RFR: 8373114: Redundant MethodCounters in the preimage generated by training run [v3]

[Igor Veresov]

On Wed, 14 Jan 2026 19:54:59 GMT, Vladimir Kozlov <kvn at openjdk.org> wrote:

>> Ashutosh Mehra has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Remove blank line
>>   
>>   Signed-off-by: Ashutosh Mehra <asmehra at redhat.com>
>
> src/hotspot/share/oops/trainingData.cpp line 122:
> 
>> 120: MethodTrainingData* MethodTrainingData::make(const methodHandle& method, bool null_if_not_found, bool use_cache) {
>> 121:   MethodTrainingData* mtd = nullptr;
>> 122:   if ((!have_data() && !need_data()) || (assembling_data() && !CDSConfig::is_at_aot_safepoint())) {
> 
> @ashu-mehra  and @veresov  from what I see `MethodTrainingData::make()` should be called only during training run when CTD is set: https://github.com/openjdk/jdk/blob/master/src/hotspot/share/compiler/compileBroker.cpp#L356C21-L356C30
> 
> Am I missing path where it is called in assembly phase or production?
> 
> `need_data()` is true only during training:
> https://github.com/openjdk/jdk/blob/master/src/hotspot/share/oops/trainingData.hpp#L290

During assembly we have `have_data()` return true, and since we're running some java code this will make us cache MTDs in MCs. That's what I think this code is trying to avoid.

But I have a question though. Why do we treat the presence of the cache MTD value in MC as a proof that MTD refers to a MC?

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

PR Review Comment: https://git.openjdk.org/jdk/pull/28670#discussion_r2692024222