RFR: 8184017: [MVT] Support Q-types in MHs.insertArguments()

Maurizio Cimadamore maurizio.cimadamore at oracle.com
Mon Jul 10 16:48:55 UTC 2017


Looks good

Maurizio


On 10/07/17 14:06, Vladimir Ivanov wrote:
> Thanks, Maurizio & Paul.
>
> Yes, all tests in jdk/test/valhalla/mvt pass.
>
> I decided to write a dedicated unit test and here's an updated version:
>   http://cr.openjdk.java.net/~vlivanov/valhalla/8184017/webrev.01/
>
> Also, added explicit null check to align the behavior on MVTs with 
> primitives.
>
> Best regards,
> Vladimir Ivanov
>
> On 7/7/17 7:58 PM, Maurizio Cimadamore wrote:
>> Looks good - did you run existing tests just in case? They are in the 
>> jdk jtreg test group called jdk_valhalla_mvt
>>
>> Maurizio
>>
>>
>> On 07/07/17 17:37, Vladimir Ivanov wrote:
>>> http://cr.openjdk.java.net/~vlivanov/valhalla/8184017/webrev.00
>>>
>>> Q-typed argument binding support. The value is stored in boxed form 
>>> inside BMH and unboxed on access.
>>>
>>> Before: a1:Q
>>> invoke061_LQL_Q=Lambda(a0:L/SpeciesData<LLLL>,a1:Q,a2:L)=>{
>>>
>>> After: t4:Q
>>> invoke065_LL_Q=Lambda(a0:L/SpeciesData<LLLLLL>,a1:L)=>{
>>>     t2:L=BoundMethodHandle$Species_L6.argL5(a0:L);
>>>     t3:L=BoundMethodHandle$Species_L6.argL4(a0:L);
>>>     t4:Q=MethodHandle.invokeBasic(t2:L,t3:L);
>>>
>>> I haven't added any test case yet, but there will be one in an 
>>> updated test on vectors.
>>>
>>> Best regards,
>>> Vladimir Ivanov
>>



More information about the valhalla-dev mailing list