mikemccand commented on issue #13768:
URL: https://github.com/apache/lucene/issues/13768#issuecomment-2346091188
OK I disabled Panama (via temporary code change in
`VectorizationProvider.java` -- we don't accept sysprops to disable this
anymore right?):
```
recall latency (ms) nDoc topK fanout maxConn beamWidth quantized
index s force merge s num segments index size (MB)
0.318 0.312 1500000 10 6 32 50 4 bits
175.21 125.64 1 5013.21
0.319 0.285 1500000 10 6 32 50 4 bits
174.54 124.80 1 5562.51
```
Indeed there is some performance penalty now (285 usec -> 312 usec, ~9.5%)
... recall also bounced around a bit, but prolly that's acceptable HNSW
randomness noise. And wow look how much slower indexing / force merging got
... those SIMD instructions clearly help ;)
But I don't think we should block removing `compress` option due to non-SIMD
results?
--
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]
