RFR: JDK-8141634 Implement VarHandles/Unsafe intrinsics on SPARC
Zoltán Majó
zoltan.majo at oracle.com
Mon Aug 29 07:41:14 UTC 2016
Hi Trevor,
On 08/26/2016 05:12 PM, Trevor Watson wrote:
> I've updated the code based on David's feedback and successfully run
> JPRT hotspot and jtreg compiler/unsafe tests.
>
> Revised code is at:
> http://cr.openjdk.java.net/~alanbur/JDK-8141634/
that looks good to me. I'll run some additional tests, will keep you
updated.
Thank you!
Best regards,
Zoltan
>
> Many thanks,
> Trevor
>
> On 26/08/16 09:34, Trevor Watson wrote:
>> Ok, thanks. I'll remove that code from my changes, retest and resubmit
>> the webrev.
>>
>> On 26/08/16 09:02, David Holmes wrote:
>>> Hi Trevor,
>>>
>>> On 26/08/2016 6:00 PM, Trevor Watson wrote:
>>>> Hi David,
>>>>
>>>> Thanks for that information. I already suspected that 32-bit was no
>>>> longer supported, but as I'm new to working on this code I didn't want
>>>> to go changing to many things at once.
>>>>
>>>> Would you be okay if I logged a separate bug to remove the 32-bit
>>>> checks
>>>> and supports_cx8 predicate? That was I can focus on just those things.
>>>
>>> I was only suggesting not including them in this new code. Removing
>>> them
>>> from existing code is a different matter.
>>>
>>> Thanks,
>>> David
>>>
>>>> Trevor
>>>>
>>>> On 26/08/16 04:44, David Holmes wrote:
>>>>> Hi Trevor,
>>>>>
>>>>> Not a full review as I'm not familiar with .ad file syntax and
>>>>> structure. Two observations:
>>>>>
>>>>> 1. We no longer support 32-bit so that code could be elided.
>>>>> 2. supports_cx8 is true for sparc so doesn't need to verified in
>>>>> sparc
>>>>> specific code (it is intended for use in shared code).
>>>>>
>>>>> Cheers,
>>>>> David
>>>>>
>>>>> On 26/08/2016 12:01 AM, Trevor Watson wrote:
>>>>>> Hi All,
>>>>>>
>>>>>> Please can I get a review of these updates to sparc.ad which adds in
>>>>>> the
>>>>>> C2 compareAndExchange intriniscs for int, long, pointer and object?
>>>>>>
>>>>>> I've run the hotspot jtreg tests locally and the JPRT hotspot tests
>>>>>> successfully.
>>>>>>
>>>>>> The implementation of these intrinsics gives a 9x to 20x performance
>>>>>> improvement for compareAndExchange operatios.
>>>>>>
>>>>>> http://cr.openjdk.java.net/~alanbur/JDK-8141634/
>>>>>>
>>>>>> Thanks,
>>>>>> Trevor
More information about the hotspot-dev
mailing list