8313713: Allow CompileCommand flag to specify compilation level by kirill-shirokov · Pull Request #30352 · openjdk/jdk

kirill-shirokov · 2026-03-21T04:29:44Z

Implemented intx arguments in -XX:CompileCommand={break,compileonly,exclude,print} commands that allow to select specific compilation levels.

E.g. -XX:CompileCommand=compileonly,java/lang/Object::wait,9 would allow to compile the wait() method only on CompLevel 1 and 4 (1 << (4 - 1) | 1 << (1 - 1) == 9).

CompLevel	-XX:CompileCommand bitmask
0 (interpreter)	not applicable
1 (C1)	1
2 (C1 + counters)	2
3 (C1 + counters + mdo)	4
4 (C2/JVMCI)	8
All C1 levels	7
All levels	15

For compatibility with the old format, specifying the commands without intx argument is also supported. In this case all levels will be covered (using a default value of 15). Specifying boolean values is not supported anymore.

New tests are provided and I ran them manually. The tests are not included into any tiers yet.

Some implementation choices I made (and I open to discussing/modifying them):

intx for the value, not uintx as most flags use intx
I tried to avoid any changes to compilation level state machine. The only modification I did is to avoid downgrading level from 3 to 2 if level 2 is excluded in CompilationPolicy::select_task(), see line 817+
I refactored direct freeing CompileTask::_directive pointer in CompileBroker::compile_method(), line 2405 via a new CompileTask method (this approach conforms to OO encapsulation principles more)
JVMCI CodeInstaller::install() method processes compilation level like this:
DirectiveSet* directive = DirectivesStack::getMatchingDirective(method, compiler == nullptr || compiler->is_c1() ? CompLevel_simple : CompLevel_full_optimization);
Please let me know if this needs changing.
Same for C2V_VMENTRY_0(jboolean, hasNeverInlineDirective,(JNIEnv* env, jobject, ARGUMENT_PAIR(method))):
return !Inline || CompilerOracle::should_not_inline(method, CompLevel_full_optimization) || method->dont_inline();
(level 4 is assumed, please let me know if this has to be changed)

Verified by running tier1 tests on linux + macosx aarch64.

Progress

Change must be properly reviewed (1 review required, with at least 1 Reviewer)
Change must not contain extraneous whitespace
Commit message must refer to an issue

Issue

JDK-8313713: Allow CompileCommand flag to specify compilation level (Enhancement - P4)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/30352/head:pull/30352
$ git checkout pull/30352

Update a local copy of the PR:
$ git checkout pull/30352
$ git pull https://git.openjdk.org/jdk.git pull/30352/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 30352

View PR using the GUI difftool:
$ git pr show -t 30352

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/30352.diff

Using Webrev

Link to Webrev Comment

…ial draft implementation

…uld_not_inline(). Replace compiler arument with comp_level in DirectivesStack::getMatchingDirective()

…. Fix level checking in WB_IsMethodCompilable()

…leCommands

…itmask

…n ClearDirectivesFileStackTest. Sort includes. Update copyright year.

…evel is excluded for the method. Fix WB_IsMethodCompilable for levels 1-3

…mpileonly and exclude CompileCommands with a level argument

…tput without WhiteBox

…o communicate with testee via stdin

…se added to CompileLevelParseTest.java. CompileTask::release_direcive() added for ownership clarity. Null pointer crash fixed in CompilationPolicy::select_task()

bridgekeeper · 2026-03-21T04:30:50Z

👋 Welcome back kshiroko! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2026-03-21T04:30:56Z

❗ This change is not yet ready to be integrated.
See the Progress checklist in the description for automated requirements.

openjdk · 2026-03-21T04:32:03Z

@kirill-shirokov The following label will be automatically applied to this pull request:

hotspot

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing list. If you would like to change these labels, use the /label pull request command.

mlbridge · 2026-03-21T04:41:34Z

Webrevs

01: Full - Incremental (79407adb)
00: Full (cbd30672)

vnkozlov

Nice work, thank you. I have few comments.

vnkozlov · 2026-03-21T18:06:52Z

src/hotspot/share/jvmci/jvmciCodeInstaller.cpp

      if (compile_state == nullptr) {
        // This compile didn't come through the CompileBroker so perform the printing here
-        DirectiveSet* directive = DirectivesStack::getMatchingDirective(method, compiler);
+        DirectiveSet* directive = DirectivesStack::getMatchingDirective(method, compiler == nullptr || compiler->is_c1() ? CompLevel_simple : CompLevel_full_optimization);


At this point compiler should be JVMCI compiler and not NULL. It used in nmethod::new_nmethod() during registration. So, just use CompLevel_full_optimization here.

vnkozlov · 2026-03-21T18:08:40Z

src/hotspot/share/compiler/compilerOracle.hpp

  option(Quiet, "quiet", Unknown) \
  option(Log, "log", Bool) \
-  option(Print, "print", Bool) \
+  option(Print, "print", Intx) \


For an other RFE. intx is 64 bits in 64-bits VM we have now. We should introduce normal int for small values like you added in these changes.

vnkozlov · 2026-03-21T18:13:36Z

src/hotspot/share/compiler/compilationPolicy.cpp

+  if (max_task != nullptr && max_method != nullptr) {
+    methodHandle max_method_h(THREAD, max_method);
+    DirectiveSet * directive = DirectivesStack::getMatchingDirective(max_method_h, CompLevel_limited_profile);


Is it possible to use and update task->directive() here? Because max_task->set_comp_level(CompLevel_limited_profile) will not match it - it was created for CompLevel_full_profile.

