8350209: Preserve adapters in AOT cache #24740

vnkozlov · 2025-04-17T19:11:47Z

JEP 483 preserves class information in AOT cache which helps Java startup performance.

We should also preserve adapters (i2c, c2i) to further improve performance of class linking where adapters are generated.

Short running Java application can see several percents improvement. I got 6% improvement when ran HelloWorld.java on Linux-x64 Ice Lake CPU (2.5Ghz):

(perf stat -r 100 java -XX:AOTCache=app.aotcache -cp hello.jar HelloWorld > /dev/null) 2>&1 | grep elapsed
         0.0299401 +- 0.0000504 seconds time elapsed  ( +-  0.17% )

(perf stat -r 100 java -XX:AOTCache=app.aotcache -XX:+UnlockDiagnosticVMOptions -XX:-AOTAdapterCaching -cp hello.jar HelloWorld > /dev/null) 2>&1 | grep elapsed
         0.0318654 +- 0.0000535 seconds time elapsed  ( +-  0.17% )

New diagnostic flags are introduced (use -XX:+UnlockDiagnosticVMOptions to unlock them):

-XX:+AOTAdapterCaching  - Enable or disable saving and restoring i2c2i adapters
-XX:AOTCodeMaxSize=10*M - buffer size in bytes for AOT code caching
-XX:+AbortVMOnAOTCodeFailure - Abort VM on the first occurrence of AOT code caching failure

By default AOTAdapterCaching is false and enabled ergonomically when -XX:AOTCache is specified.
This flag is ignored when AOTCache is not specified.

To use AOT adapters follow process described in JEP 483:

java -XX:AOTMode=record -XX:AOTConfiguration=app.aotconf -cp app.jar App
java -XX:AOTMode=create -XX:AOTConfiguration=app.aotconf -XX:AOTCache=app.aot -cp app.jar
java -XX:AOTCache=app.aot -cp app.jar App

There are several new UL flag combinations to trace the AOT code caching process:

-Xlog:aot+codecache+init -Xlog:aot+codecache+exit -Xlog:aot+codecache+stubs

@ashu-mehra is main author of changes. He implemented adapters caching.
I did main framework (AOTCodeCache class) for saving and loading AOT code.

Tested tier1-6,10, which includes tests with AOTClassLinking enabled. Also Xcomp,stress and JCK.

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-8350209: Preserve adapters in AOT cache (Enhancement - P3)

Reviewers

Vladimir Ivanov (@iwanowww - Reviewer)
Magnus Ihse Bursie (@magicus - Reviewer) Review applies to e29da02e
Ashutosh Mehra (@ashu-mehra - Committer)
Ioi Lam (@iklam - Reviewer) Review applies to 78f2828d

Contributors

Ashutosh Mehra <[email protected]>

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/24740/head:pull/24740
$ git checkout pull/24740

Update a local copy of the PR:
$ git checkout pull/24740
$ git pull https://git.openjdk.org/jdk.git pull/24740/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 24740

View PR using the GUI difftool:
$ git pr show -t 24740

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/24740.diff

Using Webrev

Link to Webrev Comment

bridgekeeper · 2025-04-17T19:12:27Z

👋 Welcome back kvn! 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 · 2025-04-17T19:12:36Z

@vnkozlov 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:

8350209: Preserve adapters in AOT cache

Co-authored-by: Ashutosh Mehra <[email protected]>
Reviewed-by: vlivanov, asmehra, ihse, iklam

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 14 new commits pushed to the master branch:

c116b18: 8355559: Benchmark modification/extension shouldn't affect the behavior of other benchmarks
9e26b9f: 8355372: GenShen: Test gc/shenandoah/generational/TestOldGrowthTriggers.java fails with UseCompactObjectHeaders enabled
394961c: 8328914: Document the java.security.debug property in javadoc
... and 11 more: https://git.openjdk.org/jdk/compare/e2ae50d877b13b121912e2496af4b5209b315a05...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.

openjdk · 2025-04-17T19:13:22Z

@vnkozlov The following labels will be automatically applied to this pull request:

build
hotspot

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

vnkozlov · 2025-04-17T19:17:24Z

/contributor add @ashu-mehra

openjdk · 2025-04-17T19:17:53Z

@vnkozlov
Contributor Ashutosh Mehra <[email protected]> successfully added.

mlbridge · 2025-04-17T20:37:07Z

Webrevs

ashu-mehra · 2025-04-18T06:01:43Z

@vnkozlov I forgot to remove AOTCodeSection from code/aotCodeCache.hpp when I refactored the APIs. This struct is no longer used. Can you please remove it.

vnkozlov · 2025-04-18T14:10:54Z

@vnkozlov I forgot to remove AOTCodeSection from code/aotCodeCache.hpp when I refactored the APIs. This struct is no longer used. Can you please remove it.

