8276711: compiler/codecache/cli tests failing when SegmentedCodeCache used with -Xint #7650

coleenp · 2022-03-01T20:50:23Z

In Loom, when using -Xint, the +SegmentedCodeCache option cannot be used because it doesn't generate a code heap for nmethods, and in loom the compiler needs to generate an nmethod for Continuation.enterSpecial even with -Xint.
This change is @rickard 's loom change with the tests fixed so they pass for it. One tests for the new warning message and code cache usage, and the others remove the Xint tests.
Tested with tier1-4.

Progress

Change must not contain extraneous whitespace
Commit message must refer to an issue
Change must be properly reviewed

Issue

JDK-8276711: compiler/codecache/cli tests failing when SegmentedCodeCache used with -Xint

Reviewers

Vladimir Kozlov (@vnkozlov - Reviewer)
David Holmes (@dholmes-ora - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.java.net/jdk pull/7650/head:pull/7650
$ git checkout pull/7650

Update a local copy of the PR:
$ git checkout pull/7650
$ git pull https://git.openjdk.java.net/jdk pull/7650/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 7650

View PR using the GUI difftool:
$ git pr show -t 7650

Using diff file

Download this PR as a diff file:
https://git.openjdk.java.net/jdk/pull/7650.diff

… used with -Xint

bridgekeeper · 2022-03-01T20:51:58Z

👋 Welcome back coleenp! 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 · 2022-03-01T20:54:37Z

@coleenp 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 · 2022-03-01T20:58:26Z

Webrevs

dholmes-ora · 2022-03-01T21:27:18Z

src/hotspot/share/runtime/arguments.cpp

+  if (_mode == _int) {
+    if (SegmentedCodeCache) {
+      warning("SegmentedCodeCache has no meaningful effect with -Xint");
+      FLAG_SET_DEFAULT(SegmentedCodeCache, false);
+    }
+  }


Isn't this better placed in Arguments::set_mode_flags where the other compiler related flags get turned off?

SegmentedCodeCache should be set in compiler/compilerDefinitions.cpp:
https://github.com/openjdk/jdk/blob/master/src/hotspot/share/compiler/compilerDefinitions.cpp#L294

I replied on email but I don't see it and this is better here. I should really change this to be:

if (FLAG_IS_CMDLINE(SegmentedCodeCache)) {
}
Since this code is checking for someone setting it on the command line with -Xint. The code in compilerDefinitions.cpp sets the flag ergonomically if the code cache is big and if tiered compilation is on. I believe this is after the code in arguments.cpp. So this code won't reset it to true, if I'm following it right.
edit: yes, I'll check the flag in compilerDefinitions.cpp.

I agree that the warning should be issued if flag is set on command line together with -Xint.
In this case you need add this check to CompilerConfig::check_args_consistency() where different compiler flags are set to false when we run with Interpreter:
https://github.com/openjdk/jdk/blob/master/src/hotspot/share/compiler/compilerDefinitions.cpp#L493

Thanks Coleen, seems the use of set_mode_flags is somewhat flawed as you can change any of the flags that it tries to force on/off simply by having them later on the command-line.

You could do as @vnkozlov suggests and move this check into compilerDefinitions.cpp as CompilerConfig::ergo_initialize() is called just a few lines above this.

Thanks Vladimir, that's a better place for it. There are other checks for inconsistent flags with -Xint in check_args_consistency(). One of the tests still fails with that change so I'll have to figure out why.

