RFR(L): 8212245 Remove Thread-Local Value Buffer

Frederic Parain frederic.parain at oracle.com
Wed Oct 17 13:36:01 UTC 2018


Tobias,

Thank you for your review.
The compiler test has been cleaned up:

http://cr.openjdk.java.net/~fparain/8212245/webrev.01/

Regards,

Fred


> On Oct 17, 2018, at 04:57, Tobias Hartmann <tobias.hartmann at oracle.com> wrote:
> 
> Okay, you did remove the VTBufferTest but BigValueTypeThreshold is still used by the compiler tests.
> 
> Thanks,
> Tobias
> 
> On 17.10.2018 10:56, Tobias Hartmann wrote:
>> Hi Frederic,
>> 
>> this looks good to me, thanks for also removing the C2 code.
>> 
>> I think you missed to remove the corresponding tests and also the flag usages. For example:
>> 
>> test/hotspot/jtreg/runtime/valhalla/valuetypes/VTBufferTest.java: * @run main/othervm -Xint
>> -XX:+EnableValhalla -XX:ValueTypesBufferMaxMemory=0 VTBufferTest generate-and-run
>> 
>> test/hotspot/jtreg/compiler/valhalla/valuetypes/TestCallingConvention.java:        case 1: return
>> new String[] {"-XX:ValueTypesBufferMaxMemory=0"};
>> 
>> Best regards,
>> Tobias
>> 
>> On 16.10.2018 21:42, Frederic Parain wrote:
>>> Please review this large but simple patch removing the Thread-Local Value Buffer.
>>> This buffer used by the interpreter didn’t demonstrate performance improvements
>>> and it causes complexity in the JIT code.
>>> 
>>> CR: https://bugs.openjdk.java.net/browse/JDK-8212245
>>> Webrev: http://cr.openjdk.java.net/~fparain/8212245/webrev.00/index.html
>>> 
>>> Testing: hotspot_jtreg:hotspot_valhalla on Linux/Mac/Windows.
>>> 
>>> Thank you,
>>> 
>>> Fred
>>> 



More information about the valhalla-dev mailing list