kaivalnp commented on issue #15697:
URL: https://github.com/apache/lucene/issues/15697#issuecomment-3931276740

   > Seeing the same kind of warmup behavior on a Mac M2
   
   Ah thanks, my machine was an AWS Graviton3 -- so this issue(?) may be 
related to ARM
   
   Something strange: the degradation does not occur if we separate 
[this](https://github.com/apache/lucene/blob/e1879e450b75b3a58fde2b0dad77ae6b499504dd/lucene/core/src/java25/org/apache/lucene/internal/vectorization/PanamaVectorUtilSupport.java#L596-L612)
 into two separate loops (upper and lower):
   
   ```
   Benchmark                                                       (size)   
Mode  Cnt   Score   Error   Units
   VectorUtilBenchmark.binaryHalfByteDotProductSinglePackedScalar    1024  
thrpt   15   2.447 ± 0.008  ops/us
   VectorUtilBenchmark.binaryHalfByteDotProductSinglePackedVector    1024  
thrpt   15  13.275 ± 0.341  ops/us
   ```
   
   ..which seems counter intuitive, because the "packed" vector is loaded twice 
(perhaps C2 compiles it "correctly" under the hood).
   
   I'm also unsure if this is an actual issue, or just occurs in JMH 
benchmarks..


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to