✨ อัปเดตล่าสุด: eclipse openj/openj/pull/

AditiS11 · 2026-03-12T16:26:58Z

Remove valueHashCode function
Add implementation for convertValueObjectAtOffsetToHash to iteratively compute hash for value objects using field values.
Add function inlineConvertObjectToHash to handle hash computation for both value objects and identity objects.
Update inlineObjectHashCode to handle hash calculation for value objects.

GC changes:

Update GC to skip value type hash computation during copy operations.
Hash calculations should be triggered by the VM when it reads the hash slot and discovers 0 in the slot.
Add wrapper functions convertValueToHashForObject and computeObjectAddressToHashForObject to return 0 for value types during copy operations.
Add convertObjectToHash as GC-side wrapper to handle hash calculation for both value and identity objects.

Test changes:

Add tests to verify hash code behaviour for empty and similar value classes.
Re-enable tests related to hash for value objects.

Fixes: #15768

This change follows the approach described in 15768(comment)

AditiS11 · 2026-03-12T16:28:48Z

With these changes the following OpenJDK tests are passing:

valhalla/valuetypes/MethodHandleTest.java
valhalla/valuetypes/StreamTest.java
valhalla/valuetypes/RecursiveValueClass.java
runtime/valhalla/inlinetypes/TestCloneableValue.java
valhalla/valuetypes/ValueObjectMethodsTest.java
runtime/valhalla/inlinetypes/ObjectMethods.java#compressed-class-pointers
runtime/valhalla/inlinetypes/ObjectMethods.java#no-verify
runtime/valhalla/inlinetypes/ObjectMethods.java#no-compressed-class-pointers

Passing Test:
https://hyc-runtimes-jenkins.swg-devops.com/view/Test_grinder/job/Grinder/60415/testReport/
https://hyc-runtimes-jenkins.swg-devops.com/view/Test_grinder/job/Grinder/60414/testReport/

theresa-m

First pass.

theresa-m · 2026-03-13T14:59:15Z

 private:
 protected:
 public:
+	struct ValueTypeHashQueueEntry {


This struct and the new helper methods can be under private in this file.

I've encountered compilers that don't handle private types properly in all situations.

theresa-m

What is the behavior of the ri if hashCode is called prior to all strict instance fields being set in ? The easiest way to test this out is to use the @NullRestricted annotation which will mark the instance field as strict.

theresa-m · 2026-03-18T18:20:24Z

+						break;
+					}
+					default:
+						  break;


This should result in an error.

I initially thought of using Assert_VM_unreachable(), but I think it can't be used here.
Could you please suggest on what error to throw here.

Use Assert_VM_unreachable() (or similar) seems reasonable; reaching line 324 means the signature is invalid - which should have been detected long before we get here.

I think you can add a comment like https://github.com/eclipse-openj9/openj9/blob/master/runtime/oti/ObjectHash.hpp#L121 instead. edit: didn't see Keith's answer before posting this.

keithc-ca · 2026-03-19T17:41:22Z

 private:
 protected:
 public:
+	struct ValueTypeHashQueueEntry {


I've encountered compilers that don't handle private types properly in all situations.

AditiS11 · 2026-03-20T18:22:18Z

What is the behavior of the ri if hashCode is called prior to all strict instance fields being set in ? The easiest way to test this out is to use the @NullRestricted annotation which will mark the instance field as strict.

It throws a compiler error: reference to hashCode() may only appear after an explicit constructor invocation

theresa-m

I suggest summarizing how hashcodes are updated for value types or link #15768 (comment) in the description so it is easy to find.

dmitripivkine · 2026-03-26T15:10:28Z

Sorry, I probably missed something, I was under impression that during VT hash code calculation if visited sub-object has hash code stored in the slot already, this code can be used instead of walking this object again. Is it right? If so, where it is implemented in the code? Just want to double check.

AditiS11 · 2026-03-26T17:52:14Z

Sorry, I probably missed something, I was under impression that during VT hash code calculation if visited sub-object has hash code stored in the slot already, this code can be used instead of walking this object again. Is it right? If so, where it is implemented in the code? Just want to double check.

https://github.com/eclipse-openj9/openj9/pull/23502/changes#diff-9aa50703bc7320e2410d7093b1f619ded056a3aeed5f75e2477603d7a5772712R269-R329
The current implementation relies on inlineObjectHashCode() for reference fields. This helper checks whether the object already has a hash stored in its hash slot and returns it if present; otherwise it computes and stores a new one.

