prudhvigodithi opened a new pull request, #16306:
URL: https://github.com/apache/lucene/pull/16306
### Description
After split LRUQueryCache into 16 partitions with separate locks, concurrent
cache insertions no longer wait for each other. UsageTrackingQueryCachingPolicy
still does, every onUse() and frequency() call funnels through one shared
synchronized (this) block. This PR replaces the single
FrequencyTrackingRingBuffer field with 16 independent ring buffers, partitioned
by hash code.
| Benchmark | Baseline (M ops/s) | Candidate (M
ops/s) | Δ throughput |
|---------------------------------|-------------------:|--------------------:|-------------:|
| `singleThread_onUse` | 18.5 |
14.1 | −24% |
| `singleThread_shouldCache` | 26.9 |
23.5 | −13% |
| `concurrent_uniform_4t` | 7.8 |
13.3 | +70% |
| `concurrent_uniform_8t` | 8.5 |
17.1 | +101% |
| `concurrent_uniform_16t` | 8.5 |
16.4 | +93% |
| `concurrent_uniform_32t` | 8.6 |
16.4 | +90% |
| `concurrent_mixed_zipf_16t` | 9.3 |
16.9 | +82% |
| `concurrent_mixed_zipf_32t` | 9.6 |
16.7 | +73% |
--
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]
