8380753: C2: RegMask::find_first_set() incorrectly returns partially contained sets for scalable vector registers by shqking · Pull Request #30890 · openjdk/jdk

shqking · 2026-04-23T01:15:31Z

Issue: compiler/loopopts/superword/TestAlignVectorFuzzer.java fails intermittently only on Neoverse-V1 machine (256-bit SVE) after JDK-8325467 with:

  Internal Error (/tmp/jdk-src/src/hotspot/share/opto/chaitin.cpp:1451), pid=101028, tid=101050
  assert(mask.can_represent(assigned)) failed: sanity

Root cause: On Neoverse-V1, the physical size of a scalable vector register exceeds RegMask::SlotsPerVecA (i.e. 2x larger). When querying a stack slot, RegMask::find_first_set() returns a register that is partially covered by the mask, violating can_represent().

Note-1: The failure is intermittent. I didn't check the jtreg code in details. I suppose the randomized and stress tests hit this corner case.

Note-2: JDK-8325467 converted a previously handled can_represent() check into an assertion. I suppose the bug is independent of other parts of JDK-8325467.

In this patch, we further validate the query result in RegMask::find_first_set().

Test-1: Run TestAlignVectorFuzzer.java case 100 times on Neoverse-V1 and all passes.

Test-2: Run tier1, tier2 and tier3 on Linux AArch64 and x86_64. No new failure.

Test-3: Add one unit test in gtest:RegMask. It fails on AArch64 if the change to regmask.cpp (that is done in this patch) is removed.

Co-contributed-by: Daniel Lunden daniel.lunden@oracle.com

I confirm that I make this contribution in accordance with the OpenJDK Interim AI Policy.

Progress

Change must not contain extraneous whitespace
Commit message must refer to an issue
Change must be properly reviewed (2 reviews required, with at least 1 Reviewer, 1 Author)

Issue

JDK-8380753: C2: RegMask::find_first_set() incorrectly returns partially contained sets for scalable vector registers (Bug - P4)

Reviewers

Anton Seoane Ampudia (@anton-seoane - Author)
Vladimir Kozlov (@vnkozlov - Reviewer)
Fei Yang (@RealFYang - Reviewer)

Contributors

Daniel Lundén <dlunden@openjdk.org>

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/30890/head:pull/30890
$ git checkout pull/30890

Update a local copy of the PR:
$ git checkout pull/30890
$ git pull https://git.openjdk.org/jdk.git pull/30890/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 30890

View PR using the GUI difftool:
$ git pr show -t 30890

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/30890.diff

Using Webrev

Link to Webrev Comment

…contained sets for scalable vector registers Issue: compiler/loopopts/superword/TestAlignVectorFuzzer.java fails intermittently only on Neoverse-V1 machine (256-bit SVE) after JDK-8325467 with: ``` Internal Error (/tmp/jdk-src/src/hotspot/share/opto/chaitin.cpp:1451), pid=101028, tid=101050 assert(mask.can_represent(assigned)) failed: sanity ``` Root cause: On Neoverse-V1, the physical size of a scalable vector register exceeds RegMask::SlotsPerVecA (i.e. 2x larger). When querying a stack slot, RegMask::find_first_set() returns a register that is partially covered by the mask, violating can_represent(). Note-1: The failure is intermittent. I didn't check the jtreg code in details. I suppose the randomized and stress tests hit this corner case. Note-2: JDK-8325467 converted a previously handled can_represent() check into an assertion. I suppose the bug is independent of other parts of JDK-8325467. In this patch, we further validate the query result in RegMask::find_first_set(). Test-1: Run TestAlignVectorFuzzer.java case 100 times on Neoverse-V1 and all passes. Test-2: Run tier1, tier2 and tier3 on Linux AArch64 and x86_64. No new failure. Test-3: Add one unit test in gtest:RegMask. It fails on AArch64 if the change to regmask.cpp (that is done in this patch) is removed. Co-contributed-by: Daniel Lunden <daniel.lunden@oracle.com>

bridgekeeper · 2026-04-23T01:16:43Z

👋 Welcome back haosun! 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 · 2026-04-23T01:17:06Z

@shqking 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:

8380753: C2: RegMask::find_first_set() incorrectly returns partially contained sets for scalable vector registers

Co-authored-by: Daniel Lundén <dlunden@openjdk.org>
Reviewed-by: aseoane, kvn, fyang

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 117 new commits pushed to the master branch:

2342f6b: 8381550: Shenandoah: Fix confusing end of degenerated cycle log message
5d49a68: 8383459: [BACKOUT] [macos] VoiceOver Reads "Header" instead of "Heading"
38a8e3c: 8381436: Remove Obsolete Translation Resources
... and 114 more: https://git.openjdk.org/jdk/compare/1f782b7a5209cdcde0d95adbe38ca6911e080c2b...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 · 2026-04-23T01:17:58Z

@shqking The following label will be automatically applied to this pull request:

hotspot-compiler

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.

openjdk · 2026-04-23T01:18:00Z

The total number of required reviews for this PR has been set to 2 based on the presence of this label: hotspot-compiler. This can be overridden with the /reviewers command.

