[lworld] RFR: lworld aaload/aastore

David Simms david.simms at oracle.com
Fri Mar 16 15:55:14 UTC 2018


Ah it must be Friday, I fire off this email too early...

There is also a bug left in "ValueKlass::raw_value_byte_size()" (not 
checking for ref/flattened), and more importantly, aastore still needs 
to check oops "is buffer" when storing to objArray

On 16/03/18 16:46, David Simms wrote:
>
> So I updated the aaload/store interpreter for lworld, adding more 
> tests and covariant array copy, here's an updated webrev:
>
> http://cr.openjdk.java.net/~dsimms/valhalla/lworld_value_aaloadstore/webrev1/ 
>
>
> The ProfileInterpreter bug is still present however...going to need to 
> understand why (doesn't appear to be stack change, could be wrong)...
>
>
> On 14/03/18 15:41, David Simms wrote:
>>
>> Thanks for the heads up...I will double check with my latest fixes + 
>> latest repo.
>>
>> Since posting the RFR yesterday I've increased tests and found a few 
>> bugs. Also started implementing covariance and arraycopy to handle it...
>>
>> We'll see if I can update the webrev today or early tomorrow.
>>
>> /D
>>
>> On 14/03/18 14:21, Tobias Hartmann wrote:
>>> Update: All tests pass with -XX:-ProfileInterpreter -Xint.
>>>
>>>
>>> On 14.03.2018 14:17, Tobias Hartmann wrote:
>>>> Hi David,
>>>>
>>>> I've executed 
>>>> test/hotspot/jtreg/compiler/valhalla/valuetypes/TestArrays.java 
>>>> with -Xint and your
>>>> patch but I'm seeing crashes:
>



More information about the valhalla-dev mailing list