RFR(S) 8235629 Remove unnecessary calls to Thread::current

[Ioi Lam]

On 12/11/19 11:56 AM, Calvin Cheung wrote:
>
> On 12/11/19 10:19 AM, Ioi Lam wrote:
>> https://bugs.openjdk.java.net/browse/JDK-8235629
>> http://cr.openjdk.java.net/~iklam/jdk14/8235629-remove-thread-current.v01/ 
>>
>>
>> + In mutex.cpp, put "assert_owner(Thread::current())" into DEBUG_ONLY 
>> to avoid
>>   the evaluation of the call in product builds.
>> + Changed various "ResourceMark rm;" to "rm(THREAD)".
>
> Found a few more of the above in the following functions:
>
> void MetaspaceShared::preload_and_dump(TRAPS)
> int MetaspaceShared::preload_classes
> bool MetaspaceShared::try_link_class
>
> methodHandle Method::make_method_handle_intrinsic
>
> bool klassVtable::needs_new_vtable_entry
>

Hi Calvin,

Thanks for the info. I've added these to my patch. Here's the delta from 
the last webrev

http://cr.openjdk.java.net/~iklam/jdk14/8235629-remove-thread-current.v02.delta/

I think there are probably many other cases that aren't found in my 
profile. I'll try to write a script to find all the cases and fix them 
in a separate RFE.

Thanks
- Ioi


> thanks,
>
> Calvin
>
>> + In UL logs, use "ResourceMark rm" only conditionally.
>>
>> I found these locations by running "java -version" in gdb, setting a 
>> breakpoint
>> at Thread::current() and logging all the callstacks. So these are by 
>> no means complete,
>> but are rather an easy-to-fix part of JVM start-up.
>>
>> Before: 0.0420053188 sec
>> After:  0.0417649239 sec
>>         99.42770374% = about .6% improvement
>>
>> Thanks
>> - Ioi
>>
>>