8271128: InlineIntrinsics support for 32-bit ARM #4927

mychris · 2021-07-29T09:40:08Z

Hi,

please review this patch, which adds support for InlineIntrinsics to the 32-bit ARM port. The old aarch32 port had this intrinsic implemented and enabled by default.

Like on many other platforms, the 32-bit ARM port simply calls into the SharedRuntime to intrinsify the basic java.lang.Math methods. InlineIntrinsics is already implemented for C1 on 32-bit ARM, which does the same thing.

testing: hotspot tier1 on ARMv5TE (soft-float) and ARMv7-A (hard-float)

There is already the micro benchmark test/micro/org/openjdk/bench/java/lang/MathBench.java which I used. The soft-float benchmarks are not that meaningful, since I performed them in QEMU.

hard-float -Xint -XX:+InlineIntrinsics

Benchmark	Mode	Cnt	Score	Error	Units
MathBench.absDouble	thrpt	5	1169.574	+/- 133.694	ops/ms
MathBench.cosDouble	thrpt	5	759.902	+/- 573.852	ops/ms
MathBench.expDouble	thrpt	5	854.753	+/- 67.217	ops/ms
MathBench.log10Double	thrpt	5	902.034	+/- 22.413	ops/ms
MathBench.logDouble	thrpt	5	895.470	+/- 113.811	ops/ms
MathBench.powDouble	thrpt	5	936.136	+/- 40.661	ops/ms
MathBench.sinDouble	thrpt	5	864.670	+/- 68.329	ops/ms
MathBench.sqrtDouble	thrpt	5	1082.589	+/- 92.570	ops/ms
MathBench.tanDouble	thrpt	5	853.715	+/- 122.427	ops/ms

hard-float -Xint -XX:-InlineIntrinsics

Benchmark	Mode	Cnt	Score	Error	Units
MathBench.absDouble	thrpt	5	450.907	+/- 10.402	ops/ms
MathBench.cosDouble	thrpt	5	592.242	+/- 14.011	ops/ms
MathBench.expDouble	thrpt	5	167.614	+/- 7.530	ops/ms
MathBench.log10Double	thrpt	5	572.099	+/- 55.089	ops/ms
MathBench.logDouble	thrpt	5	596.588	+/- 24.976	ops/ms
MathBench.powDouble	thrpt	5	212.673	+/- 4.060	ops/ms
MathBench.sinDouble	thrpt	5	584.873	+/- 42.774	ops/ms
MathBench.sqrtDouble	thrpt	5	514.690	+/- 30.568	ops/ms
MathBench.tanDouble	thrpt	5	566.586	+/- 23.995	ops/ms

soft-float -Xint -XX:+InlineIntrinsics

Benchmark	Mode	Cnt	Score	Error	Units
MathBench.absDouble	thrpt	5	279.575	+/- 56.455	ops/ms
MathBench.cosDouble	thrpt	5	137.005	+/- 72.561	ops/ms
MathBench.expDouble	thrpt	5	117.778	+/- 30.186	ops/ms
MathBench.log10Double	thrpt	5	107.957	+/- 10.158	ops/ms
MathBench.logDouble	thrpt	5	101.341	+/- 3.914	ops/ms
MathBench.powDouble	thrpt	5	222.220	+/- 3.854	ops/ms
MathBench.sinDouble	thrpt	5	112.715	+/- 9.088	ops/ms
MathBench.sqrtDouble	thrpt	5	119.341	+/- 76.528	ops/ms
MathBench.tanDouble	thrpt	5	105.224	+/- 30.477	ops/ms

soft-float -Xint -XX:-InlineIntrinsics

Benchmark	Mode	Cnt	Score	Error	Units
MathBench.absDouble	thrpt	5	173.150	+/- 36.279	ops/ms
MathBench.cosDouble	thrpt	5	129.774	+/- 8.795	ops/ms
MathBench.expDouble	thrpt	5	53.524	+/- 1.679	ops/ms
MathBench.log10Double	thrpt	5	132.503	+/- 4.274	ops/ms
MathBench.logDouble	thrpt	5	135.483	+/- 1.150	ops/ms
MathBench.powDouble	thrpt	5	54.266	+/- 0.699	ops/ms
MathBench.sinDouble	thrpt	5	105.636	+/- 4.647	ops/ms
MathBench.sqrtDouble	thrpt	5	204.550	+/- 7.206	ops/ms
MathBench.tanDouble	thrpt	5	101.072	+/- 3.701	ops/ms

