RFR: 8271820: Implementation of JEP 416: Reimplement Core Reflection with Method Handle [v8]

[Peter Levart]

On Wed, 1 Sep 2021 01:05:32 GMT, Mandy Chung <mchung at openjdk.org> wrote:

>> This reimplements core reflection with method handles.
>> 
>> For `Constructor::newInstance` and `Method::invoke`, the new implementation uses `MethodHandle`.  For `Field` accessor, the new implementation uses `VarHandle`.    For the first few invocations of one of these reflective methods on a specific reflective object we invoke the corresponding method handle directly. After that we spin a dynamic bytecode stub defined in a hidden class which loads the target `MethodHandle` or `VarHandle` from its class data as a dynamically computed constant. Loading the method handle from a constant allows JIT to inline the method-handle invocation in order to achieve good performance.
>> 
>> The VM's native reflection methods are needed during early startup, before the method-handle mechanism is initialized. That happens soon after System::initPhase1 and before System::initPhase2, after which we switch to using method handles exclusively.
>> 
>> The core reflection and method handle implementation are updated to handle chained caller-sensitive method calls [1] properly. A caller-sensitive method can define with a caller-sensitive adapter method that will take an additional caller class parameter and the adapter method will be annotated with `@CallerSensitiveAdapter` for better auditing.   See the detailed description from [2].
>> 
>> Ran tier1-tier8 tests.   
>> 
>> [1] https://bugs.openjdk.java.net/browse/JDK-8013527
>> [2] https://bugs.openjdk.java.net/browse/JDK-8271820?focusedCommentId=14439430&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14439430
>
> Mandy Chung has updated the pull request incrementally with one additional commit since the last revision:
> 
>   minor cleanup and more test case.

Hi Mandy,
I'm still looking at this great work and have a few questions that I want to ask upfront.
I understand the need for hard-coded specialization as opposed to inserting more MH transformations. You want to squeeze the startup costs as much as possible. But what I would like to understand is the need for MHInvoker and generated implementations of that interface. I can see that by making the target MH a constant in such generated MHInvoker, JVM is able to optimize the MH invocation chain better when JIT kicks-in. So instead of holding the reference to a target MethodHandle in say DirectMethodHandleAccessor, you hold a reference to MHInvoker. You trade constant MH optimization for indirection to a non-constant MHInvoker instance (I see @Stable annotation there, but it would only be effective when the holder MethodAccessor instance was also constant which unfortunately isn't as it is held by a volatile field in Method so even if Method was constant, its MethodAccessor would not be). So my question is whether this trade pays off.
I wonder what the performance would be if:
- MHInvoker was eliminated
- the DirectMethodHandleAccessor just used the target MH directly (via @Stable field) but still using hard-coded specializations
- AdaptiveMethodHandleAccessor would not be needed then
- the Method had @Stable reference to MethodAccessor instead of volatile (data races are used everywhere so why not for this field too?)
I guess you already tried this approach and later added MHInvoker "middleman" to optimize the warmed-up performance. If not, I can try that variant and come up with benchmark results to compare...

-------------

PR: https://git.openjdk.java.net/jdk/pull/5027