8352675: Support Intel AVX10 converged vector ISA feature detection #24329

jatin-bhateja · 2025-03-31T13:57:22Z

Intel AVX10[1] extends and enhances the capabilities of Intel AVX-512 to benefit all Intel® products and will be the vector ISA of choice moving into the future.
It supports a new ISA versioning scheme which simplifies the existing AVX512 feature enumeration scheme. Feature set supported by an AVX10 ISA version will be supported by all the versions above it.
The initial, fully-featured version of Intel® AVX10 will be enumerated as Version 2 (denoted as Intel® AVX10.2). This will include the new ISA extension over the existing AVX512 instructions.
An early version of Intel® AVX10 (Version 1, or Intel® AVX10.1) that only enumerates the Intel® AVX-512 instruction set at 128, 256, and 512 bits will be enabled on the Granite Rapids Server for software pre-enabling.

This patch adds the necessary CPUID feature detection for AVX10 ISA version 1 and 2. In terms of architectural state save restoration, AVX10 is isomorphic to AVX512 support up till Granite Rapids. State components affected by AVX10 extension include SSE, AVX, Opmask, ZMM_Hi256, and Hi16_ZMM registers.

The patch has been regressed through tier1 and jvmci tests

Please review and share your feedback.

Best Regards,
Jatin

[1] https://www.intel.com/content/www/us/en/content-details/844829/intel-advanced-vector-extensions-10-2-intel-avx10-2-architecture-specification.html

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-8352675: Support Intel AVX10 converged vector ISA feature detection (Enhancement - P4)

Reviewers

Vladimir Ivanov (@iwanowww - Reviewer) Review applies to c65f0777
Yudi Zheng (@mur47x111 - Committer) Review applies to c65f0777
Sandhya Viswanathan (@sviswa7 - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/24329/head:pull/24329
$ git checkout pull/24329

Update a local copy of the PR:
$ git checkout pull/24329
$ git pull https://git.openjdk.org/jdk.git pull/24329/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 24329

View PR using the GUI difftool:
$ git pr show -t 24329

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/24329.diff

Using Webrev

Link to Webrev Comment

jatin-bhateja · 2025-03-31T13:57:43Z

/label add hotspot-compiler-dev

bridgekeeper · 2025-03-31T13:58:03Z

👋 Welcome back jbhateja! 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 · 2025-03-31T13:58:57Z

@jatin-bhateja 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:

8352675: Support Intel AVX10 converged vector ISA feature detection

Reviewed-by: sviswanathan, vlivanov, yzheng

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 15 new commits pushed to the master branch:

0f2a6c2: 8356577: Migrate ClassFileVersionTest to be feature-agnostic
8fadf29: 8351443: Improve robustness of StringBuilder
68a1185: 8310003: Improve logging when default truststore is inaccessible
... and 12 more: https://git.openjdk.org/jdk/compare/411a63ea1b0c6e8bfea219427bf1c317c5dadabf...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 · 2025-03-31T13:59:22Z

@jatin-bhateja
The hotspot-compiler label was successfully added.

mlbridge · 2025-04-02T19:00:24Z

Webrevs

openjdk · 2025-04-03T02:55:16Z

@jatin-bhateja Please do not rebase or force-push to an active PR as it invalidates existing review comments. Note for future reference, the bots always squash all changes into a single commit automatically as part of the integration. See OpenJDK Developers’ Guide for more information.

eme64

Just leaving a few drive-by comments, I'm really not very familiar with this code. It would be nice if someone from Intel reviewed this also.

Also: you should probably update some more copyright dates ;)

src/jdk.internal.vm.ci/share/classes/jdk/vm/ci/hotspot/HotSpotJVMCIBackendFactory.java

iwanowww · 2025-04-09T19:16:45Z

src/hotspot/cpu/x86/vm_version_x86.cpp

  int res = jio_snprintf(
              buf, sizeof(buf),
              "(%u cores per cpu, %u threads per core) family %d model %d stepping %d microcode 0x%x",
              cores_per_cpu(), threads_per_core(),
              cpu_family(), _model, _stepping, os::cpu_microcode_revision());
  assert(res > 0, "not enough temporary space allocated");
-  insert_features_names(buf + res, sizeof(buf) - res, _features_names);
+  insert_features_names(_features, buf + res, sizeof(buf) - res, _features_names);


x86 is the only platform which uses insert_features_names. Other platforms rely on macros. Maybe it's time to do the same on x86?