Done.

…for AOTCodeMaxSize values.

vnkozlov · 2025-04-21T21:44:05Z

/cc hotspot-runtime
/cc hotspot-compiler

openjdk · 2025-04-21T21:44:42Z

@vnkozlov
The hotspot-runtime label was successfully added.

vnkozlov · 2025-04-21T21:44:49Z

/label remove hotspot

openjdk · 2025-04-21T21:45:12Z

@vnkozlov
The hotspot-compiler label was successfully added.

openjdk · 2025-04-21T21:45:54Z

@vnkozlov
The hotspot label was successfully removed.

iklam · 2025-04-22T17:01:20Z

src/hotspot/share/cds/cdsConfig.cpp

+bool CDSConfig::is_dumping_aot_code_enabled() {
+  return _is_dumping_aot_code_enabled;


Other functions in CDSConfig don't have the _enabled() suffix. E.g., is_dumping_method_handles(). It doesn't mean we are doing method handle right now, but rather we have the ability to do so.

Since we rarely ask "am I in the middle of dumping X right now", I think adding _enabled() will be redundant.

vnkozlov · 2025-04-22T22:43:05Z

Added missing changes for Aarch64 to always generate far jumps in AOT code. generate_i2c2i_adapters() use far_jump().
Note, for x64 ForceUnreachable flag is set to true in AOTCodeCache::initialize(). Unfortunately Aarch64 does not use this flag, so I need to add check to each query functions.

iwanowww · 2025-04-30T22:09:24Z

src/hotspot/share/runtime/sharedRuntime.cpp

+}
+
+#ifndef PRODUCT
+void AdapterHandlerLibrary::print_adapter_handler_info(AdapterHandlerEntry* handler, AdapterBlob* adapter_blob) {


Suggestion: pass tty explicitly as outputStream*

void AdapterHandlerLibrary::print_adapter_handler_info_on(outptutStream* st, AdapterHandlerEntry* handler, AdapterBlob* adapter_blob) {

iwanowww · 2025-04-30T22:22:18Z

src/hotspot/share/code/relocInfo.hpp

@@ -1287,6 +1289,24 @@ class trampoline_stub_Relocation : public Relocation {

  void pack_data_to(CodeSection * dest) override;
  void unpack_data() override;
+#if defined(AARCH64)


It's unfortunate to see AArch64-specific code in shared code.

But I don't see anything besidespd_destination() and pd_set_destination() declarations. Where're their bodies?

Good catch. I copied it from leyden/premain but it is used only for trampoline jump relocation for calls in nmethods on Aarch64. They are not present in adapters. I remove this code.

iwanowww · 2025-04-30T22:29:40Z

src/hotspot/share/code/aotCodeCache.hpp

+/*
+ * AOT Code Cache collects code from Code Cache and corresponding metadata
+ * during application training run.
+ * In following "production" runs this code and data can me loaded into


iwanowww · 2025-04-30T22:37:06Z

src/hotspot/share/code/codeBlob.hpp

@@ -188,6 +204,8 @@ class CodeBlob {
  nmethod* as_nmethod_or_null() const         { return is_nmethod() ? (nmethod*) this : nullptr; }
  nmethod* as_nmethod() const                 { assert(is_nmethod(), "must be nmethod"); return (nmethod*) this; }
  CodeBlob* as_codeblob() const               { return (CodeBlob*) this; }
+  AdapterBlob* as_adapter_blob() const        { assert(is_adapter_blob(), "must be adapter blob"); return (AdapterBlob*) this; }
+  ExceptionBlob* as_exception_blob() const    { assert(is_exception_stub(), "must be exception stub"); return (ExceptionBlob*) this; }


