[13] RFR (S): 8218758: [TESTBUG] compiler/cha/StrengthReduceInterfaceCall.java misses recompilation event
Vladimir Ivanov
vladimir.x.ivanov at oracle.com
Thu Feb 14 21:10:11 UTC 2019
Thanks, Igor.
Best regards,
Vladimir Ivanov
On 13/02/2019 19:58, Igor Ignatyev wrote:
> Vladimir,
>
> in case you are waiting for my response, I'm fine w/ the current state of webrev.00.
>
> -- Igor
>
>> On Feb 12, 2019, at 4:12 PM, Vladimir Ivanov <vladimir.x.ivanov at oracle.com> wrote:
>>
>> Yes, that's also an option, but I think just ensuring there's a nmethod registered (and keeping existing method if any present) should work fine as well.
>>
>> Best regards,
>> Vladimir Ivanov
>>
>> On 12/02/2019 16:06, Igor Ignatyev wrote:
>>> you can use WB.deoptimizeMethod to guarantee the the method is not compiled before the loop.
>>> -- Igor
>>>> On Feb 12, 2019, at 4:00 PM, Vladimir Ivanov <vladimir.x.ivanov at oracle.com> wrote:
>>>>
>>>> Thanks, Igor.
>>>>
>>>> I made a slight adjustment in the original patch:
>>>>
>>>> public void compile(Runnable r) {
>>>> - assertFalse(WB.isMethodCompiled(TEST));
>>>> while (!WB.isMethodCompiled(TEST)) {
>>>>
>>>> If recompilation happens before compile() is called, it would cause problems there as well.
>>>>
>>>> Best regards,
>>>> Vladimir Ivanov
>>>>
>>>> On 12/02/2019 15:52, Igor Ignatyev wrote:
>>>>> Hi Vladimir,
>>>>> looks good to me.
>>>>> Thanks,
>>>>> -- Igor
>>>>>> On Feb 12, 2019, at 2:07 PM, Vladimir Ivanov <vladimir.x.ivanov at oracle.com> wrote:
>>>>>>
>>>>>> http://cr.openjdk.java.net/~vlivanov/8218758/webrev.00/
>>>>>> https://bugs.openjdk.java.net/browse/JDK-8218758
>>>>>>
>>>>>> The test checks that nmethod has been invalidated by ensuring there's no nmethod installed at the moment. But it's not robust enough: it is possible that JVM initiates compilation before the check happens.
>>>>>>
>>>>>> The fix is to harden the logic to additionaly check that if there's nmethod installed it differs from the one observed earlier.
>>>>>>
>>>>>> Testing: hs-precheckin-comp, tier1-5
>>>>>>
>>>>>> Best regards,
>>>>>> Vladimir Ivanov
>
More information about the hotspot-compiler-dev
mailing list