I moved the check but TieredStopAtLevel=0 still allocates code heaps for nmethods, so the check only checks for the -Xint in arguments. I don't know if this makes sense or not in the code heap allocation code, but the codecache/cli tests test for sizes of code heaps pass with this change.

mlbridge · 2022-03-02T00:01:23Z

Mailing list message from coleen.phillimore at oracle.com on hotspot-dev:

On 3/1/22 4:29 PM, David Holmes wrote:

On Tue, 1 Mar 2022 20:50:23 GMT, Coleen Phillimore <coleenp at openjdk.org> wrote:

In Loom, when using -Xint, the +SegmentedCodeCache option cannot be used because it doesn't generate a code heap for nmethods, and in loom the compiler needs to generate an nmethod for Continuation.enterSpecial even with -Xint.
This change is @rickard 's loom change with the tests fixed so they pass for it. One tests for the new warning message and code cache usage, and the others remove the Xint tests.
Tested with tier1-4.
src/hotspot/share/runtime/arguments.cpp line 4122:

4120: FLAG_SET_DEFAULT(SegmentedCodeCache, false);
4121: }
4122: }
Isn't this better placed in Arguments::set_mode_flags where the other compiler related flags get turned off?

Well, the default for SegmentedCodeCache is off.? It's set here because
it's after that option is found on the command line.? Maybe the line
should be

if (FLAG_IS_CMDLINE(SegmentedCodeCache)) {
}

to make it more clear.

Coleen

dholmes-ora · 2022-03-02T02:24:12Z

src/hotspot/share/compiler/compilerDefinitions.cpp

@@ -524,6 +523,13 @@ bool CompilerConfig::check_args_consistency(bool status) {
 #endif
  }

+  // TieredStopAtLevel==0 allocates nmethod space in the code heap with
+  // SegmentedCodeCache so only disallow the option for -Xint.
+  if (Arguments::is_interpreter_only() && FLAG_IS_CMDLINE(SegmentedCodeCache)) {


You need to check SegmentedCodeCache==true else you will generate the warning if someone is explicitly turning it off in combination with -Xint.

Yes, code should follow pattern in previous lines:

if (SegmentedCodeCache) { if (!FLAG_IS_DEFAULT(SegmentedCodeCache)) { warning("SegmentedCodeCache has no meaningful effect with -Xint"); } FLAG_SET_CMDLINE(SegmentedCodeCache, false); }

And you should include it in previous scope which checks if (Arguments::is_interpreter_only())

The previous scope checks CompilerConfig::is_interpreter_only() which includes TieredStopAtLevel == 0, so it's not the same there.

vnkozlov · 2022-03-02T02:44:18Z

src/hotspot/share/compiler/compilerDefinitions.cpp

@@ -524,6 +523,13 @@ bool CompilerConfig::check_args_consistency(bool status) {
 #endif
  }

+  // TieredStopAtLevel==0 allocates nmethod space in the code heap with


I would prefer to not have special case TieredStopAtLevel==0. Unless you want to test loom when JIT compilers are disabled. Currently we don't run any tests with such setting.

I tested with this setting and it works fine with -XX:+SegmentedCodeCache because the code for segmented code cache only checks Arguments::is_interpreter_only(). This doesn't seem right to me, but the testing also expects this to work. I could fix the tests to not do this. Now I don't know what anyone wants!

… a nmethod heaps, and fix tests accordingly.

coleenp · 2022-03-02T13:57:34Z

I pushed a change to make TieredStopAtLevel=0 also disable SegmentCodeCache and fixed the tests for that also, but am rerunning tests tier1-4, then 5-6.

vnkozlov

Good.

openjdk · 2022-03-02T17:33:34Z

@coleenp This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

8276711: compiler/codecache/cli tests failing when SegmentedCodeCache used with -Xint

Reviewed-by: kvn, dholmes

You can use pull request commands such as /summary, /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been 12 new commits pushed to the master branch:

d0eb6fa: 8281569: Create tests for Frame.setMinimumSize() method
02aa7ce: 8282515: More clean up on NativeLibraries just for JNI library use
1485883: 8281628: KeyAgreement : generateSecret intermittently not resetting
b6c35ae: 8209784: Include hsdis in the JDK
ce18ff8: 8282551: Properly initialize L32X64MixRandom state
234c17e: 8274788: Support archived heap objects in ParallelGC
fc918a7: 8281234: The -protected option is not always checked in keytool and jarsigner
d80f697: 8282523: Fix 'hierachy' typo
f12200c: 8267796: vmTestbase/nsk/jvmti/scenarios/hotswap/HS201/hs201t002/TestDescription.java fails with NoClassDefFoundError
12a822a: 8282381: Parallel: Remove unnecessary PCReferenceProcessor
... and 2 more: https://git.openjdk.java.net/jdk/compare/732d891f9f8987edfc188c049eb19de6ba790c42...master

As there are no conflicts, your changes will automatically be rebased on top of these commits when integrating. If you prefer to avoid this automatic rebasing, please check the documentation for the /integrate command for further details.

