Can't get lambda forest to work

Maurizio Cimadamore maurizio.cimadamore at oracle.com
Tue Jun 15 06:49:47 PDT 2010


I did a merge with b97, I hope that helps.

Maurizio

On 15/06/10 09:29, Andrew Haley wrote:
> On 06/14/2010 06:52 PM, Rémi Forax wrote:
>    
>> Le 14/06/2010 18:58, Andrew Haley a écrit :
>>      
>>> On 06/14/2010 05:39 PM, Andrew Haley wrote:
>>>
>>>        
>>>> On 06/14/2010 05:28 PM, Maurizio Cimadamore wrote:
>>>>
>>>>          
>>>>> I tried launching the 'java' launcher from a recently built lambda repo;
>>>>> I added the options:
>>>>>
>>>>> -XX:+UnlockExperimentalVMOptions -XX:+EnableMethodHandles
>>>>> -XX:+EnableInvokeDynamic
>>>>>
>>>>>
>>>>> and that seemed to work, at least it did for me. I noticed you are using
>>>>> a 64-bit setting, while I'm on 32-bit, could that be the issue?
>>>>>
>>>>>            
>>>> Maybe.  I'm just going to step through the VM in a debugger and have a look.
>>>> I'll let you know what I find.  :-)
>>>>
>>>>          
>>> OK, I found.  The VM is looking for java_dyn_CallSite.target
>>> in java_dyn_CallSite::compute_offsets()
>>>
>>> in the Lambda forest, this field is commented out:
>>>
>>>       /*
>>>
>>>       // Fields used only by the JVM.  Do not use or change.
>>>       private Object vmmethod;
>>>       int callerMID, callerBCI;  // supplied by the JVM
>>>
>>>       private MethodHandle target;
>>>
>>>       final Object caller;  // usually a class
>>>       final String name;
>>>       final MethodType type;
>>>       */
>>>
>>> but this code was commented out in this change:
>>>
>>> user:        jrose
>>> date:        Thu Jan 07 16:16:45 2010 -0800
>>> summary:     6914665: update jdk code for JSR 292 (post 6858164)
>>>
>>> Cc'ing hotspot-dev in the hope of enlightenment...
>>>        
>> This is a known bug :)
>> Since b95, John Rose commits support for lightweight callsite (among
>> other things)
>> but forget to commit the JDK runtime part.
>>
>> see http://mail.openjdk.java.net/pipermail/mlvm-dev/2010-May/001728.html
>> for a workaround.
>>      
> OK, thanks.
>
> Andrew.
>
>    



More information about the lambda-dev mailing list