8269598: Regressions up to 5% on aarch64 seems due to JDK-8268858 #4766

JoshuaZhuwj · 2021-07-13T12:55:00Z

Hi,

Please review this change.
It fixes the SPECjvm2008 regression on aarch64 caused by JDK-8268858.

Compressor::compress() in case "compress" of SPECjvm2008 performed
degraded by 5% due to extra spillings. There existed an LRG that covered
almost the entire function, hence it interfered with more live ranges.
These interferences caused the LRG no freedom and lastly RA had to spill
it to stack. After the fix, the LRG is split into two shorter live
ranges. Both LRGs are assigned to registers due to fewer interferences.

See details at: http://cr.openjdk.java.net/~jzhu/8269598/analysis.pdf
OptoAssembly: http://cr.openjdk.java.net/~jzhu/8269598/

The root cause is:
When PreserveFramePointer is off, frame pointer is allocatable, but
different from other SOC registers, it is excluded from fatproj's mask
because its save type is No-Save. Therefore fatproj was not treated as
high-pressure transition point in the previous int_pressure_limit()
computation. This fix ensures high pressure at fatproj when
PreserveFramePointer is off. See check_pressure_at_fatproj() [1].

After this fix, the regression disappeared.
The result of regression cases in SPECjvm2008:
http://cr.openjdk.java.net/~jzhu/8269598/Book1.pdf

[1]

jdk/src/hotspot/share/opto/chaitin.hpp

Line 620 in 375fc2a

void check_pressure_at_fatproj(uint fatproj_location, RegMask& fatproj_mask) {

Best Regards,
Joshua

Progress

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

Issue

JDK-8269598: Regressions up to 5% on aarch64 seems due to JDK-8268858

Reviewers

Vladimir Kozlov (@vnkozlov - Reviewer)
Dean Long (@dean-long - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.java.net/jdk pull/4766/head:pull/4766
$ git checkout pull/4766

Update a local copy of the PR:
$ git checkout pull/4766
$ git pull https://git.openjdk.java.net/jdk pull/4766/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 4766

View PR using the GUI difftool:
$ git pr show -t 4766

Using diff file

Download this PR as a diff file:
https://git.openjdk.java.net/jdk/pull/4766.diff

bridgekeeper · 2021-07-13T12:55:27Z

👋 Welcome back jzhu! 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 · 2021-07-13T12:57:52Z

@JoshuaZhuwj 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.

mlbridge · 2021-07-13T13:01:20Z

Webrevs

00: Full (d3561e17)

vnkozlov

Good.

openjdk · 2021-07-13T16:09:58Z

@JoshuaZhuwj 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:

8269598: Regressions up to 5% on aarch64 seems due to JDK-8268858

Reviewed-by: kvn, dlong

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 39 new commits pushed to the master branch:

7d0edb5: Merge
7b4d84c: 8270422: Test build/AbsPathsInImage.java fails after JDK-8259848
72db09b: 8266313: (JEP-356) - RandomGenerator spec implementation requirements tightly coupled to JDK internal classes
3bbd233: 8270075: SplittableRandom extends AbstractSplittableGenerator
381bd62: 8266889: [macosx-aarch64] Crash with SIGBUS in MarkActivationClosure::do_code_blob during vmTestbase/nsk/jvmti/.../bi04t002 test run
82c256e: 8259499: Handling type arguments from outer classes for inner class in javadoc
e5db9a9: 8268620: InfiniteLoopException test may fail on x86 platforms
67273ae: 8269865: Async UL needs to handle ERANGE on exceeding SEM_VALUE_MAX
0f54707: 8270056: Generated lambda class can not access protected static method of target class
a033866: 8269637: javax/swing/JFileChooser/FileSystemView/SystemIconTest.java fails on windows
... and 29 more: https://git.openjdk.java.net/jdk/compare/07e90524576f159fc16523430f1db62327c89a3b...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 (@vnkozlov, @dean-long) 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).

dean-long

I don't understand why aarcht64 needs these adjustments but x86 does not. Both are doing the same tricks with PreserveFramePointer, right?

JoshuaZhuwj · 2021-07-14T09:11:39Z

I don't understand why aarcht64 needs these adjustments but x86 does not. Both are doing the same tricks with PreserveFramePointer, right?

Yes. If make the same adjustment on x86, the same wide LRG in
Compressor::compress() will also be split and be assigned to register
just like aarch64.

But this adjustment is a trade-off in my opinion. Lowering pressure
threshold would also bring unnecessary spillings in some scenarios,
result in performance degradation especially on x86 platform that has
a small quantity of GPRs. On aarch64 that has more GPRs, the benefits
of lowering pressure threshold by 1 or 2 outweigh the disadvantages.

I also performed the same SPECjvm2018 "compress" test on x86_64.
Compared to 5% performance improvement on aarch64, the same adjustment
degraded performance by 1.28% nevertheless.

JoshuaZhuwj · 2021-07-15T01:49:07Z

Thanks @vnkozlov @dean-long for your review.