Progress

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

Issue

JDK-8271128: InlineIntrinsics support for 32-bit ARM

Reviewers

Aleksey Shipilev (@shipilev - Reviewer) ⚠️ Review applies to 3c20c98

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.java.net/jdk pull/4927/head:pull/4927
$ git checkout pull/4927

Update a local copy of the PR:
$ git checkout pull/4927
$ git pull https://git.openjdk.java.net/jdk pull/4927/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 4927

View PR using the GUI difftool:
$ git pr show -t 4927

Using diff file

Download this PR as a diff file:
https://git.openjdk.java.net/jdk/pull/4927.diff

bridgekeeper · 2021-07-29T09:41:23Z

👋 Welcome back cgo! 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 · 2021-07-29T09:43:30Z

@mychris 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.

mychris · 2021-07-29T09:44:20Z

/cc hotspot-runtime

openjdk · 2021-07-29T09:45:04Z

@mychris
The hotspot-runtime label was successfully added.

mlbridge · 2021-07-29T09:48:44Z

Webrevs

shipilev

This looks okay to me, but I think hotspot-tier1 testing is inadequate for this. Please run at least full tier1, which should include JDK tests for java.lang.Math.

mychris · 2021-07-29T11:58:12Z

Sorry, I forgot to mention this.
I tested the java.lang.Math methods for which the new intrinsics are implemented manually, by comparing the results of the intrinsics with the results of the corresponding java.lang.StrictMath method. Since both, StrictMath and the intrinsics use the same algorithm on 32-bit ARM, it is possible to do that.
But I agree, doing some more testing doesn't hurt. I will start jdk tier1 as well and will report back as soon as they are done, will definitely take some time though.

mychris · 2021-08-04T07:31:06Z

jdk:tier1 tests on soft-float and hard-float are good. IeeeRecommendedTests on soft-float has some issues, but this is because I don't use an external soft float library, but the glibc implementation for floating point arithmetic, which doesn't have the required precision. Probably one should pick up JDK-8215902 again and fix this situation.

shipilev · 2021-08-04T07:54:54Z

jdk:tier1 tests on soft-float and hard-float are good.

Thanks, we are good then.

IeeeRecommendedTests on soft-float has some issues, but this is because I don't use an external soft float library, but the glibc implementation for floating point arithmetic, which doesn't have the required precision.

This is pre-existing, right? Meaning, it is not worse with this PR?

mychris · 2021-08-04T08:01:43Z

IeeeRecommendedTests on soft-float has some issues, but this is because I don't use an external soft float library, but the glibc implementation for floating point arithmetic, which doesn't have the required precision.

This is pre-existing, right? Meaning, it is not worse with this PR?

Yes, this PR doesn't change anything. The test case also uses the Math and StrictMath functions and the problem is identical for both implementations.

shipilev

Good to go then.

openjdk · 2021-08-04T08:04:29Z

@mychris 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:

8271128: InlineIntrinsics support for 32-bit ARM

Reviewed-by: shade

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 17 new commits pushed to the master branch:

c2b7fac: 8271896: Remove unnecessary top address checks in BOT
e7b6f48: 8265602: -XX:DumpLoadedClassList should support custom loaders
ea02dad: 8272067: Initial nroff manpage generation for JDK 18
adb0ae5: 8261441: JFR: Filename expansion
e38e365: 8271208: Typo in ModuleDescriptor.read javadoc
14692d5: Merge
dfacda4: 8270872: Final nroff manpage update for JDK 17
90f85ff: 8271588: JFR Recorder Thread crashed with SIGSEGV in write_klass
f312f28: 8271863: ProblemList serviceability/sa/TestJmapCore.java on linux-x64 with ZGC
62e72ad: 8271293: Monitor class should use ThreadBlockInVMPreprocess
... and 7 more: https://git.openjdk.java.net/jdk/compare/685fc3c677cd0e71ef4443214ae14c7eed355140...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.

As you do not have Committer status in this project an existing Committer must agree to sponsor your change. Possible candidates are the reviewers of this PR (@shipilev) but any other Committer may sponsor as well.

