RFR: Small fixes - shipment one
Coleen Phillimore
coleenp at openjdk.java.net
Tue Mar 22 15:41:53 UTC 2022
On Tue, 22 Mar 2022 15:26:14 GMT, Robbin Ehn <rehn at openjdk.org> wrote:
>> src/hotspot/share/runtime/continuation.cpp line 1213:
>>
>>> 1211:
>>> 1212: chunk = allocate_chunk(cont_size + ContinuationHelper::frame_metadata);
>>> 1213: if (UNLIKELY(chunk == nullptr || !_thread->cont_fastpath() || _barriers)) { // OOME/probably humongous
>>
>> This is just a strange set of conditions. Why would you allocate_chunk if _thread->cont_fastpath() is false? or _barriers, whatever that means. Really don't expect a function like allocate_chunk to have side effects.
>> Maybe in the latter two conditions, you still want to zero out the chunk in the tail and don't want to remove this.
>
> The variable _barriers is set by allocate_chunk, so it must be check afterwards.
> If allocation safepoints and there is JVM TI qeury it can 'remove' us as available for fastpath.
> If allocation fails due to OOME.
>
> The init_chunk() should always zero enought.
Ok, thanks for confirming that it's already zeroed.
-------------
PR: https://git.openjdk.java.net/loom/pull/114
More information about the loom-dev
mailing list