[GitHub] [pinot] richardstartin commented on pull request #8818: regexp_like fusing

[GitBox]

richardstartin commented on PR #8818:
URL: https://github.com/apache/pinot/pull/8818#issuecomment-1154392815

   Just circling back here having taken a quick look at 
`BenchmarkFuseRegexp.decreasing9Fusing` - I got similar results to Gonzalo:
   
   ```
   Benchmark                              (_fstType)  Mode  Cnt   Score   Error 
 Units
   BenchmarkFuseRegexp.decreasing9Fusing      LUCENE  avgt    5  32.525 ± 2.733 
 ms/op
   BenchmarkFuseRegexp.decreasing9Fusing      NATIVE  avgt    5   0.487 ± 0.123 
 ms/op
   BenchmarkFuseRegexp.decreasing9Fusing        null  avgt    5   0.845 ± 0.204 
 ms/op
   ```
   
    Lucene spends virtually all of its time in _minimization_ (which I 
mistakenly referred to as _determinization_ earlier):
   
   <img width="936" alt="Screenshot 2022-06-13 at 21 10 51" 
src="https://user-images.githubusercontent.com/16439049/173436813-a13a5830-ad7b-4a8c-8e5f-3cf8003640df.png";>
    
   Now if I enable minimization in the native implementation by reverting 
#8237, the scores change drastically:
   
   ```
   Benchmark                              (_fstType)  Mode  Cnt    Score    
Error  Units
   BenchmarkFuseRegexp.decreasing9Fusing      LUCENE  avgt    5   32.399 ±  
3.005  ms/op
   BenchmarkFuseRegexp.decreasing9Fusing      NATIVE  avgt    5  366.520 ± 
40.574  ms/op
   BenchmarkFuseRegexp.decreasing9Fusing        null  avgt    5    0.819 ±  
0.172  ms/op
   ```
   
   As do the profiles. Without minimization, there is no clear single 
bottleneck:
   <img width="879" alt="Screenshot 2022-06-13 at 21 14 47" 
src="https://user-images.githubusercontent.com/16439049/173437464-980859ce-e7c5-4e8f-b43e-ba609ab39396.png";>
   
   With minimization, there is a single bottleneck, and it's in minimization :
   
   <img width="931" alt="Screenshot 2022-06-13 at 21 23 10" 
src="https://user-images.githubusercontent.com/16439049/173438712-5f8efc36-1f8d-4ea5-8bb6-b135d21ff4fc.png";>
   
   
   As to whether this is cause for concern is beside the point; this is the 
root cause of the difference observed, rather than a quirk of the benchmark 
which needs to be analysed and removed.
   


-- 
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: commits-unsubscr...@pinot.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org
For additional commands, e-mail: commits-h...@pinot.apache.org