➡️ To integrate this PR with the above commit message to the master branch, type /integrate in a new comment.

coleenp · 2022-03-02T18:23:09Z

Thanks Vladimir. Tiers 1-4 passed. Now 5-6 also.

dholmes-ora · 2022-03-02T22:34:24Z

src/hotspot/share/code/codeCache.cpp

@@ -368,7 +368,7 @@ bool CodeCache::heap_available(int code_blob_type) {
  if (!SegmentedCodeCache) {
    // No segmentation: use a single code heap
    return (code_blob_type == CodeBlobType::All);
-  } else if (Arguments::is_interpreter_only()) {
+  } else if (CompilerConfig::is_interpreter_only()) {


Doesn't this introduce a very subtle change in behaviour in relation to the handling of MethodNonProfiled?
@vnkozlov please confirm this is what you intended - I don't know the implications.

According to my conversation with Rickard, TierStopAtLevel=0 is the same as -Xint. If TieredStopAtLevel=0, the compiler won't allocate nmethods in the MethodProfiled and MethodNonProfiled areas anyway, so this won't allocate them now.

I think this code was missed during Igor's V. changes which introduced CompilerConfig::is_interpreter_only().
Coleen is correct, with TieredStopAtLevel=0 all JIT compilers are disabled. MethodNonProfiled is codeheap for tier1 (C1) and tier4 (C2) compiled nmethods. It does not make sense to have such codeheap when compilation disabled.

Okay, so does that mean this block:

} else { // No TieredCompilation: we only need the non-nmethod and non-profiled code heap return (code_blob_type == CodeBlobType::NonNMethod) || (code_blob_type == CodeBlobType::MethodNonProfiled); }

is actually unreachable?

It is reachable for -XX:-TieredCompilation when only C2 is used. Or when only tier1 C1 is used with -XX:TieredStopAtLevel=1.

Ah I see. Thanks for clarifying @vnkozlov !

-XX:TieredStopAtLevel=1 is the old -client switch. Thanks Vladimir for confirming.

coleenp · 2022-03-03T13:19:45Z

Thanks David.
/integrate

openjdk · 2022-03-03T13:21:13Z

Going to push as commit 7822cbc.
Since your change was applied there have been 17 commits pushed to the master branch:

832729b: 6911375: mouseWheel has no effect without vertical scrollbar
0402a28: 8037965: NullPointerException in TextLayout.getBaselineFromGraphic() for JTextComponents
d4d1fbc: 8282484: G1: Predicted old time in log always zero
b1f935c: 8277055: Assert "missing inlining msg" with -XX:+PrintIntrinsics
2da6777: 8281122: [IR Framework] Cleanup IR matching code in preparation for JDK-8280378
d0eb6fa: 8281569: Create tests for Frame.setMinimumSize() method
02aa7ce: 8282515: More clean up on NativeLibraries just for JNI library use
1485883: 8281628: KeyAgreement : generateSecret intermittently not resetting
b6c35ae: 8209784: Include hsdis in the JDK
ce18ff8: 8282551: Properly initialize L32X64MixRandom state
... and 7 more: https://git.openjdk.java.net/jdk/compare/732d891f9f8987edfc188c049eb19de6ba790c42...master

Your commit was automatically rebased without conflicts.

openjdk · 2022-03-03T13:21:26Z

@coleenp Pushed as commit 7822cbc.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

8276711: compiler/codecache/cli tests failing when SegmentedCodeCache…

c369bc4

… used with -Xint

openjdk bot added the rfr Pull request is ready for review label Mar 1, 2022

openjdk bot added the hotspot hotspot-dev@openjdk.org label Mar 1, 2022

dholmes-ora reviewed Mar 1, 2022

View reviewed changes

Move check for -Xint.

