Guo Feng

I checked some other usage, seeing that for operations like `bytebuffer#readBytes` we also just checked once. If the `length`

Thanks @jpountz @uschindler , this is the benchmark result based on the newest codes ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value HighTermTitleBDVSort 168.66 (13.3%) 163.17 (11.2%) -3.3% (...

@uschindler Thank you for the benchmark guidance! I made a new change in this [commit](https://github.com/apache/lucene/pull/592/commits/347fdb13d60f620064290123f5122e6ea78ac288) that ensure valid before each call of `ByteBuffer#get`. And I run a benchmark on wikimediumall...

Thanks all! I've finished benchmarks (20 JVM * 200 repeat): **Read byte from guard** > Here is the [code](https://github.com/apache/lucene/pull/592/commits/5fae0b2f4ffc6deab48dd0f447f7d76587a328bf) ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value HighTermTitleBDVSort 60.79...

Thanks @jpountz . I opened https://issues.apache.org/jira/browse/LUCENE-10376 (https://github.com/apache/lucene/pull/602) It can be seen that `IntNRQ` is getting a 10% speed up but i'm not very sure if it is a noise since...

To test if it is the `try-catch` block preventing inline, I made another experiment: call `readByte()` instead of `guard.get(curBuf)` to read byte, like this (Yes it is just a copy...

As it can be the same as super, I tried to reuse super logic ([call super.xxx](https://github.com/apache/lucene/pull/592/commits/480f383fa74229389cc2acd44fa2e86ae0c71130)). still seeing the speed up: ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value...

To ensure the benchmark is warmed up, I also run a benchmark that repeat `AndHighLow` task 20000 times (which was usually seen a good speed up in benchmarks before). Here...

Thanks @uschindler ! > I was also thinking about the following: Why not make readVInt/readVLong final in the DataInput class? This should be possible after the refactoring in #602 !...

Here is the report of [making vint final](https://github.com/apache/lucene/compare/main...gf2121:make_vint_final?expand=1) (50 repeat * 20 JVM): ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value HighTermTitleBDVSort 66.42 (20.7%) 61.57 (13.7%) -7.3% ( -34%...