JoshuaZhuwj · 2021-07-15T01:49:39Z

/integrate

openjdk · 2021-07-15T01:50:19Z

@JoshuaZhuwj
Your change (at version d3561e1) is now ready to be sponsored by a Committer.

y1yang0 · 2021-07-15T01:55:28Z

/sponsor

openjdk · 2021-07-15T01:57:42Z

Going to push as commit 7c23491.
Since your change was applied there have been 39 commits pushed to the master branch:

7d0edb5: Merge
7b4d84c: 8270422: Test build/AbsPathsInImage.java fails after JDK-8259848
72db09b: 8266313: (JEP-356) - RandomGenerator spec implementation requirements tightly coupled to JDK internal classes
3bbd233: 8270075: SplittableRandom extends AbstractSplittableGenerator
381bd62: 8266889: [macosx-aarch64] Crash with SIGBUS in MarkActivationClosure::do_code_blob during vmTestbase/nsk/jvmti/.../bi04t002 test run
82c256e: 8259499: Handling type arguments from outer classes for inner class in javadoc
e5db9a9: 8268620: InfiniteLoopException test may fail on x86 platforms
67273ae: 8269865: Async UL needs to handle ERANGE on exceeding SEM_VALUE_MAX
0f54707: 8270056: Generated lambda class can not access protected static method of target class
a033866: 8269637: javax/swing/JFileChooser/FileSystemView/SystemIconTest.java fails on windows
... and 29 more: https://git.openjdk.java.net/jdk/compare/07e90524576f159fc16523430f1db62327c89a3b...master

Your commit was automatically rebased without conflicts.

openjdk · 2021-07-15T01:57:59Z

@kelthuzadx @JoshuaZhuwj Pushed as commit 7c23491.

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

dean-long · 2021-07-15T20:39:18Z

@JoshuaZhuwj Please try SPECjvm2008 compress on x86/x64 with -XX:+PreserveFramePointer. We are seeing a regression after JDK-8268858.

JoshuaZhuwj · 2021-07-16T02:27:44Z

@dean-long Thanks for your information. Previously I already found performance degradation with -XX:+PreserveFramePointer in Alibaba Flink Nexmark benchmark. It's caused by inefficient codes generated by RA. That's why I begin working on register pressure. I already created JDK-8270808 [1] to track it. Meanwhile I will also track your mentioned regression issue together with it. Hope they're caused by the same reason.

[1] https://bugs.openjdk.java.net/browse/JDK-8270808

JoshuaZhuwj · 2021-07-16T12:20:08Z

@JoshuaZhuwj Please try SPECjvm2008 compress on x86/x64 with -XX:+PreserveFramePointer. We are seeing a regression after JDK-8268858.

Just now I ran six rounds of "compress" on x64 machine with "-XX:+PreserveFramePointer".
Thread number of "compress" test is configured as 1 in my testing environment.
See result at http://cr.openjdk.java.net/~jzhu/8269598/Book2_x86.pdf
There is no regression in my testing. How about yours?

8269598: Regressions up to 5% on aarch64 seems due to JDK-8268858

d3561e1

openjdk bot added the rfr Pull request is ready for review label Jul 13, 2021

openjdk bot added the hotspot-compiler hotspot-compiler-dev@openjdk.org label Jul 13, 2021

vnkozlov approved these changes Jul 13, 2021

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Jul 13, 2021

dean-long suggested changes Jul 13, 2021

View reviewed changes

dean-long approved these changes Jul 14, 2021

View reviewed changes

openjdk bot added the sponsor Pull request is ready to be sponsored label Jul 15, 2021

openjdk bot closed this Jul 15, 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 Jul 15, 2021

8269598: Regressions up to 5% on aarch64 seems due to JDK-8268858 #4766

8269598: Regressions up to 5% on aarch64 seems due to JDK-8268858 #4766

JoshuaZhuwj commented Jul 13, 2021 •

edited by openjdk bot

bridgekeeper bot commented Jul 13, 2021

openjdk bot commented Jul 13, 2021

mlbridge bot commented Jul 13, 2021

vnkozlov left a comment

openjdk bot commented Jul 13, 2021 •

edited

dean-long left a comment

JoshuaZhuwj commented Jul 14, 2021

JoshuaZhuwj commented Jul 15, 2021

JoshuaZhuwj commented Jul 15, 2021

openjdk bot commented Jul 15, 2021

y1yang0 commented Jul 15, 2021

openjdk bot commented Jul 15, 2021

openjdk bot commented Jul 15, 2021

dean-long commented Jul 15, 2021

JoshuaZhuwj commented Jul 16, 2021

JoshuaZhuwj commented Jul 16, 2021

8269598: Regressions up to 5% on aarch64 seems due to JDK-8268858 #4766

8269598: Regressions up to 5% on aarch64 seems due to JDK-8268858 #4766

Conversation

JoshuaZhuwj commented Jul 13, 2021 • edited by openjdk bot

Progress

Issue

Reviewers

Reviewing

bridgekeeper bot commented Jul 13, 2021

openjdk bot commented Jul 13, 2021

mlbridge bot commented Jul 13, 2021

Webrevs

vnkozlov left a comment

Choose a reason for hiding this comment

openjdk bot commented Jul 13, 2021 • edited

dean-long left a comment

Choose a reason for hiding this comment

JoshuaZhuwj commented Jul 14, 2021

JoshuaZhuwj commented Jul 15, 2021

JoshuaZhuwj commented Jul 15, 2021

openjdk bot commented Jul 15, 2021

y1yang0 commented Jul 15, 2021

openjdk bot commented Jul 15, 2021

openjdk bot commented Jul 15, 2021

dean-long commented Jul 15, 2021

JoshuaZhuwj commented Jul 16, 2021

JoshuaZhuwj commented Jul 16, 2021

JoshuaZhuwj commented Jul 13, 2021 •

edited by openjdk bot

openjdk bot commented Jul 13, 2021 •

edited