For value types, the hash slot is not checked, and the current implementation walks through all the fields to compute the hash.

AditiS11 · 2026-05-28T12:37:11Z

Sorry about the delay. I have updated the changes.

dmitripivkine · 2026-05-28T14:40:54Z


-			uint32_t *hashCodePointer = (uint32_t*)((uint8_t*) destinationObjectPtr + hashOffset);
-			*hashCodePointer = convertValueToHash(_javaVM, (uintptr_t)forwardedHeader->getObject());
+			int32_t *hashCodePointer = (int32_t*)((uint8_t*)destinationObjectPtr + hashOffset);


If you are going to touch code after this point please add spaces before stars:
int32_t *hashCodePointer = (int32_t *)((uint8_t *)destinationObjectPtr + hashOffset);

keithc-ca

Please squash and rebase.

keithc-ca · 2026-06-02T18:56:52Z

Please update the commit message to spell "hash code" consistently (two words), punctuate consistently, and ensure that no line in the body is more than 72 characters long.

- Remove valueHashCode function. - Add implementation for convertValueObjectAtOffsetToHash to iteratively compute hash codes for value objects using field values. - Add function inlineConvertObjectToHash to handle hash computation for both value objects and identity objects. - Update inlineObjectHashCode to handle hash calculation for value objects. GC changes: - Update GC to skip value type hash computation during copy operations. - Hash calculations should be triggered by the VM when it reads the hash slot and discovers 0 in the slot. - Add wrapper functions convertValueToHashForObject and computeObjectAddressToHashForObject to return 0 for value types during copy operations. - Add convertObjectToHash as a GC-side wrapper to handle hash calculation for both value and identity objects. Test changes: - Add tests to verify hash code behaviour for empty and similar value classes. - Re-enable tests related to hash codes for value objects. Fixes: eclipse-openj9#15768 Signed-off-by: Aditi Srinivas M <Aditi.Srini@ibm.com>

keithc-ca · 2026-06-03T13:51:36Z

The commit message is much better. Thank you.

keithc-ca · 2026-06-03T14:02:16Z

Jenkins test sanity,extended aix,amac,plinux,win jdk25

hangshao0 · 2026-06-03T14:07:38Z

Jenkins test sanity.functional,extended zlinuxval jdknext

hangshao0 · 2026-06-03T14:08:20Z

jenkins compile win,win32 jdk8

keithc-ca · 2026-06-03T14:39:29Z

+	 * @return finalized hash value
+	 */
+	static VMINLINE U_32
+	finalizeMurmur3Hash(U_32 hashValue, U_32 numBytesHashed)


DDR should be updated to match these changes; see ObjectHash.

I created #24053 to track that (future) work.

keithc-ca · 2026-06-04T01:21:57Z

The one test failure appears to be a known issue: #21871.

theresa-m self-requested a review March 12, 2026 16:30

theresa-m added comp:vm project:valhalla Used to track Project Valhalla related work labels Mar 12, 2026

AditiS11 force-pushed the valuetypeHashCode branch 2 times, most recently from 29d6177 to 6f419cf Compare March 13, 2026 11:42

theresa-m suggested changes Mar 13, 2026

View reviewed changes

AditiS11 marked this pull request as draft March 16, 2026 12:06

AditiS11 force-pushed the valuetypeHashCode branch 3 times, most recently from 59a45e3 to 437659e Compare March 18, 2026 11:17

AditiS11 marked this pull request as ready for review March 18, 2026 12:22

AditiS11 mentioned this pull request Mar 18, 2026

Value Types: valhalla/valuetypes/MethodHandleTest.java #22648

Closed

theresa-m suggested changes Mar 19, 2026

View reviewed changes

keithc-ca requested changes Mar 19, 2026

View reviewed changes

AditiS11 force-pushed the valuetypeHashCode branch from 437659e to c5f3c7e Compare March 20, 2026 18:20

