shbhar commented on PR #16859:
URL: https://github.com/apache/lucene/pull/16859#issuecomment-6275209858

   The size & complexity of this change is still iffy, so trying to make sure 
everything has earned its place and see what can be removed or simplified.
   
   Starting with XXHash3 64 vs existing Murmur3 128 implementation (take the 
first long for 64 bit hash)
   
   JMH results (5 forks, 10 iterations, 1s each with 5 iterations warmup) on 
JDK25
   
   ```
   Graviton4 (c8g.metal-24xl)
   
----------------------------------------------------------------------------------------------------------------------------------
   Benchmark                                            (function)  (keyShape)  
 Mode  Cnt    Score   Error   Units
   PrimaryKeyHashFunctionBenchmark.throughput                 xxh3      uuid16  
thrpt   50  377.272 ± 0.101  ops/us
   PrimaryKeyHashFunctionBenchmark.throughput  murmur3StringHelper      uuid16  
thrpt   50   99.298 ± 0.176  ops/us
   PrimaryKeyHashFunctionBenchmark.latency                    xxh3      uuid16  
 avgt   50   10.074 ± 0.009   ns/op
   PrimaryKeyHashFunctionBenchmark.latency     murmur3StringHelper      uuid16  
 avgt   50   17.684 ± 0.055   ns/op
   
   
----------------------------------------------------------------------------------------------------------------------------------
   Intel Xeon 6975P-C (c8i.metal-48xl)
   Benchmark                                            (function)  (keyShape)  
 Mode  Cnt    Score   Error   Units
   PrimaryKeyHashFunctionBenchmark.throughput                 xxh3      uuid16  
thrpt   50  506.192 ± 4.584  ops/us
   PrimaryKeyHashFunctionBenchmark.throughput  murmur3StringHelper      uuid16  
thrpt   50  144.471 ± 0.229  ops/us
   PrimaryKeyHashFunctionBenchmark.latency                    xxh3      uuid16  
 avgt   50    7.281 ± 0.003   ns/op
   PrimaryKeyHashFunctionBenchmark.latency     murmur3StringHelper      uuid16  
 avgt   50   12.557 ± 0.060   ns/op
   ```
   
   So UUID16, xxHash is ~1.7x lower avg latency and ~3.5x higher throughput. 
This roughly matches benchmarks reported by XXHash project itself: 
https://github.com/Cyan4973/xxHash


-- 
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