RFR 8183131: [MVT] 2-slot type (long & double) support is broken in LambdaFormBuilder

Maurizio Cimadamore maurizio.cimadamore at oracle.com
Wed Jul 5 11:08:41 UTC 2017


I did a slightly more ambitious refactoring of the lambda form builder code:

* all methods concerned with code generation have been moved to 
LambdaFormCodeBuilder - more in the spirit of the design of the bytecode API
* names of methods have been renamed - the 'emit' prefix has been 
dropped - to make it more builder-friendly
* load/store of builders are now overrides of basic load/store - they 
map the index using the LF local map and they delegate to the parent method

Let me know what you think

http://cr.openjdk.java.net/~mcimadamore/8183131_v2/

Maurizio


On 03/07/17 22:13, Maurizio Cimadamore wrote:
> I see what you mean - I will give this a try
>
> Maurizio
>
>
> On 03/07/17 21:53, Vladimir Ivanov wrote:
>> Looks good.
>>
>> Probably, it's better to keep builder.store() wrapped in a helper 
>> method (as it is done in InvokerBytecodeGenerator) to keep 
>> localsMap[] conversion implicit.
>>
>> Best regards,
>> Vladimir Ivanov
>>
>> On 7/3/17 5:19 PM, Maurizio Cimadamore wrote:
>>> Hi,
>>> This patch fixes an issue with lambda form builder not taking into 
>>> account single vs. double slots indices.
>>>
>>> http://cr.openjdk.java.net/~mcimadamore/8183131/
>>>
>>> Thanks
>>> Maurizio
>>>
>>>
>



More information about the valhalla-dev mailing list