iwanowww · 2025-04-09T19:24:06Z

src/hotspot/share/runtime/abstract_vm_version.hpp

@@ -56,6 +56,9 @@ class Abstract_VM_Version: AllStatic {

  // CPU feature flags, can be affected by VM settings.
  static uint64_t _features;
+  // Extra CPU feature flags used when all 64 bits of _features are exhausted for
+  // on a given target, currently only used for x86_64, can be affected by VM settings.
+  static uint64_t _extra_features;


That's unfortunate. Maybe it's time to turn _features into a fixed size (platform-specific) bitmap instead? (RegMask is one existing example.) Having 2 independent fields is error-prone (look at _cpu_features).

src/hotspot/cpu/x86/vm_version_x86.cpp

jatin-bhateja · 2025-04-23T05:42:49Z

/label add graal-dev

openjdk · 2025-04-23T05:44:05Z

@jatin-bhateja
The graal label was successfully added.

iwanowww

It looks much better! Thanks, Jatin.

I'm curious why don't you represent feature bitmap as a POD (with all the accessors on it) and pass it around by value when needed? (It's size will vary across platforms, but will be fixed at runtime.) It should significantly simplify the implementation.

As an example, take a look at RegMask in C2. It accommodates significantly more bits than needed for VM_Version.

jatin-bhateja · 2025-04-24T18:39:04Z

It looks much better! Thanks, Jatin.

I'm curious why don't you represent feature bitmap as a POD (with all the accessors on it) and pass it around by value when needed? (It's size will vary across platforms, but will be fixed at runtime.) It should significantly simplify the implementation.

As an example, take a look at RegMask in C2. It accommodates significantly more bits than needed for VM_Version.

Hi @iwanowww,
RegMask is part of opto code, and it may not be accessible to the JVMCI interface, Currently, JVMCI captures the native address of various fields of VM_Struct, which are of interest to Graal. In the proposed solution, we are adding a new dynamically sized feature vector whose each element is 64 bits wide. JVMCI book-keeps the dynamic feature vector and its size, then uses the UNSAFE access API to compute the enabled feature set on the Java side.

iwanowww · 2025-04-24T22:15:32Z

RegMask is part of opto code, and it may not be accessible to the JVMCI interface

I'm not suggesting to reuse RegMask, but introduce a separate class (e.g., VMFeatures) and embed its instances into Abstract_VM_Version (as VMFeatures _features and VMFeatures _cpu_features). You can keep all the accessors and bit manipulation logic on VMFeatures class.

JVMCI can still operate on in-memory representation at Abstract_VM_Version::_features. But it now needs to query its size (which becomes platform-specific constant).

(BTW all CPU feature constants in AMD64HotSpotVMConfig change their meaning. I don't see any usages in JDK code. Should they go away now?)

merykitty · 2025-05-06T11:47:47Z

src/hotspot/cpu/x86/vm_version_x86.hpp

+
+class VM_Features {
+ public:
+  using FeatureVector = uint64_t [MAX_FEATURE_VEC_SIZE];


Do you think it would be better to refactor this into a separate class analogous to std::bitset? You can start with only implementing test, set, reset. This would help in other use cases, too.

https://en.cppreference.com/w/cpp/utility/bitset

In essence, what we have currently is a bitmap implementation, but its utility is limited to VM_Version for now. The current approach simplifies the JVMCI side of handling. We have an existing utility for bitset src/hotspot/share/utilities/bitMap.hpp, we have multiple implementations for feature detection currently for different targets, it will be good to have the unified solution in the future. For now our intent is just to lift the hard limation of 64 feature bits for x86 target.

src/hotspot/cpu/x86/vm_version_x86.hpp

iwanowww

Very nice!

I made a cleanup pass over the code [1]. Feel free to incorporate it or let me know if you have any questions/concerns.

Meanwhile, submitted it for testing.

[1] iwanowww@35aeb88

src/hotspot/cpu/x86/vm_version_x86.hpp

mur47x111

JVMCI changes look good. Will run some Graal tests on this PR

src/jdk.internal.vm.ci/share/classes/jdk/vm/ci/hotspot/HotSpotJVMCIBackendFactory.java

iwanowww

There are some SA-related failures. Fixed by [1]. Otherwise, testing results are good.

