closed) bug in the C2 compiler. The 
>> timeout arises when the tiered compilation policy decides to 
>> recompile the caller method and therefore marks it as non-entrant. 
>> The method is still in use but call sites are not patched. See line 
>> 1179 of sharedRuntime.cpp:
>>
>> // We do not patch the call site if the caller nmethod has been made 
>> non-entrant.
>>
>> As a result, all call sites have to be resolved on every call which 
>> is costly and leads to the timeout (4mins vs. 17s).
>>
>> *Solution
>> *The call site patching code in sharedRuntime is changed to patch not 
>> entrant call sites as well.
>> *
>> **Testing
>> *- JPRT
>> - SPECjbb2005, SPECjbb2013
>> - SPECjvm2008-Compiler*
>> *- Nashorn + Octane*
>> *- UTE full run with Kitchensink
>>
>> Thanks,
>> Tobias
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20140624/f8d299c8/attachment-0001.html>