RFR: 8190925: Deprecate FastTLABRefill and set default off

Robbin Ehn robbin.ehn at oracle.com
Tue Nov 14 08:56:53 UTC 2017


Thanks!

/Robbin

On 11/13/2017 10:30 PM, David Holmes wrote:
> Looks good!
> 
> Thanks,
> David
> 
> On 13/11/2017 11:33 PM, Robbin Ehn wrote:
>> Hi all, please review.
>>
>> The option FastTLABRefill work only in C1 (interpreter and c2 don't use this 
>> option) when using either SerialGC or ParrallelGC (and deprecated CMS). Both 
>> the newly integrated Memory-Leak Profiler and upcoming Low Overhead Profiler 
>> don't work when FastTLABRefill is enabled.
>>
>> For more info please see CSR:
>> https://bugs.openjdk.java.net/browse/JDK-8190926
>>
>> Issue:
>> https://bugs.openjdk.java.net/browse/JDK-8190925
>>
>> Code below, thanks Robbin!
>>
>> diff -r cf127be65014 src/hotspot/share/runtime/arguments.cpp
>> --- a/src/hotspot/share/runtime/arguments.cpp    Sat Nov 11 01:21:09 2017 +0100
>> +++ b/src/hotspot/share/runtime/arguments.cpp    Mon Nov 13 14:26:27 2017 +0100
>> @@ -385,2 +385,3 @@
>>     { "UseMembar",                    JDK_Version::jdk(10), 
>> JDK_Version::jdk(11), JDK_Version::jdk(12) },
>> +  { "FastTLABRefill",               JDK_Version::jdk(10), 
>> JDK_Version::jdk(11), JDK_Version::jdk(12) },
>>     { "IgnoreUnverifiableClassesDuringDump", JDK_Version::jdk(10), 
>> JDK_Version::undefined(), JDK_Version::undefined() },
>> diff -r cf127be65014 src/hotspot/share/runtime/globals.hpp
>> --- a/src/hotspot/share/runtime/globals.hpp    Sat Nov 11 01:21:09 2017 +0100
>> +++ b/src/hotspot/share/runtime/globals.hpp    Mon Nov 13 14:26:27 2017 +0100
>> @@ -2015,4 +2015,4 @@
>>                                                                               \
>> -  product(bool, FastTLABRefill, true,                                       \
>> -          "Use fast TLAB refill code")                                      \
>> +  product(bool, FastTLABRefill, false,                                      \
>> +          "(Deprecated) Use fast TLAB refill code")                         \
>>                                                                               \


More information about the hotspot-dev mailing list