Chris Hegarty comments

Results 113 comments of


                                            Chris Hegarty

Integrating GPU based Vector Search using cuVS

> P.P.S. Elasticsearch has a Gradle plugin to strip preview flags. Basically it patches one byte in all class files that are created by Javac. Yes, we can do this....

Integrating GPU based Vector Search using cuVS

FYI - I've made the cuvs-java api Java 21 friendly, with an spi and a java-22 specific impl in the versioned section of an mrjar - MemorySegment and Arena have...

Integrating GPU based Vector Search using cuVS

> I think bumping main only for each non LTS release would be cool. Then we keep it at the next LTS (Java 25)? I filed the following issue to...

Integrating GPU based Vector Search using cuVS

I just committed a rewrite for the cuVS format implementation. After the rewrite all the BaseKnnVectorsFormatTestCase tests pass. There are still some lurking intermittent failures, but the tests pass successfully...

Examine performance of individual data accessor methods of MemorySegmentIndexInput when IndexInputs are closed in other threads (deoptimizations,...)

Thanks for your quick reply @uschindler. Your analysis makes sense to me.

Examine performance of individual data accessor methods of MemorySegmentIndexInput when IndexInputs are closed in other threads (deoptimizations,...)

Thanks for jumping on this so quickly @uschindler.

Examine performance of individual data accessor methods of MemorySegmentIndexInput when IndexInputs are closed in other threads (deoptimizations,...)

> or at least Arena.ofShared() -- at all. This is the issue - the __shared__ Arena. Can we please get some perf numbers with Arena.ofConfined(), for this particular use-case.

Examine performance of individual data accessor methods of MemorySegmentIndexInput when IndexInputs are closed in other threads (deoptimizations,...)

> Here is some response I sent: https://mail.openjdk.org/pipermail/panama-dev/2024-July/020535.html Thanks @uschindler - this is very helpful. > Actually the ideas I have: > > * Use READ_ONCE in combination with single...

New JMH benchmark method - vdot8s that implement int8 dotProduct in C…

> > With the updated compile flags, the performance of auto-vectorized code is slightly better than explicitly vectorized code (see results). Interesting thing to note is that both C-based implementations...

New JMH benchmark method - vdot8s that implement int8 dotProduct in C…

Could Lucene ever have this directly in one of its modules? We currently use the `FlatVectorsScorer` to plugin the "native code optimized" alternative, when scoring Scalar Quantized vectors. But Lucene...