8370822: Improve performance for MemorySegment::fill

[hextriclosan]

The manual byte-replication expression replaced with the multiplicative expansion trick - a more compact and efficient equivalent.

---

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-8370822: Improve performance for MemorySegment::fill (Enhancement - P4)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/28031/head:pull/28031
$ git checkout pull/28031

Update a local copy of the PR:
$ git checkout pull/28031
$ git pull https://git.openjdk.org/jdk.git pull/28031/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 28031

View PR using the GUI difftool:
$ git pr show -t 28031

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/28031.diff

Using Webrev

Link to Webrev Comment

[bridgekeeper]

👋 Welcome back hextriclosan! 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]

❗ This change is not yet ready to be integrated.
See the Progress checklist in the description for automated requirements.

[openjdk]

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

• core-libs

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.

[hextriclosan]

/solves JDK-8370822

[openjdk]

@hextriclosan This issue is referenced in the PR title - it will now be updated.

[mlbridge]

Webrevs

• 00: Full (e3016f39)

[minborg]

I ran some benchmarks and was unable to see any significant performance improvements. I ran on Linux (a64/x64), macOS a64, and Windows x64. Can you provide benchmarks that support this is a performance improvement?

Benchmark	Linux aarch64	Linux x64	MacOSX aarch64	Windows x64
openjdk.bench.java.lang.foreign.SegmentBulkFill.nativeSegmentFillJava-ELEM_SIZE:2	-0.05%	-4.60%	-0.12%	-0.22%
openjdk.bench.java.lang.foreign.SegmentBulkFill.nativeSegmentFillJava-ELEM_SIZE:3	-0.01%	-0.17%	-1.18%	0.04%
openjdk.bench.java.lang.foreign.SegmentBulkFill.nativeSegmentFillJava-ELEM_SIZE:4	-0.01%	0.00%	-0.01%	0.28%
openjdk.bench.java.lang.foreign.SegmentBulkFill.nativeSegmentFillJava-ELEM_SIZE:5	0.35%	-0.00%	0.04%	0.18%
openjdk.bench.java.lang.foreign.SegmentBulkFill.nativeSegmentFillJava-ELEM_SIZE:512	-0.17%	-0.84%	1.57%	-0.05%
openjdk.bench.java.lang.foreign.SegmentBulkFill.nativeSegmentFillJava-ELEM_SIZE:6	0.00%	0.02%	-0.01%	0.36%
openjdk.bench.java.lang.foreign.SegmentBulkFill.nativeSegmentFillJava-ELEM_SIZE:64	-0.03%	-0.75%	-5.88%	-2.61%
openjdk.bench.java.lang.foreign.SegmentBulkFill.nativeSegmentFillJava-ELEM_SIZE:7	-0.11%	-0.25%	-0.02%	0.16%
openjdk.bench.java.lang.foreign.SegmentBulkFill.nativeSegmentFillJava-ELEM_SIZE:8	-0.01%	-0.20%	-0.07%	0.15%

[hextriclosan]

I ran some benchmarks and was unable to see any significant performance improvements. I ran on Linux (a64/x64), macOS a64, and Windows x64. Can you provide benchmarks that support this is a performance improvement?

Hello Per-Ake,

I explored the corresponding code using JITWatch. My analysis shows that the patch reduces the number of instructions (in both bytecode and JIT-compiled assembly). However, as your benchmarks demonstrate, this does not lead to measurable overall performance improvement, the differences fall within normal noise.

So, the idea for the patch may be clarity and maintainability, as it simplifies and shortens the code.

Thanks.

[minborg]

Hello Per-Ake,

I explored the corresponding code using JITWatch. My analysis shows that the patch reduces the number of instructions (in both bytecode and JIT-compiled assembly). However, as your benchmarks demonstrate, this does not lead to measurable overall performance improvement, the differences fall within normal noise.

So, the idea for the patch may be clarity and maintainability, as it simplifies and shortens the code.

Thanks.

Interesting observation. I think reducing bytecode is also important. Are you able to share some JIT-compiled assembly before/after?

[hextriclosan]

Here is syntactic example

Before

After