Add information regarding synchronous compilations in javacore by dev-koan · Pull Request #24236 · eclipse-openj9/openj9

dev-koan · 2026-06-25T14:35:00Z

Added a struct in J9JITConfig to store information regarding synchronous compilations and print them in javacore file.

Signed-off-by: Dev Agarwal dev.agarwal@ibm.com

mpirvu

We have one decision to make: do we want to measure the times needed for synchronous compilations to be performed or the waiting time of application threads? there could be several application threads waiting for the same compilation to finish.

mpirvu

LGTM. Please post an example of the output.

mpirvu · 2026-07-06T19:15:18Z

Also, please squash the 4 commits into 1.

dev-koan · 2026-07-06T19:17:57Z

LGTM. Please post an example of the output.

amicic · 2026-07-08T15:07:57Z

We have one decision to make: do we want to measure the times needed for synchronous compilations to be performed or the waiting time of application threads? there could be several application threads waiting for the same compilation to finish.

Ideally, i'd like to see each app thread maintain its own longest wait time (and associated info). For brevity, only threads that have wait times longer than a certain threshold (like 3ms or 10ms) should be reported it to javacore.

mpirvu · 2026-07-08T15:17:11Z

If each thread maintains the name of the sync method, with long method names and many threads there could be some footprint increase. I like the idea of applying some filtering and not include quick compilations.

mpirvu · 2026-07-08T20:35:16Z

@amicic To clarify: each app thread that is subject to a wait due to a synchronous compilation will have a dynamically allocated string in J9VMThread struct (to hold the name of the method it waited for). This string needs to be freed when the thread is destroyed.
Caveat: if the thread gets destroyed before the javacore generation happen, the wait info will be lost.

mpirvu

LGTM. Please post an example of the output in the description of this PR.

mpirvu · 2026-07-20T14:32:10Z

+typedef struct J9JITSyncCompilationStatistics {
+	uint32_t totalCount;
+    uint64_t totalWaitTime; // microseconds (us)
+	J9JITLongestSyncComp longestWaitMethods[J9_LONGEST_SYNC_COMP];


Please add a comment showing how this array is sorted.

mpirvu · 2026-07-21T18:03:59Z

@amicic If you approve it, I am going to start testing and merge. Thanks

amicic · 2026-07-21T18:11:28Z

+			_OutputStream.writeCharacters("2nd Longest");
+		}
+		else {
+			_OutputStream.writeCharacters("3rd Longest");


Not quite the most robust (in case J9_LONGEST_SYNC_COMP increases a bit). you could simplify and just print numbers 1. longest, 2. longest

Just an observation. If you still prefer this way, i'm fine with it.

Updated this.

Formatting: } else ....

Added a struct in J9JITConfig to store information regarding synchronous compilations and print them in javacore file. Signed-off-by: Dev Agarwal <dev.agarwal@ibm.com>

keithc-ca · 2026-07-21T19:56:59Z

+	for (int i = 0; i < J9_LONGEST_SYNC_COMP; i++) {
+		stats->longestWaitMethods[i].waitTime = 0;


This seems racy: other threads may be active.

keithc-ca · 2026-07-21T20:02:00Z

Please post an example of the output in the description

Can we have a text version of that sample (as opposed to an image)?

dev-koan · 2026-07-23T18:53:18Z

Please post an example of the output in the description

Can we have a text version of that sample (as opposed to an image)?

NULL           ------------------------------------------------------------------------
0SECTION       Synchronous compilations info dump routine
NULL           ==============================
1NOTE          This data is reset after each javacore file is written
NULL           ------------------------------------------------------------------------
1JITSYNCSTATS  Synchronous Compilation Statistics
NULL           ------------------------------------------------------------------------
2JITSYNCCOUNT  Total synchronous compilations: 793
2JITTOTALWAIT  Total application wait time: 1239420us
2JITLONGEST    1. Longest Synchronous Compilation
3JITWAITTIME   Wait time: 29445us
3JITENDTIME    Wait time end: 2026-07-23T14:52:20.902
3JITMETHOD     Method: jdk/internal/classfile/impl/StackMapGenerator.processBlock(Ljdk/internal/classfile/impl/RawBytecodeHelper;)Z
3JITTHREAD     Thread: (unnamed thread)
2JITLONGEST    2. Longest Synchronous Compilation
3JITWAITTIME   Wait time: 18123us
3JITENDTIME    Wait time end: 2026-07-23T14:52:22.000
3JITMETHOD     Method: java/lang/Class.getEnumConstantsShared()[Ljava/lang/Object;
3JITTHREAD     Thread: main
2JITLONGEST    3. Longest Synchronous Compilation
3JITWAITTIME   Wait time: 18063us
3JITENDTIME    Wait time end: 2026-07-23T14:52:22.018
3JITMETHOD     Method: java/lang/Access.getEnumConstantsShared(Ljava/lang/Class;)[Ljava/lang/Enum;
3JITTHREAD     Thread: main