➡️ To flag this PR as ready for integration with the above commit message, type /integrate in a new comment. (Afterwards, your sponsor types /sponsor in a new comment to perform the integration).

mychris · 2021-08-05T09:22:10Z

Thanks for the review Aleksey, very much appreciated.
/integrate

openjdk · 2021-08-05T09:23:07Z

@mychris
Your change (at version d8926df) is now ready to be sponsored by a Committer.

shipilev · 2021-08-05T09:36:48Z

Sorry for not catching this earlier, but I think java_lang_math_abs and java_lang_math_sqrt are not transcendental, so it is awkward to call generate_transcendental_entry for them? I think that's just a naming problem, maybe call it generate_math_runtime_entry? (If you fix this, maybe also pull from master to get clean GHA runs).

* master: (97 commits) 8270903: sun.net.httpserver.HttpConnection: Improve toString 8271722: [TESTBUG] gc/g1/TestMixedGCLiveThreshold.java can fail if G1 Full GC uses >1 workers 8270058: Use Objects.check{Index,FromIndexSize} for java.desktop 4819544: SwingSet2 JTable Demo throws NullPointerException 8271878: UnProblemList jdk/jfr/event/gc/detailed/TestEvacuationFailedEvent.java in JDK18 8271895: UnProblemList javax/swing/JComponent/7154030/bug7154030.java in JDK18 8271863: ProblemList serviceability/sa/TestJmapCore.java on linux-x64 with ZGC 8271898: disable os.release_multi_mappings_vm on macOS-X64 8271893: mark hotspot runtime/PerfMemDestroy/PerfMemDestroy.java test as ignoring external VM flags 8271887: mark hotspot runtime/CDSCompressedKPtrs tests which ignore external VM flags 8271891: mark hotspot runtime/Safepoint tests which ignore external VM flags 8271886: mark hotspot runtime/InvocationTests tests which ignore external VM flags 8271890: mark hotspot runtime/Dictionary tests which ignore external VM flags 8271894: ProblemList javax/swing/JComponent/7154030/bug7154030.java in JDK17 8271888: build error after JDK-8271599 8271456: Avoid looking up standard charsets in "java.desktop" module 8271589: fatal error with variable shift count integer rotate operation. 8271599: Javadoc of floorDiv() and floorMod() families is inaccurate in some places 8271877: ProblemList jdk/jfr/event/gc/detailed/TestEvacuationFailedEvent.java in JDK17 8271217: Fix race between G1PeriodicGCTask checks and GC request ...

mychris · 2021-08-05T12:27:20Z

Sorry for not catching this earlier, but I think java_lang_math_abs and java_lang_math_sqrt are not transcendental, so it is awkward to call generate_transcendental_entry for them? I think that's just a naming problem, maybe call it generate_math_runtime_entry? (If you fix this, maybe also pull from master to get clean GHA runs).

I chose that name because of the aarch64 implementation, which uses the same name (but doesn't generate calls for neither abs, nor sqrt). I renamed it to generate_math_runtime_call, since it only generates the call to the runtime function.

shipilev · 2021-08-05T15:52:20Z

Sorry for not catching this earlier, but I think java_lang_math_abs and java_lang_math_sqrt are not transcendental, so it is awkward to call generate_transcendental_entry for them? I think that's just a naming problem, maybe call it generate_math_runtime_entry? (If you fix this, maybe also pull from master to get clean GHA runs).

I chose that name because of the aarch64 implementation, which uses the same name (but doesn't generate calls for neither abs, nor sqrt). I renamed it to generate_math_runtime_call, since it only generates the call to the runtime function.

Thanks, that looks better. transcendental_entry local variable too?

mychris · 2021-08-06T08:04:01Z

Thanks, that looks better. transcendental_entry local variable too?

Sorry, should have seen this myself.
Done.

shipilev

I'd prefer use_runtime_function to be called use_runtime_call, but this is fine as well. I can sponsor this for you.

mychris · 2021-08-06T08:44:42Z

I'd prefer use_runtime_function to be called use_runtime_call

Done.

I can sponsor this for you.

Thanks, I guess I have to integrate again?

/integrate

openjdk · 2021-08-06T08:46:04Z

@mychris
Your change (at version 6de922f) is now ready to be sponsored by a Committer.

shipilev · 2021-08-06T08:46:57Z

Yup, I'll wait for GHA to clear the ARM build, and then sponsor.

shipilev · 2021-08-06T10:22:13Z

ARM builds are clean.