ExceptionBlob is C2-specific, but as_exception_blob() is unused.

iwanowww · 2025-04-30T22:54:07Z

src/hotspot/share/code/aotCodeCache.cpp

+  if (AbortVMOnAOTCodeFailure) {
+    vm_exit_during_initialization("Unable to use AOT Code Cache.", nullptr);
+  }
+  log_info(aot, codecache, init)("Unable to use AOT Code Cache.");


Should it be a warning instead?

iwanowww · 2025-04-30T22:54:38Z

src/hotspot/share/code/aotCodeCache.cpp

+    tty->print_cr("Unable to create AOT Code Cache.");
+    vm_abort(false);
+  }
+  log_info(aot, codecache, exit)("Unable to create AOT Code Cache.");


Same here (log_warning?).

I will remove exit_vm_on_ from these 2 methods to avoid confusion. And will add comment.

iwanowww · 2025-04-30T22:55:38Z

src/hotspot/share/code/aotCodeCache.cpp

+  AOTAdapterCaching = false;
+}
+
+static void exit_vm_on_store_failure() {


The naming (exit_vm_on_load_failure and exit_vm_on_store_failure) still look confusing to me. By default, they disable AOTAdapterCaching and issue a message, but the name strongly suggests that execution halts there:

if (!open_cache(is_dumping, is_using)) { if (is_using) { exit_vm_on_load_failure(); } else { exit_vm_on_store_failure(); } return; }

iwanowww · 2025-04-30T22:59:03Z

src/hotspot/share/runtime/sharedRuntime.cpp

+    entry_offset[1] = handler->get_c2i_entry() - i2c_entry;
+    entry_offset[2] = handler->get_c2i_unverified_entry() - i2c_entry;
+    entry_offset[3] = handler->get_c2i_no_clinit_check_entry() - i2c_entry;
+    AOTCodeCache::store_code_blob(*adapter_blob, AOTCodeEntry::Adapter, id, name, AdapterHandlerEntry::ENTRIES_COUNT, entry_offset);


What the intended behavior here when AOTCodeCache::store_code_blob fails?

If something happened when we tried to store blob (adapter in this case) into buffer (for example, if reserved for AOT code buffer is too small) the next will be called:

set_failed(); exit_vm_on_store_failure();

So we either abort VM based on the flag or issue warning (as you suggested) and continue execution.

set_failed() will prevent following attempts to cache blobs and prevent final dump any cached code into AOT cache:

bool for_dump() const { return _for_dump && !_failed; }

Maybe add an assert here?

bool success = AOTCodeCache::store_code_blob(...); assert(success || !AOTCodeCache::is_dumping_adapters(), "");

assert(success || !AOTCodeCache::is_dumping_adapters(), "");

This condtion !AOTCodeCache::is_dumping_adapters() in the assert is not very intuitive. I think what we need to assert is future stores in the aot code cache are disabled. So having a method like AOTCodeCache::is_store_disabled() would better communicate the intent here. But I don't mind keeping this condition for this initial PR. I will just suggest to add a better assert message like:

assert(success || !AOTCodeCache::is_dumping_adapters(), "storing of adapter must be disabled");

And I think we should also be setting _adapter_caching to false in report_load_failure and report_store_failure to be consistent, otherwise we may end up in a situation where AOTAdapterCaching is false but _adapter_caching is true. In fact, I feel we should only be setting _adapter_caching and not AOTAdapterCaching in report_load/store_failure because _adapter_caching is the flag used to gate store/load operations.

Thank you, @ashu-mehra. You have good points. I will work on them.

FTR I suggested !AOTCodeCache::is_dumping_adapters() because that's the guarding check for AOTCodeCache::store_code_blob() call.

I will add assert message but keep check.

iwanowww · 2025-04-30T23:02:14Z

src/hotspot/share/code/aotCodeCache.cpp

+  AOTAdapterCaching = false;
+}
+
+static void exit_vm_on_store_failure() {


Setting AOTAdapterCaching to false on failure is simple indication that adapter caching is switched off for someone who will look on final state of flag.