[1] iwanowww@9d4b85a

src/hotspot/cpu/x86/vm_version_x86.hpp

iwanowww

Testing results (hs-tier1 - hs-tier4) are clean.

mur47x111

CPU features in Graal remain the same after this PR. Passed all Graal compiler unit tests.

sviswa7 · 2025-05-08T22:54:51Z

src/hotspot/cpu/x86/vm_version_x86.cpp

@@ -452,13 +461,11 @@ class VM_Version_StubGenerator: public StubCodeGenerator {
    __ lea(rsi, Address(rbp, in_bytes(VM_Version::std_cpuid1_offset())));
    __ movl(rcx, 0x18000000); // cpuid1 bits osxsave | avx
    __ andl(rcx, Address(rsi, 8)); // cpuid1 bits osxsave | avx
-    __ cmpl(rcx, 0x18000000);
-    __ jccb(Assembler::notEqual, done); // jump if AVX is not supported
+    __ jccb(Assembler::equal, done); // jump if AVX is not supported


This and all the following places with multi-bit check still need to be fixed. If you walk through stock and new code in this PR when Address(rsi, 8) on line 468 has 0x10000000, you will observe that stock code will jump to done and new code will not jump to done. Let me know if I am missing something.

test/lib-test/jdk/test/whitebox/CPUInfoTest.java

test/hotspot/jtreg/serviceability/sa/ClhsdbLongConstant.java

sviswa7

Rest of the PR looks good to me.

src/hotspot/cpu/x86/vm_version_x86.cpp

sviswa7

Looks good to me.

jatin-bhateja · 2025-05-09T23:32:21Z

Thanks @iwanowww , @sviswa7 , @mur47x111 , @merykitty for your reviews.

jatin-bhateja · 2025-05-09T23:32:34Z

/integrate

openjdk · 2025-05-09T23:33:33Z

Going to push as commit 3b336a9.
Since your change was applied there have been 15 commits pushed to the master branch:

0f2a6c2: 8356577: Migrate ClassFileVersionTest to be feature-agnostic
8fadf29: 8351443: Improve robustness of StringBuilder
68a1185: 8310003: Improve logging when default truststore is inaccessible
... and 12 more: https://git.openjdk.org/jdk/compare/411a63ea1b0c6e8bfea219427bf1c317c5dadabf...master

Your commit was automatically rebased without conflicts.

openjdk · 2025-05-09T23:33:41Z

@jatin-bhateja Pushed as commit 3b336a9.

💡 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 Mar 31, 2025

jatin-bhateja marked this pull request as ready for review April 2, 2025 18:56

openjdk bot added the rfr Pull request is ready for review label Apr 2, 2025

8352675: Support Intel AVX10 converged vector ISA feature detection

b95ac21

jatin-bhateja force-pushed the JDK-8352675 branch from ff03a06 to b95ac21 Compare April 3, 2025 02:54

graalvmbot mentioned this pull request Apr 3, 2025

[JDK-8353630] Adapt JDK-8352675: Support Intel AVX10 converged vector ISA feature detection oracle/graal#10977

Closed

eme64 reviewed Apr 8, 2025

View reviewed changes

src/jdk.internal.vm.ci/share/classes/jdk/vm/ci/hotspot/HotSpotJVMCIBackendFactory.java Outdated Show resolved Hide resolved

iwanowww reviewed Apr 9, 2025

View reviewed changes

jatin-bhateja marked this pull request as draft April 16, 2025 16:01

openjdk bot removed the rfr Pull request is ready for review label Apr 16, 2025

jatin-bhateja added 3 commits April 17, 2025 08:00

dropping unneeded feature enabling/checks

6a02fe9

Merge branch 'master' of http://github.com/openjdk/jdk into JDK-8352675

5173ff4

Add dynamic sized feature vectors

5d09adb

jatin-bhateja force-pushed the JDK-8352675 branch from 5b3f3b6 to 5d09adb Compare April 23, 2025 05:35