/sponsor

openjdk · 2021-08-06T10:23:25Z

Going to push as commit b6a19f1.
Since your change was applied there have been 17 commits pushed to the master branch:

c2b7fac: 8271896: Remove unnecessary top address checks in BOT
e7b6f48: 8265602: -XX:DumpLoadedClassList should support custom loaders
ea02dad: 8272067: Initial nroff manpage generation for JDK 18
adb0ae5: 8261441: JFR: Filename expansion
e38e365: 8271208: Typo in ModuleDescriptor.read javadoc
14692d5: Merge
dfacda4: 8270872: Final nroff manpage update for JDK 17
90f85ff: 8271588: JFR Recorder Thread crashed with SIGSEGV in write_klass
f312f28: 8271863: ProblemList serviceability/sa/TestJmapCore.java on linux-x64 with ZGC
62e72ad: 8271293: Monitor class should use ThreadBlockInVMPreprocess
... and 7 more: https://git.openjdk.java.net/jdk/compare/685fc3c677cd0e71ef4443214ae14c7eed355140...master

Your commit was automatically rebased without conflicts.

openjdk · 2021-08-06T10:23:37Z

@shipilev @mychris Pushed as commit b6a19f1.

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

8271128: InlineIntrinsics support for 32-bit ARM

d8926df

openjdk bot added the rfr Pull request is ready for review label Jul 29, 2021

openjdk bot added the hotspot hotspot-dev@openjdk.org label Jul 29, 2021

openjdk bot added the hotspot-runtime hotspot-runtime-dev@openjdk.org label Jul 29, 2021

shipilev reviewed Jul 29, 2021

View reviewed changes

shipilev approved these changes Aug 4, 2021

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Aug 4, 2021

openjdk bot added the sponsor Pull request is ready to be sponsored label Aug 5, 2021

mychris added 2 commits August 5, 2021 13:26

Rename generate_transcendental_entry -> generate_math_runtime_call

faadc2d

openjdk bot removed the sponsor Pull request is ready to be sponsored label Aug 5, 2021

Rename transcendental_entry -> use_runtime_function

3c20c98

shipilev approved these changes Aug 6, 2021

View reviewed changes

Rename use_runtime_function -> use_runtime_call

6de922f

openjdk bot added the sponsor Pull request is ready to be sponsored label Aug 6, 2021

openjdk bot closed this Aug 6, 2021

openjdk bot added integrated Pull request has been integrated and removed ready Pull request is ready to be integrated rfr Pull request is ready for review sponsor Pull request is ready to be sponsored labels Aug 6, 2021

8271128: InlineIntrinsics support for 32-bit ARM #4927

8271128: InlineIntrinsics support for 32-bit ARM #4927

Uh oh!

Conversation

mychris commented Jul 29, 2021 • 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!

bridgekeeper bot commented Jul 29, 2021

Uh oh!

openjdk bot commented Jul 29, 2021

Uh oh!

mychris commented Jul 29, 2021

Uh oh!

openjdk bot commented Jul 29, 2021

Uh oh!

mlbridge bot commented Jul 29, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Webrevs

Uh oh!

shipilev left a comment

Choose a reason for hiding this comment

Uh oh!

mychris commented Jul 29, 2021

Uh oh!

mychris commented Aug 4, 2021

Uh oh!

shipilev commented Aug 4, 2021

Uh oh!

mychris commented Aug 4, 2021

Uh oh!

shipilev left a comment

Choose a reason for hiding this comment

Uh oh!

openjdk bot commented Aug 4, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mychris commented Aug 5, 2021

Uh oh!

openjdk bot commented Aug 5, 2021

Uh oh!

shipilev commented Aug 5, 2021

Uh oh!

mychris commented Aug 5, 2021

Uh oh!

shipilev commented Aug 5, 2021

Uh oh!

mychris commented Aug 6, 2021

Uh oh!

shipilev left a comment

Choose a reason for hiding this comment

Uh oh!

mychris commented Aug 6, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openjdk bot commented Aug 6, 2021

Uh oh!

shipilev commented Aug 6, 2021

Uh oh!

shipilev commented Aug 6, 2021

Uh oh!

openjdk bot commented Aug 6, 2021

Uh oh!

openjdk bot commented Aug 6, 2021

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

mychris commented Jul 29, 2021 •

edited by openjdk bot

Loading

mlbridge bot commented Jul 29, 2021 •

edited

Loading

openjdk bot commented Aug 4, 2021 •

edited

Loading

mychris commented Aug 6, 2021 •

edited

Loading