But how does it affect execution? What's the intended behavior when a failure happens during store attempt? What are the consistency guarantees for AOT code cache during dumping in presense of store failures? What I see right now is that errors reported by AOTCodeCache::store_code_blob() are silently ignored. How does _cache notice a store failure during dumping phase?

vnkozlov · 2025-05-01T00:30:08Z

How does _cache notice a store failure during dumping phase?
set_failed() sets flags AOTCodeCache::_failed. Which is checked when we trying to get _cache pointer:

  AOTCodeCache* cache = open_for_use();
  if (cache == nullptr) {
    return nullptr;
  }
  
  or
  
  AOTCodeCache* cache = open_for_dump();
  if (cache == nullptr) {
    return false;
  }

and

AOTCodeCache* AOTCodeCache::open_for_use() {
  if (AOTCodeCache::is_on_for_use()) {
    return AOTCodeCache::cache();
  }
  return nullptr;
}

AOTCodeCache* AOTCodeCache::open_for_dump() {
  if (AOTCodeCache::is_on_for_dump()) {
    AOTCodeCache* cache = AOTCodeCache::cache();
    cache->clear_lookup_failed(); // Reset bit <<<<<<<<<< this is different failure: we did not find one entry
    return cache;
  }
  return nullptr;
}

and 

  static bool is_on_for_use()  { return is_on() && _cache->for_use(); }
  static bool is_on_for_dump() { return is_on() && _cache->for_dump(); }

  bool for_use()  const { return _for_use  && !_failed; }
  bool for_dump() const { return _for_dump && !_failed; }

vnkozlov · 2025-05-01T00:52:57Z

Thank you, @iwanowww, for review. I think I addressed your comments and answered your question.

iwanowww

Looks good!

Minor suggestions follow.

iwanowww · 2025-05-01T01:51:07Z

src/hotspot/share/runtime/sharedRuntime.cpp

+    entry_offset[1] = handler->get_c2i_entry() - i2c_entry;
+    entry_offset[2] = handler->get_c2i_unverified_entry() - i2c_entry;
+    entry_offset[3] = handler->get_c2i_no_clinit_check_entry() - i2c_entry;
+    AOTCodeCache::store_code_blob(*adapter_blob, AOTCodeEntry::Adapter, id, name, AdapterHandlerEntry::ENTRIES_COUNT, entry_offset);


Maybe add an assert here?

bool success = AOTCodeCache::store_code_blob(...); assert(success || !AOTCodeCache::is_dumping_adapters(), "");

iwanowww · 2025-05-01T01:53:20Z

src/hotspot/share/code/aotCodeCache.cpp

+#include <sys/stat.h>
+#include <errno.h>
+
+static void load_failure() {


Maybe on_(load|store)_failure() or report_(load|store)_failure()?

I will use report_

vnkozlov · 2025-05-01T02:55:43Z

Thank you, @iwanowww. I need 2 re-reviews so I asked @ashu-mehra to review again.
Meanwhile I am re-running testing with latest mainline code.

…heck.

vnkozlov · 2025-05-01T18:50:19Z

I reverted back log_warning() to log_info() in report_*_failure() methods.

I removed AOTCodeCache::_adapter_caching field and use AOTAdapterCaching flag check instead. Setting the flag to false in case of a failure make more sense now.

iwanowww

Looks good.

vnkozlov · 2025-05-01T19:27:48Z

/reviewers 1

openjdk · 2025-05-01T19:28:32Z

@vnkozlov
The total number of required reviews for this PR (including the jcheck configuration and the last /reviewers command) is now set to 1 (with at least 1 Reviewer).

ashu-mehra

lgtm

vnkozlov · 2025-05-01T21:02:36Z

Thank you again, @iwanowww and @ashu-mehra

vnkozlov · 2025-05-01T21:02:44Z

/integrate

openjdk · 2025-05-01T21:03:25Z

Going to push as commit aae2bb6.
Since your change was applied there have been 14 commits pushed to the master branch:

c116b18: 8355559: Benchmark modification/extension shouldn't affect the behavior of other benchmarks
9e26b9f: 8355372: GenShen: Test gc/shenandoah/generational/TestOldGrowthTriggers.java fails with UseCompactObjectHeaders enabled
394961c: 8328914: Document the java.security.debug property in javadoc
... and 11 more: https://git.openjdk.org/jdk/compare/e2ae50d877b13b121912e2496af4b5209b315a05...master

Your commit was automatically rebased without conflicts.

openjdk · 2025-05-01T21:03:32Z

@vnkozlov Pushed as commit aae2bb6.