462c49c

dholmes-ora reviewed Mar 2, 2022

View reviewed changes

vnkozlov reviewed Mar 2, 2022

View reviewed changes

Make TieredStopAtLevel=0 also ignore SegmentedCodeHeap and not create…

c362ffa

… a nmethod heaps, and fix tests accordingly.

vnkozlov approved these changes Mar 2, 2022

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Mar 2, 2022

dholmes-ora reviewed Mar 2, 2022

View reviewed changes

dholmes-ora approved these changes Mar 3, 2022

View reviewed changes

openjdk bot added the integrated Pull request has been integrated label Mar 3, 2022

openjdk bot closed this Mar 3, 2022

openjdk bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels Mar 3, 2022

coleenp deleted the xint branch March 3, 2022 13:21

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

8276711: compiler/codecache/cli tests failing when SegmentedCodeCache used with -Xint #7650

8276711: compiler/codecache/cli tests failing when SegmentedCodeCache used with -Xint #7650

coleenp commented Mar 1, 2022 •

edited by openjdk bot

Loading

bridgekeeper bot commented Mar 1, 2022

openjdk bot commented Mar 1, 2022

mlbridge bot commented Mar 1, 2022 •

edited

Loading

dholmes-ora Mar 1, 2022

vnkozlov Mar 1, 2022

coleenp Mar 1, 2022 •

edited

Loading

vnkozlov Mar 1, 2022

dholmes-ora Mar 1, 2022

coleenp Mar 1, 2022

coleenp Mar 2, 2022

mlbridge bot commented Mar 2, 2022

dholmes-ora Mar 2, 2022

vnkozlov Mar 2, 2022

vnkozlov Mar 2, 2022

coleenp Mar 2, 2022

vnkozlov Mar 2, 2022

coleenp Mar 2, 2022

coleenp commented Mar 2, 2022

vnkozlov left a comment

openjdk bot commented Mar 2, 2022 •

edited

Loading

coleenp commented Mar 2, 2022 •

edited

Loading

dholmes-ora Mar 2, 2022

coleenp Mar 2, 2022

vnkozlov Mar 3, 2022

dholmes-ora Mar 3, 2022

vnkozlov Mar 3, 2022

dholmes-ora Mar 3, 2022

coleenp Mar 3, 2022

coleenp commented Mar 3, 2022

openjdk bot commented Mar 3, 2022

openjdk bot commented Mar 3, 2022

8276711: compiler/codecache/cli tests failing when SegmentedCodeCache used with -Xint #7650

8276711: compiler/codecache/cli tests failing when SegmentedCodeCache used with -Xint #7650

Conversation

coleenp commented Mar 1, 2022 • edited by openjdk bot Loading

Progress

Issue

Reviewers

Reviewing

bridgekeeper bot commented Mar 1, 2022

openjdk bot commented Mar 1, 2022

mlbridge bot commented Mar 1, 2022 • edited Loading

Webrevs

Choose a reason for hiding this comment

Choose a reason for hiding this comment

coleenp Mar 1, 2022 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mlbridge bot commented Mar 2, 2022

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

coleenp commented Mar 2, 2022

vnkozlov left a comment

Choose a reason for hiding this comment

openjdk bot commented Mar 2, 2022 • edited Loading

coleenp commented Mar 2, 2022 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

coleenp commented Mar 3, 2022

openjdk bot commented Mar 3, 2022

openjdk bot commented Mar 3, 2022

coleenp commented Mar 1, 2022 •

edited by openjdk bot

Loading

mlbridge bot commented Mar 1, 2022 •

edited

Loading

coleenp Mar 1, 2022 •

edited

Loading

openjdk bot commented Mar 2, 2022 •

edited

Loading

coleenp commented Mar 2, 2022 •

edited

Loading