Add support for Nursery only compact by shadman2606 · Pull Request #8248 · eclipse-omr/omr

shadman2606 · 2026-05-11T18:29:42Z

Add support for Nursery only compact

add a flag to control if whole heap compacted or Nursery only
use the fixup_only subarea state to skip moving Tenure objects

Issue: eclipse-openj9/openj9#20515

Signed-off-by: Shadman Siddiqui shadman2606@gmail.com

amicic · 2026-05-12T21:22:37Z

@@ -48,6 +48,7 @@ class MM_ParallelCompactTask : public MM_ParallelTask
 	MM_CompactScheme *_compactScheme;
 	bool _rebuildMarkBits;
 	bool _aggressive;
+	bool nurseryOnly;


members should start with '_'

amicic · 2026-05-12T21:25:11Z

+	if (CompactReason::COMPACT_ABORTED_SCAVENGE == compactStats->_compactReason)
+	{
+		nurseryOnly = true;
+	}


you can collapse it into one line:

bool nurseryOnly = (CompactReason::COMPACT_ABORTED_SCAVENGE == compactStats->_compactReason);

amicic · 2026-05-12T21:30:28Z

+	{
+		nurseryOnly = true;
+	}
+	MM_ParallelCompactTask compactTask(env, _dispatcher, _compactScheme, rebuildMarkBits, env->_cycleState->_gcCode.shouldAggressivelyCompact(), nurseryOnly);


Honestly, don't know why rebuildMarkBits and aggressive (and now nurseryOnly) are passed to the CompactTask, rather than CompactScheme.

If they were part of CompactScheme then we would not need to pass them later through various methods via params, instead of just fetching them from CompactScheme when needed.

Perhaps we can fix/simplify that later. For now, let's keep the focus on minimal changes.

amicic · 2026-05-20T15:27:22Z

+			if(nurseryOnly && MEMORY_TYPE_OLD == memorySubSpace->getTypeFlags())
+			{
+				state = SubAreaEntry::fixup_only;
+			}


Suggested change

}

if (nurseryOnly && (MEMORY_TYPE_OLD == memorySubSpace->getTypeFlags())) {

state = SubAreaEntry::fixup_only;

}

fix the formatting for other similar spots accordingly

add extra oval brackets

curly bracket is up one line

fully correct check is actually

MEMORY_TYPE_OLD == (getTypeFlags() & MEMORY_TYPE_OLD))

since other bit-wise flags could be set too (like MEMORY_TYPE_RAM)

amicic · 2026-05-20T16:18:24Z

 	        	if (changeSubAreaAction(env, &subAreaTable[i], SubAreaEntry::fixing_heap_for_walk)) {
 	        		omrobjectptr_t start = subAreaTable[i].firstObject;
-					omrobjectptr_t end   = subAreaTable[i].firstObject;
+					omrobjectptr_t end   = subAreaTable[i+1].firstObject;


we typically provide extra space around binary operators

[i + 1]

amicic · 2026-05-20T16:57:48Z


 		while(NULL != (memoryPool = poolIterator.nextPool())) {
-			memoryPool->postProcess(env, MM_MemoryPool::forCompact);
+			if(!nurseryOnly || (MEMORY_TYPE_OLD != (memoryPool->getSubSpace()->getTypeFlags() & MEMORY_TYPE_OLD))) {


there should be space between keywords if, while, for,... and opening oval bracket

fix anywhere in the code you updated or even in vicinity of those that you have not updated

amicic · 2026-05-20T17:00:44Z

 	MM_CompactScheme *_compactScheme;
 	bool _rebuildMarkBits;
 	bool _aggressive;
+	bool _nurseryOnly;


add comment
/**< if true, Nursery is only compacted and Tenure fixed-up via RS, otherwise whole heap is compacted/fixed-up */

amicic · 2026-05-20T17:01:21Z

 	compactStats->_startTime = omrtime_hires_clock();
-	MM_ParallelCompactTask compactTask(env, _dispatcher, _compactScheme, rebuildMarkBits, env->_cycleState->_gcCode.shouldAggressivelyCompact());
+	bool nurseryOnly = false;
+	nurseryOnly = (CompactReason::COMPACT_ABORTED_SCAVENGE == compactStats->_compactReason);


could be one line:

bool nurseryOnly = (CompactReason::COMPACT_ABORTED_SCAVENGE == compactStats->_compactReason);

- add a flag to control if whole heap compacted or Nursery only - use the fixup_only subarea state to skip moving Tenure objects Issue: eclipse-openj9/openj9#20515 Signed-off-by: Shadman Siddiqui shadman2606@gmail.com

amicic added the comp:gc label May 12, 2026

amicic reviewed May 12, 2026

View reviewed changes

shadman2606 force-pushed the slide-compact branch 2 times, most recently from 6c15ce7 to a4b04f0 Compare May 14, 2026 15:51

amicic reviewed May 20, 2026

View reviewed changes

amicic suggested changes May 20, 2026

View reviewed changes

amicic reviewed May 20, 2026

View reviewed changes

shadman2606 force-pushed the slide-compact branch from a4b04f0 to c8f8040 Compare May 20, 2026 16:52

shadman2606 changed the title ~~Optimize sliding compact for tenure after aborted scavenge~~ Add support for Nursery only compact May 20, 2026

amicic reviewed May 20, 2026

View reviewed changes

amicic suggested changes May 20, 2026

View reviewed changes

shadman2606 force-pushed the slide-compact branch 2 times, most recently from 1820145 to 7cf582c Compare May 22, 2026 17:45

Add support for Nursery only compact

2b54835

- add a flag to control if whole heap compacted or Nursery only - use the fixup_only subarea state to skip moving Tenure objects Issue: eclipse-openj9/openj9#20515 Signed-off-by: Shadman Siddiqui shadman2606@gmail.com

shadman2606 force-pushed the slide-compact branch from ecab2c7 to 2b54835 Compare June 9, 2026 10:11

Uh oh!

Conversation

shadman2606 commented May 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

amicic May 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

amicic May 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

shadman2606 commented May 11, 2026 •

edited

Loading

amicic May 20, 2026 •

edited

Loading

amicic May 20, 2026 •

edited

Loading