costin opened a new pull request, #16286:
URL: https://github.com/apache/lucene/pull/16286
Add dedicated method for bulk processing. Mirrors
NumericDocValues#longValues for binary fields.
Provides a default implementation with default per-doc fallback plus
Lucene90 codec override that reads directly from
the data slice (fixed-length: readBytes(doc * length), variable-length:
batch address lookup).
### Benchmark
AMD EPYC 7R32 (c5a.2xlarge, AVX2), JDK 25 Temurin, 1M docs, batchSize=1024
#### Codec override benefit
`binaryValuesBulk` on default per-doc impl vs `binaryValuesBulk` with
Lucene90 override:
| encoding | valueLength | default (ops/s) | codec override (ops/s) | ratio |
|---|---:|---:|---:|---:|
| fixed | 8 | 39,821 | 65,940 | **1.66x** |
| fixed | 32 | 38,643 | 59,785 | **1.55x** |
| fixed | 128 | 31,571 | 46,174 | **1.46x** |
| variable | 8 | 19,930 | 26,154 | **1.31x** |
| variable | 32 | 15,222 | 17,754 | **1.17x** |
| variable | 128 | 12,676 | 15,243 | **1.20x** |
#### End-to-end improvement
Per-doc access on main vs bulk API with codec override:
| encoding | valueLength | per-doc (ops/s) | bulk+codec (ops/s) | ratio |
|---|---:|---:|---:|---:|
| fixed | 8 | 48,048 | 65,940 | **1.37x** |
| fixed | 32 | 48,226 | 59,785 | **1.24x** |
| fixed | 128 | 39,443 | 46,174 | **1.17x** |
| variable | 8 | 20,641 | 26,154 | **1.27x** |
| variable | 32 | 15,968 | 17,754 | **1.11x** |
| variable | 128 | 14,060 | 15,243 | **1.08x** |
--
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]