jatin-bhateja marked this pull request as ready for review April 23, 2025 05:40

openjdk bot added the rfr Pull request is ready for review label Apr 23, 2025

openjdk bot added the graal graal-dev@openjdk.org label Apr 23, 2025

iwanowww reviewed Apr 24, 2025

View reviewed changes

Merge branch 'master' of http://github.com/openjdk/jdk into JDK-8352675

862217b

Fix windows build

f413e0e

merykitty reviewed May 6, 2025

View reviewed changes

src/hotspot/cpu/x86/vm_version_x86.hpp Outdated Show resolved Hide resolved

cleanups & refactorings

35aeb88

iwanowww reviewed May 6, 2025

View reviewed changes

src/hotspot/cpu/x86/vm_version_x86.hpp Show resolved Hide resolved

Making _features_bitmap size configurable

cfc09d0

mur47x111 reviewed May 7, 2025

View reviewed changes

src/jdk.internal.vm.ci/share/classes/jdk/vm/ci/hotspot/HotSpotJVMCIBackendFactory.java Show resolved Hide resolved

iwanowww reviewed May 7, 2025

View reviewed changes

src/hotspot/cpu/x86/vm_version_x86.hpp Outdated Show resolved Hide resolved

src/hotspot/cpu/x86/vm_version_x86.hpp Outdated Show resolved Hide resolved

jatin-bhateja added 3 commits May 8, 2025 18:33

Reveiw suggestions incorporated

8acbd7a

Code re-factoring from Vladimir

1a3bce9

Addressing Yudi's comments

c65f077

iwanowww approved these changes May 8, 2025

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label May 8, 2025

mur47x111 approved these changes May 8, 2025

View reviewed changes

sviswa7 reviewed May 9, 2025

View reviewed changes

jatin-bhateja added 2 commits May 9, 2025 19:55

Merge branch 'master' of http://github.com/openjdk/jdk into JDK-8352675

b4838d0

Sandhya's review comments resoultion

f583a52

openjdk bot removed the ready Pull request is ready to be integrated label May 9, 2025

Review comments resolutions

b4654fa

sviswa7 reviewed May 9, 2025

View reviewed changes

src/hotspot/cpu/x86/vm_version_x86.cpp Outdated Show resolved Hide resolved

src/hotspot/cpu/x86/vm_version_x86.cpp Show resolved Hide resolved

sviswa7 approved these changes May 9, 2025

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label May 9, 2025

openjdk bot added the integrated Pull request has been integrated label May 9, 2025

openjdk bot closed this May 9, 2025

openjdk bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels May 9, 2025

graalvmbot mentioned this pull request May 16, 2025

[GR-65034] Update labsjdk to 25+23-jvmci-b01 oracle/graal#11217

Merged

8352675: Support Intel AVX10 converged vector ISA feature detection #24329

8352675: Support Intel AVX10 converged vector ISA feature detection #24329

Uh oh!

Conversation

jatin-bhateja commented Mar 31, 2025 • edited by openjdk bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewers

Reviewing

Uh oh!

jatin-bhateja commented Mar 31, 2025

Uh oh!

bridgekeeper bot commented Mar 31, 2025

Uh oh!

openjdk bot commented Mar 31, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openjdk bot commented Mar 31, 2025

Uh oh!

mlbridge bot commented Apr 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Webrevs

Uh oh!

openjdk bot commented Apr 3, 2025

Uh oh!

eme64 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

iwanowww Apr 9, 2025

Choose a reason for hiding this comment

Uh oh!

iwanowww Apr 9, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jatin-bhateja commented Apr 23, 2025

Uh oh!

openjdk bot commented Apr 23, 2025

Uh oh!

iwanowww left a comment

Choose a reason for hiding this comment

Uh oh!

jatin-bhateja commented Apr 24, 2025

Uh oh!

iwanowww commented Apr 24, 2025

Uh oh!

merykitty May 6, 2025

Choose a reason for hiding this comment

Uh oh!

jatin-bhateja May 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

iwanowww left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

mur47x111 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

iwanowww left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

iwanowww left a comment

Choose a reason for hiding this comment

Uh oh!

mur47x111 left a comment

Choose a reason for hiding this comment

Uh oh!

sviswa7 May 8, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jatin-bhateja commented Mar 31, 2025 •

edited by openjdk bot

Loading

openjdk bot commented Mar 31, 2025 •

edited

Loading

mlbridge bot commented Apr 2, 2025 •

edited

Loading

jatin-bhateja May 6, 2025 •

edited

Loading

iwanowww left a comment •

edited

Loading