mlbridge · 2026-04-23T01:23:34Z

Webrevs

00: Full (d2fa05fb)

shqking · 2026-04-23T01:24:24Z

/contributor add @dlunde

openjdk · 2026-04-23T01:25:17Z

@shqking
Contributor Daniel Lundén <dlunden@openjdk.org> successfully added.

shqking · 2026-04-23T01:37:34Z

More tests we have done:

@dlunde helped test tier1 to tier4 (and additional Oracle-internal testing) on Windows x64, Linux x64, Linux aarch64, macOS x64, and macOS aarch64

I suppose this issue may affect RISC-V as well. I don't have the access to RISC-V. I was wondering if @RealFYang could help run some unit test? Thanks

test-1: Build with this patch and run make test TEST="gtest::RegMask". The expected result is pass.
test-2: Revert the change to regmask.cpp. Build and run make test TEST="gtest::RegMask". We expect this test would fail with the following assertion.

[ RUN      ] RegMask.find_first_set_scalable_vector_vm
/tmp/jdk-src/test/hotspot/gtest/opto/test_regmask.cpp:278: Failure
Expected equality of these values:
  rm.find_first_set(lrg, scalable_reg_slots)
    Which is: 387             <-- this value might be different on RISC-V
  OptoReg::Bad
    Which is: -1

RealFYang · 2026-04-23T02:15:37Z

@shqking : Sure. Let me try it on linux-riscv64 with big vectors. Thanks for the ping.

shqking · 2026-04-23T02:27:03Z

@shqking : Sure. Let me try it on linux-riscv64 with big vectors. Thanks for the ping.

Kindly remind that

we can run this unit test gtest::RegMask on any linux-riscv64. No need to select big vectors.
if we want to reproduce the intermittent jtreg failure compiler/loopopts/superword/TestAlignVectorFuzzer.java that we encountered on AArch64 Neoverse-V1 previously, linx-rsicv64 with big vectors is needed.

anton-seoane · 2026-04-24T05:36:04Z

Thanks for this @shqking! The changeset looks good, to be on the safe side let me (re-)run a bit of testing

anton-seoane

Testing passed, code looks good. Thanks!

vnkozlov

LGTM

RealFYang

LGTM. My local test on linux-riscv64 w/wo big vectors is good.

shqking · 2026-04-28T10:34:15Z

Thanks for your reviews and tests.

/integrate

openjdk · 2026-04-28T10:34:54Z

Going to push as commit b3f658d.
Since your change was applied there have been 125 commits pushed to the master branch:

c39cf56: 8381898: SwitchBootstraps::typeSwitch() not accepting Float, Double types
56559eb: 8380955: Value-based classes in FFM should have @valuebased
d8e2a54: 8383367: jdk/jfr/event/oldobject/TestDFSWithSmallStack.java failed with -XX:+UseZGC -XX:+UseCompactObjectHeaders
... and 122 more: https://git.openjdk.org/jdk/compare/1f782b7a5209cdcde0d95adbe38ca6911e080c2b...master

Your commit was automatically rebased without conflicts.

openjdk · 2026-04-28T10:35:04Z

@shqking Pushed as commit b3f658d.

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

openjdk Bot added the hotspot-compiler hotspot-compiler-dev@openjdk.org label Apr 23, 2026

openjdk Bot added the rfr Pull request is ready for review label Apr 23, 2026

anton-seoane approved these changes Apr 27, 2026

View reviewed changes

vnkozlov approved these changes Apr 27, 2026

View reviewed changes

openjdk Bot added the ready Pull request is ready to be integrated label Apr 27, 2026

RealFYang approved these changes Apr 28, 2026

View reviewed changes

openjdk Bot added the integrated Pull request has been integrated label Apr 28, 2026

openjdk Bot closed this Apr 28, 2026

openjdk Bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels Apr 28, 2026

shqking deleted the 8380753-regmask branch April 28, 2026 10:35

Uh oh!

Conversation

shqking commented Apr 23, 2026 • 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 23, 2026

Uh oh!

openjdk Bot commented Apr 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openjdk Bot commented Apr 23, 2026

Uh oh!

openjdk Bot commented Apr 23, 2026

Uh oh!

mlbridge Bot commented Apr 23, 2026

Webrevs

Uh oh!

shqking commented Apr 23, 2026

Uh oh!

openjdk Bot commented Apr 23, 2026

Uh oh!

shqking commented Apr 23, 2026

Uh oh!

RealFYang commented Apr 23, 2026

Uh oh!

shqking commented Apr 23, 2026

Uh oh!

anton-seoane commented Apr 24, 2026

Uh oh!

anton-seoane left a comment

Choose a reason for hiding this comment

Uh oh!

vnkozlov left a comment

Choose a reason for hiding this comment

Uh oh!

RealFYang left a comment

Choose a reason for hiding this comment

Uh oh!

shqking commented Apr 28, 2026

Uh oh!

openjdk Bot commented Apr 28, 2026

Uh oh!

openjdk Bot commented Apr 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

4 participants

shqking commented Apr 23, 2026 •

edited by openjdk Bot

Loading

openjdk Bot commented Apr 23, 2026 •

edited

Loading