Invokedynamic slower than reflection?

[Chanwit Kaewkasi]

Hi Rémi,

I did freshly build mlvm and run Eric's test.
Here's my result:

Duration invokedynamic: 103
Duration reflection: 1463

The patch seems to be working for me as InvokeDynamic is clearly
faster than Reflection on my machine.

Cheers,

Chanwit
-- 
Dr Chanwit Kaewkasi, Lecturer
School of Computer Engineering,
Suranaree University of Technology
Nakhon Ratchasima, Thailand 30000

On Thu, Jul 29, 2010 at 06:26, Rémi Forax <forax at univ-mlv.fr> wrote:
> Le 29/07/2010 01:12, John Rose a écrit :
>> On Jul 27, 2010, at 10:11 AM, Rémi Forax wrote:
>>
>>
>>> Le 16/07/2010 00:48, John Rose a écrit :
>>>
>>>> It was some sort of bitrot.
>>>>
>>>> I pushed a fix for this to mlvm/hotspot.
>>>>
>>>> Thanks for the reports!
>>>>
>>>> -- John
>>>>
>>>>
>>> Is it enable by default, it doesn't seem to work ?
>>>
>> The fix is titled this way:
>>
>>    6969574: invokedynamic call sites deoptimize instead of executing
>>
>> The bug is (a) on the mlvm patch queue, and (b) working its way through the OpenJDK RE pipeline.
>>
>> Current state of (b) may be observed by looking at these three workspaces:
>>    http://hg.openjdk.java.net/jdk7/hotspot-comp/hotspot/shortlog  (* I push mlvm patches in here)
>>    http://hg.openjdk.java.net/jdk7/hotspot/hotspot/shortlog  (* hotspot team integrates here)
>>    http://hg.openjdk.java.net/jdk7/jdk7/hotspot/shortlog  (RE builds feed from this)
>>
>> The fix 6969574 is in the starred workspaces.  Therefore, it is not yet in the binary builds from RE.
>>
>
> Sorry, I should be more specific.
> My question was more, is it enable in mlvm if I use default guards.
>
> Since, I've checked that the patch is correctly applied to my workspace
> source,
> and it's correctly applied.
>
> So I think the patch doesn't fix the issue.
>
>> -- John
>>
>
> Rémi
> _______________________________________________
> mlvm-dev mailing list
> mlvm-dev at openjdk.java.net
> http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
>