RFR (S): 8020215: Different execution plan when using JIT vs interpreter

Vladimir Kozlov vladimir.kozlov at oracle.com
Wed Jul 10 18:19:02 PDT 2013


Thanks!

Vladimir

On 7/10/13 5:56 PM, Christian Thalinger wrote:
>
> On Jul 10, 2013, at 4:44 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
>
>> Thanks, Christian
>>
>> How about returns_all(vars)?
>
> Better.  -- Chris
>
>>
>> Vladimir
>>
>> On 7/10/13 4:10 PM, Christian Thalinger wrote:
>>> The change looks good.  I just find the name of the method odd:
>>>
>>> + bool BCEscapeAnalyzer::is_returned_all(ArgumentMap vars) {
>>>
>>> -- Chris
>>>
>>> On Jul 10, 2013, at 3:11 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
>>>
>>>> http://cr.openjdk.java.net/~kvn/8020215/webrev/
>>>>
>>>> The problem is caused by incomplete result returned by bytecode analyzer.
>>>> Return result should be invalidated if args are not recorded in return state because it is unknown if these args could be returned or not without additional pass over bytecode. And bytecode analyzer does only one pass.
>>>>
>>>> Thanks,
>>>> Vladimir
>


More information about the hotspot-compiler-dev mailing list