AditiS11 force-pushed the valuetypeHashCode branch 2 times, most recently from 75c5db8 to 9aeaac9 Compare March 23, 2026 12:23

theresa-m reviewed Mar 23, 2026

View reviewed changes

Comment thread runtime/vm/BytecodeInterpreter.hpp Outdated

Comment thread runtime/oti/ObjectHash.hpp

AditiS11 mentioned this pull request Mar 23, 2026

Implement arrayInstanceIndexScale() and arrayInstanceBaseOffset() #23565

Merged

AditiS11 force-pushed the valuetypeHashCode branch 2 times, most recently from 168ca66 to 111a450 Compare March 23, 2026 18:13

keithc-ca requested changes Mar 23, 2026

View reviewed changes

dmitripivkine reviewed Mar 23, 2026

View reviewed changes

Comment thread runtime/gc_glue_java/ObjectModel.hpp Outdated

AditiS11 force-pushed the valuetypeHashCode branch from 111a450 to 01beea1 Compare March 24, 2026 12:37

dmitripivkine reviewed Mar 24, 2026

View reviewed changes

Comment thread runtime/gc_glue_java/ObjectModel.hpp Outdated

AditiS11 marked this pull request as draft March 25, 2026 11:37

dmitripivkine reviewed May 28, 2026

View reviewed changes

keithc-ca requested changes May 28, 2026

View reviewed changes

Comment thread runtime/oti/ObjectHash.hpp Outdated

AditiS11 force-pushed the valuetypeHashCode branch from ef15912 to 362e893 Compare June 1, 2026 10:53

keithc-ca requested changes Jun 1, 2026

View reviewed changes

AditiS11 mentioned this pull request Jun 1, 2026

Prevent value types from being used as monitors #24004

Merged

AditiS11 force-pushed the valuetypeHashCode branch 2 times, most recently from 65e7666 to fb05683 Compare June 2, 2026 12:37

keithc-ca requested changes Jun 2, 2026

View reviewed changes

AditiS11 force-pushed the valuetypeHashCode branch from fb05683 to 4f1b712 Compare June 2, 2026 15:44

keithc-ca changed the title ~~Implement native method to calculate hashcode for valuetypes~~ Implement native method to calculate hash code for valuetypes Jun 2, 2026

AditiS11 force-pushed the valuetypeHashCode branch from 4f1b712 to f59105c Compare June 3, 2026 12:45

hangshao0 added this to Valhalla L-World Jun 3, 2026

github-project-automation Bot moved this to TODO: VM in Valhalla L-World Jun 3, 2026

github-project-automation Bot moved this from TODO: VM to Needs review in Valhalla L-World Jun 3, 2026

keithc-ca reviewed Jun 3, 2026

View reviewed changes

keithc-ca mentioned this pull request Jun 3, 2026

Update hash code functions in DDR #24053

Closed

keithc-ca approved these changes Jun 4, 2026

View reviewed changes

github-project-automation Bot moved this from Needs review to Reviewer approved in Valhalla L-World Jun 4, 2026

keithc-ca merged commit 0490c4c into eclipse-openj9:master Jun 4, 2026
23 of 25 checks passed

AditiS11 mentioned this pull request Jun 4, 2026

Re-enable tests related to hash code for value objects adoptium/aqa-tests#7151

Merged

This was referenced Jun 5, 2026

Valhalla valhalla/valuetypes/ValueObjectMethodsTest.java #24087

Closed

Valhalla runtime/valhalla/inlinetypes/InlineTypesTest.java#force-non-tearable.InlineTypesTest_force-non-tearable #24089

Closed

Uh oh!

Conversation

AditiS11 commented Mar 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

AditiS11 commented Mar 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

theresa-m left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

theresa-m left a comment • 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!

Uh oh!

Uh oh!

Uh oh!

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!

theresa-m Mar 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!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

AditiS11 commented Mar 20, 2026

Uh oh!

theresa-m left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dmitripivkine commented Mar 26, 2026

Uh oh!

AditiS11 commented Mar 26, 2026

Uh oh!

AditiS11 commented May 28, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

AditiS11 commented Mar 12, 2026 •

edited

Loading

AditiS11 commented Mar 12, 2026 •

edited

Loading

theresa-m left a comment •

edited

Loading

theresa-m Mar 20, 2026 •

edited

Loading

keithc-ca commented Jun 2, 2026 •

edited

Loading