Natea Eshetu Beshada created FLINK-39044:
--------------------------------------------

             Summary: Optimize BinaryStringData.copy() to return this for 
immutable strings
                 Key: FLINK-39044
                 URL: https://issues.apache.org/jira/browse/FLINK-39044
             Project: Flink
          Issue Type: Improvement
          Components: Table SQL / Runtime
            Reporter: Natea Eshetu Beshada


  Summary

  BinaryStringData.copy() currently always allocates a new BinaryStringData 
object and copies the underlying byte array, even when the source is already 
immutable and can be safely shared. This is unnecessary
  overhead for strings that occupy a single memory segment at offset 0 with no 
wasted space.

  Proposed Change

  Return this from BinaryStringData.copy() when all of the following conditions 
are met:
  - The string occupies exactly one MemorySegment
  - The offset is 0
  - The segment size equals the string's byte length (no wasted space)

  This is safe because BinaryStringData is immutable once constructed - none of 
its public methods mutate the underlying data.

  Benchmark Results
  
┌──────────────────────────┬───────────────────┬────────────────────┬─────────────┐
  │        Benchmark         │ Baseline (ops/ms) │ Optimized (ops/ms) │ 
Improvement │
  
├──────────────────────────┼───────────────────┼────────────────────┼─────────────┤
  │ copyCompactString        │ 18,381 ± 22,770   │ 807,198 ± 13,825   │ +4,292% 
    │
  
├──────────────────────────┼───────────────────┼────────────────────┼─────────────┤
  │ copyLargeString          │ 3,551 ± 4,718     │ 748,334 ± 6,211    │ 
+20,975%    │
  
├──────────────────────────┼───────────────────┼────────────────────┼─────────────┤
  │ deserializeCompactString │ 60,560 ± 34,536   │ 98,886 ± 1,043     │ +63%    
    │
  
├──────────────────────────┼───────────────────┼────────────────────┼─────────────┤
  │ deserializeLargeString   │ 19,253 ± 181      │ 33,917 ± 136       │ +76%    
    │
  
├──────────────────────────┼───────────────────┼────────────────────┼─────────────┤
  │ serializeCompactString   │ 125,307 ± 2,451   │ 122,648 ± 968      │ -2% 
(noise) │
  
├──────────────────────────┼───────────────────┼────────────────────┼─────────────┤
  │ serializeLargeString     │ 27,746 ± 3,046    │ 27,646 ± 2,968     │ 0% 
(noise)  │
  
└──────────────────────────┴───────────────────┴────────────────────┴─────────────┘
  JMH config: 2 forks, 5 warmup iterations, 5 measurement iterations
  Baseline: origin/master @ 7488449d086
 Note: The baseline copy benchmarks had high variance due to GC pressure from 
allocation. The optimized version eliminates allocation entirely, resulting in 
near-zero variance



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to