stan/jdk-24
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

8309960: ParallelGC young collections very slow in DelayInducer

Browse Source 
Reviewed-by: tschatzl, iwalulya
This commit is contained in:
Albert Mingkun Yang 2023-06-15 12:01:02 +00:00
parent 83d92672d4
commit 5f3613e060
1 changed files with 8 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
src/hotspot/share/gc/parallel/psPromotionManager.cpp
View File

@ -237,9 +237,14 @@ void PSPromotionManager::drain_stacks_depth(bool totally_drain) {
// Drain overflow stack first, so other threads can steal from // Drain overflow stack first, so other threads can steal from
// claimed stack while we work. // claimed stack while we work.
while (tq->pop_overflow(task)) { while (tq->pop_overflow(task)) {
if (!tq->try_push_to_taskqueue(task)) { // In PSCardTable::scavenge_contents_parallel(), when work is distributed
process_popped_location_depth(task); // among different workers, an object is never split between multiple workers.
} // Therefore, if a worker gets owned a large objArray, it may accumulate
// many tasks (corresponding to every element in this array) in its
// task queue. When there are too many overflow tasks, publishing them
// (via try_push_to_taskqueue()) can incur noticeable overhead in Young GC
// pause, so it is better to process them locally until large-objArray-splitting is implemented.
process_popped_location_depth(task);
} }
while (tq->pop_local(task, threshold)) { while (tq->pop_local(task, threshold)) {