keithc-ca · 2026-07-23T19:48:38Z

 	IDATA verboseOutputLevel;
 	omrthread_monitor_t compilationMonitor;
 	void* compilationInfo;
+	J9JITSyncCompilationStatistics *syncCompStats;


Why need this be indirect? Why not:

J9JITSyncCompilationStatistics syncCompStats;

It is simpler to use a pointer here to make sure the same memory is being accessed everywhere (in compilationthread class and javadump class)
I can also use a reference (&) to syncCompStats in javadump and compilationthread if that is preferred.

keithc-ca · 2026-07-24T20:11:24Z


    struct CompilationStatistics _stats;
    struct CompilationStatsPerInterval _intervalStats;
+    SyncCompilationStatistics _syncCompStats;


This should instead live in J9JITConfig.

keithc-ca · 2026-07-24T20:11:38Z

 * Assisted-by: IBM Bob
 *******************************************************************************/

+#include <cstdint>


I don't think this is necessary.

keithc-ca · 2026-07-24T20:13:32Z

+            _syncCompStats.totalWaitTime += duration;
+
+            /* Obtain index to insert. */
+            int idx = -1;


This should use the same type as i.

keithc-ca · 2026-07-24T20:14:29Z

+typedef struct J9JITSyncCompilationStatistics {
+	uint32_t totalCount;
+	uint64_t totalWaitTime; /* microseconds */
+	J9JITLongestSyncComp longestWaitMethods[J9_NUM_LONGEST_SYNC_COMP]; /* sorted in increasing order */


I think you mean "decreasing" (the first entry has the longest wait time).

keithc-ca · 2026-07-24T20:15:34Z

 *******************************************************************************/

 /* Includes */
+#include <cstdint>


I don't think this is necessary.

keithc-ca · 2026-07-24T20:16:37Z

+	_OutputStream.writeCharacters("2JITTOTALWAIT  Total application wait time: ");
+	_OutputStream.writeInteger64(stats->totalWaitTime, "%llu");
+	_OutputStream.writeCharacters("us\n");
+	for (int32_t i = 0; i < J9_NUM_LONGEST_SYNC_COMP; i++) {


Suggest using UDATA instead of int32_t, in part so the format on line 2953 is correct, and for consistency with other loops in this file.

keithc-ca · 2026-07-24T20:18:36Z

+
+		_OutputStream.writeCharacters("3JITENDTIME    Wait time end: ");
+		omrstr_ftime_ex(timeStamp, _MaximumTimeStampLength, "%Y-%m-%dT%H:%M:%S",
+						longestWaitMethod.waitTimeEnd, OMRSTR_FTIME_FLAG_LOCAL);


Excessive indentation: Two tabs more than line 2961 is sufficient.

dev-koan force-pushed the feature/SyncCompInfo branch 3 times, most recently from fbec998 to 2d33ba0 Compare July 2, 2026 14:00

mpirvu self-assigned this Jul 2, 2026

mpirvu added comp:jit comp:vm labels Jul 2, 2026

mpirvu requested changes Jul 2, 2026

View reviewed changes

dev-koan force-pushed the feature/SyncCompInfo branch 2 times, most recently from 43f338e to 177844e Compare July 3, 2026 19:21

dev-koan requested a review from mpirvu July 3, 2026 19:22

mpirvu requested changes Jul 3, 2026

View reviewed changes

Comment thread runtime/compiler/control/CompilationThread.cpp Outdated

Comment thread runtime/compiler/control/CompilationThread.cpp Outdated

Comment thread runtime/oti/j9nonbuilder.h Outdated

dev-koan force-pushed the feature/SyncCompInfo branch from 543fdd6 to 0825011 Compare July 6, 2026 17:23