vnkozlov · 2026-03-21T18:38:20Z

src/hotspot/share/compiler/compilerOracle.cpp

+// All C1 levels           | 7
+// All levels              | 15
+inline int comp_level_bitmask(int comp_level) {
+  assert(comp_level > 0 && comp_level < CompLevel_count, "CompLevel out of bounds");


Use CompLevel_none instead of 0.

vnkozlov · 2026-03-21T18:54:46Z

test/hotspot/jtreg/compiler/compilercontrol/jcmd/ClearDirectivesFileStackTest.java

        }
        // clear the stack
-        builder.add(new JcmdCommand(Command.NONEXISTENT, null, null,
+        builder.add(new JcmdCommand(Command.NONEXISTENT, true, null, null,


Don you know how we were able to run it before? The JcmdCommand was changed 3 years ago.

I just came across this package and this file did not compile, so I fixed it.
Technically it does not belong to this PR. I can file a separate bug and fix it there, if needed.

kirill-shirokov added 18 commits March 20, 2026 21:06

8313713: Allow CompileCommand flag to specify compilation level: Init…

c991c6a

…ial draft implementation

Update copyright year

61afcbc

Use actual compile level when matching compile directives

be85ef5

Remove DirectiveSet::_comp_level. Add comp_level to DirectiveSet::sho…

ca608b4

…uld_not_inline(). Replace compiler arument with comp_level in DirectivesStack::getMatchingDirective()

Remove extra line added due to an error in CLion

8e00dc4

Fix flag checking and CompileOnly-checking code in CompilerOracle.cpp…

4ab7a5e

…. Fix level checking in WB_IsMethodCompilable()

Allow optional arguments for break, compileonly, exclude, print Compi…

12843b5

…leCommands

Fix crash in printing methods. Fix default CompileCommand CompLevel b…

73efcb6

…itmask

Include full CompileLevelParsingTest. Fix constructor arguments bug i…

fd4c0e4

…n ClearDirectivesFileStackTest. Sort includes. Update copyright year.

Disallow downgrading method in queue to compilation level 2 if this l…

a4ba2cf

…evel is excluded for the method. Fix WB_IsMethodCompilable for levels 1-3

Add CompileLevelWBTest.java for verifying actual compilations with co…

a693c2b

…mpileonly and exclude CompileCommands with a level argument

Add @requires to CompileLevelWBTest

684eae9

Minor fixes + formatting + extra comments

62b0281

Fix copyright year in ClearDirectivesFileStackTest.java

8d0ca08

Add a new test that verifies compilation levels from VM diagnostic ou…

571a537

…tput without WhiteBox

Update compiler/compilercontrol/commands/CompileLevelPrintTest.java t…

a5eb93c

…o communicate with testee via stdin

Compile level mask argument parsing fixed. TODOs removed. Negative ca…

60c4db2

…se added to CompileLevelParseTest.java. CompileTask::release_direcive() added for ownership clarity. Null pointer crash fixed in CompilationPolicy::select_task()

Fix copyright year

1e45337

openjdk bot added the hotspot hotspot-dev@openjdk.org label Mar 21, 2026

:retab

cbd3067

openjdk bot added the rfr Pull request is ready for review label Mar 21, 2026

vnkozlov reviewed Mar 21, 2026

View reviewed changes

Changes requested in the review comments

79407ad

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

8313713: Allow CompileCommand flag to specify compilation level#30352

8313713: Allow CompileCommand flag to specify compilation level#30352
kirill-shirokov wants to merge 20 commits intoopenjdk:masterfrom
kirill-shirokov:8313713-add-comp-level-to-compilecommand-2

kirill-shirokov commented Mar 21, 2026 •

edited by openjdk bot

Loading

Uh oh!

bridgekeeper bot commented Mar 21, 2026

Uh oh!

openjdk bot commented Mar 21, 2026

Uh oh!

openjdk bot commented Mar 21, 2026 •

edited

Loading

Uh oh!

mlbridge bot commented Mar 21, 2026 •

edited

Loading

Uh oh!

vnkozlov left a comment

Uh oh!

vnkozlov Mar 21, 2026

Uh oh!

vnkozlov Mar 21, 2026

Uh oh!

vnkozlov Mar 21, 2026

Uh oh!

vnkozlov Mar 21, 2026

Uh oh!

vnkozlov Mar 21, 2026

Uh oh!

kirill-shirokov Mar 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

Conversation

kirill-shirokov commented Mar 21, 2026 • edited by openjdk bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewing

Uh oh!

bridgekeeper bot commented Mar 21, 2026

Uh oh!

openjdk bot commented Mar 21, 2026

Uh oh!

openjdk bot commented Mar 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mlbridge bot commented Mar 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Webrevs

Uh oh!

vnkozlov left a comment

Choose a reason for hiding this comment

Uh oh!

vnkozlov Mar 21, 2026

Choose a reason for hiding this comment

Uh oh!

vnkozlov Mar 21, 2026

Choose a reason for hiding this comment

Uh oh!

vnkozlov Mar 21, 2026

Choose a reason for hiding this comment

Uh oh!

vnkozlov Mar 21, 2026

Choose a reason for hiding this comment

Uh oh!

vnkozlov Mar 21, 2026

Choose a reason for hiding this comment

Uh oh!

kirill-shirokov Mar 22, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

kirill-shirokov commented Mar 21, 2026 •

edited by openjdk bot

Loading

openjdk bot commented Mar 21, 2026 •

edited

Loading

mlbridge bot commented Mar 21, 2026 •

edited

Loading