Add 128-bit vector support - #80
Open
The-Alchemist wants to merge 1 commit into
Open
The-Alchemist wants to merge 1 commit into
The-Alchemist wants to merge 1 commit into
Conversation
The-Alchemist
changed the title
The-Alchemist
marked this pull request as ready for review
September 17, 2026 21:50
Platforms such as Apple Silicon use 128-bit ByteVector as the preferred species; rejecting that width broke ./gradlew test and made forced 256-bit runs very slow. Implement index128 with four 16-byte lanes per 64-byte block, aligned with simdjson ARM64 stage-1 (simd8x64). Add test128 to CI and check, and wire species-specific test tasks to the test classpath so test128/test256/test512 run the suite with the correct JVM flag. Signed-off-by: The-Alchemist <kap4020@gmail.com>
The-Alchemist
force-pushed
the
feature/128bit-support
branch
from
September 17, 2026 21:51
42ec31d to
afee2c3
Compare
Member
|
Thanks! Have you benchmarked this? There have been attempts to add 128-bit support, but the performance wasn’t great: |
Author
|
i didn't know about #41 , thanks for pointing it out, @piotrrzysko . All this is on a Apple M4 Max. I can try on AWS Graviton too. Benchmark Results (GraalVM, Apple M4 Max)
Benchmark Results (OpenJDK25, Apple M4 Max)
Is it worth adding ARM64 in the CI build matrix? I can submit a separate PR. |
Author
|
got some numbers on AWS Graviton:
I'll try Graviton3 next. Some quick LLM analysis says our solution is not optimal for Graviton's particular SIMD implementation. |
Author
|
Graviton 4.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
org.simdjson.species=128and accept 128-bitpreferredvectors so parsing works on Apple Silicon and similar hardware (tested on M4)StructuralIndexer.index128()using four 16-byte lanes per 64-byte block (same layout as simdjson ARM64simd8x64stage-1).test128to testsFeedback welcome!