dev-koan requested a review from mpirvu July 6, 2026 17:26

dev-koan force-pushed the feature/SyncCompInfo branch from 0825011 to 3ffce49 Compare July 6, 2026 17:46

mpirvu reviewed Jul 6, 2026

View reviewed changes

Comment thread runtime/compiler/control/CompilationThread.cpp Outdated

Comment thread runtime/compiler/control/CompilationThread.cpp Outdated

Comment thread runtime/oti/j9nonbuilder.h Outdated

dev-koan requested a review from mpirvu July 6, 2026 18:59

mpirvu approved these changes Jul 6, 2026

View reviewed changes

dev-koan force-pushed the feature/SyncCompInfo branch from 3ffce49 to 75bbae1 Compare July 6, 2026 19:18

dev-koan marked this pull request as ready for review July 6, 2026 19:19

amicic added the comp:ras label Jul 8, 2026

dev-koan force-pushed the feature/SyncCompInfo branch from 75bbae1 to 286bba4 Compare July 14, 2026 15:42

dev-koan requested a review from dsouzai as a code owner July 14, 2026 15:42

dev-koan force-pushed the feature/SyncCompInfo branch 2 times, most recently from ee2ee8c to 1ca1fb1 Compare July 14, 2026 15:48

dev-koan requested a review from mpirvu July 14, 2026 15:51

mpirvu approved these changes Jul 20, 2026

View reviewed changes

dev-koan force-pushed the feature/SyncCompInfo branch from 1ca1fb1 to a65babe Compare July 20, 2026 20:52

mpirvu requested review from amicic and removed request for dsouzai July 21, 2026 18:03

amicic reviewed Jul 21, 2026

View reviewed changes

Comment thread runtime/oti/j9nonbuilder.h Outdated

amicic reviewed Jul 21, 2026

View reviewed changes

amicic approved these changes Jul 21, 2026

View reviewed changes

keithc-ca self-requested a review July 21, 2026 19:28

dev-koan force-pushed the feature/SyncCompInfo branch from a65babe to 97d5371 Compare July 21, 2026 19:28

keithc-ca reviewed Jul 21, 2026

View reviewed changes

Comment thread runtime/oti/j9nonbuilder.h Outdated

Add information regarding synchronous compilations in javacore

88e217d

Added a struct in J9JITConfig to store information regarding synchronous compilations and print them in javacore file. Signed-off-by: Dev Agarwal <dev.agarwal@ibm.com>

dev-koan force-pushed the feature/SyncCompInfo branch from 97d5371 to 88e217d Compare July 21, 2026 19:53

dev-koan requested a review from keithc-ca July 21, 2026 19:55

keithc-ca requested changes Jul 21, 2026

View reviewed changes

keithc-ca reviewed Jul 21, 2026

View reviewed changes

Comment thread runtime/oti/j9nonbuilder.h Outdated

keithc-ca reviewed Jul 21, 2026

View reviewed changes

Comment thread runtime/oti/j9nonbuilder.h Outdated

recent review updates

5052e90

dev-koan force-pushed the feature/SyncCompInfo branch from 2376107 to 5052e90 Compare July 23, 2026 19:05

keithc-ca requested changes Jul 23, 2026

View reviewed changes

review update

a89e375

keithc-ca requested changes Jul 24, 2026

View reviewed changes

review updates

a3b6534

		for (int i = 0; i < J9_LONGEST_SYNC_COMP; i++) {
		stats->longestWaitMethods[i].waitTime = 0;

Uh oh!

Conversation

dev-koan commented Jun 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mpirvu 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!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

mpirvu left a comment

Choose a reason for hiding this comment

Uh oh!

mpirvu commented Jul 6, 2026

Uh oh!

dev-koan commented Jul 6, 2026

Uh oh!

amicic commented Jul 8, 2026

Uh oh!

mpirvu commented Jul 8, 2026

Uh oh!

mpirvu commented Jul 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mpirvu left a comment

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!

mpirvu commented Jul 21, 2026

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!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

keithc-ca commented Jul 21, 2026

Uh oh!

Uh oh!

Uh oh!

dev-koan commented Jul 23, 2026

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

dev-koan commented Jun 25, 2026 •

edited

Loading

mpirvu commented Jul 8, 2026 •

edited

Loading