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

8350209: Preserve adapters in AOT cache

544a4aa

openjdk bot added build [email protected] hotspot [email protected] labels Apr 17, 2025

vnkozlov marked this pull request as ready for review April 17, 2025 20:32

openjdk bot added the rfr Pull request is ready for review label Apr 17, 2025

vnkozlov mentioned this pull request Apr 17, 2025

8350209: Preserve adapters in AOT cache #24158

Closed

3 tasks

Removed unused AOTCodeSection class

43517eb

vnkozlov added 2 commits April 18, 2025 11:03

AOT code flags are ignored when AOTCache is not specified. Set range …

ed37319

…for AOTCodeMaxSize values.

Add sanity test for AOTAdapterCaching flag

2863a96

openjdk bot added the hotspot-runtime [email protected] label Apr 21, 2025

openjdk bot added the hotspot-compiler [email protected] label Apr 21, 2025

openjdk bot removed the hotspot [email protected] label Apr 21, 2025

iklam reviewed Apr 22, 2025

View reviewed changes

vnkozlov added 2 commits April 22, 2025 10:29

remove _enabled suffix

ba08626

Generate far jumps for AOT code on AArch64

32b997d

Fix message

91c37da

iklam approved these changes Apr 30, 2025

View reviewed changes

iwanowww reviewed Apr 30, 2025

View reviewed changes

Address Vladimir's comments

70bd029

iwanowww approved these changes May 1, 2025

View reviewed changes

vnkozlov added 2 commits April 30, 2025 19:43

renaming and new assert

9c3e368

Merge branch 'master' into JDK-8350209

286e0e6

iwanowww approved these changes May 1, 2025

View reviewed changes

vnkozlov requested a review from ashu-mehra May 1, 2025 02:54

Revert log_warning change. Replace _adapter_caching check with flag c…

325efec

…heck.

vnkozlov requested a review from iwanowww May 1, 2025 18:58

iwanowww approved these changes May 1, 2025

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label May 1, 2025

ashu-mehra approved these changes May 1, 2025

View reviewed changes

openjdk bot added the integrated Pull request has been integrated label May 1, 2025

openjdk bot closed this May 1, 2025

openjdk bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels May 1, 2025

vnkozlov deleted the JDK-8350209 branch May 1, 2025 21:07

vnkozlov mentioned this pull request May 2, 2025

8355514: [premain] rename AOT code flags and adjust their ergonomic settings openjdk/leyden#65

Closed

2 tasks

RealFYang mentioned this pull request Jul 4, 2025

8360000: RISC-V: implement aot #26101

Open

3 tasks

		bool CDSConfig::is_dumping_aot_code_enabled() {
		return _is_dumping_aot_code_enabled;

8350209: Preserve adapters in AOT cache #24740

8350209: Preserve adapters in AOT cache #24740

Uh oh!

Conversation

vnkozlov commented Apr 17, 2025 • edited by openjdk bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewers

Contributors

Reviewing

Uh oh!

bridgekeeper bot commented Apr 17, 2025

Uh oh!

openjdk bot commented Apr 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openjdk bot commented Apr 17, 2025

Uh oh!

vnkozlov commented Apr 17, 2025

Uh oh!

openjdk bot commented Apr 17, 2025

Uh oh!

mlbridge bot commented Apr 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Webrevs

Uh oh!

ashu-mehra commented Apr 18, 2025

Uh oh!

vnkozlov commented Apr 18, 2025

Uh oh!

vnkozlov commented Apr 21, 2025

Uh oh!

openjdk bot commented Apr 21, 2025

Uh oh!

vnkozlov commented Apr 21, 2025

Uh oh!

openjdk bot commented Apr 21, 2025

Uh oh!

openjdk bot commented Apr 21, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

vnkozlov commented Apr 22, 2025

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!

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!

Choose a reason for hiding this comment

vnkozlov commented Apr 17, 2025 •

edited by openjdk bot

Loading

openjdk bot commented Apr 17, 2025 •

edited

Loading

mlbridge bot commented Apr 17, 2025 •

edited

Loading

vnkozlov commented May 1, 2025 •

edited

Loading