kaivalnp commented on PR #15272:
URL: https://github.com/apache/lucene/pull/15272#issuecomment-3372247607
Turns out these compiler directives make things worse in actual benchmarks
(Cohere vectors, 768d):
baseline
```
recall latency(ms) netCPU avgCpuCount nDoc topK fanout maxConn
beamWidth quantized visited index(s) index_docs/s force_merge(s)
num_segments index_size(MB) vec_disk(MB) vec_RAM(MB) indexType
0.915 1.531 1.523 0.995 200000 100 50 32
250 8 bits 4583 12.41 16113.44 18.41 1
751.21 735.474 149.536 HNSW
```
candidate (baseline + the following JVM arguments)
```
-XX:CompileCommand=dontinline,org.apache.lucene.internal.vectorization.PanamaVectorUtilSupport$ArrayLoader::*
-XX:CompileCommand=dontinline,org.apache.lucene.internal.vectorization.PanamaVectorUtilSupport$MemorySegmentLoader::*
```
```
recall latency(ms) netCPU avgCpuCount nDoc topK fanout maxConn
beamWidth quantized visited index(s) index_docs/s force_merge(s)
num_segments index_size(MB) vec_disk(MB) vec_RAM(MB) indexType
0.917 9.765 9.745 0.998 200000 100 50 32
250 8 bits 4592 13.03 15346.84 96.83 1
751.29 735.474 149.536 HNSW
```
--
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]
