HaoYang670 commented on PR #2127:
URL: https://github.com/apache/arrow-rs/pull/2127#issuecomment-1193121071
After doing lots of bench tests today, I found that the performance
regression of `bench_bool` is really weird. It is caused by value builder, but
not the null buffer builder. I have tried to remove the impact of the null
buffer builder:
```rust
/// Appends a slice of type `T` into the builder
#[inline]
pub fn append_slice(&mut self, v: &[bool]) {
//self.null_buffer_builder.append_n_true(v.len());
self.values_builder.append_slice(v);
}
```
and run `cargo bench --bench builder -- --baseline master `,
but still got the weird result: (Tested on Intel Ubuntu)
```
bench_bool/bench_bool time: [560.89 us 561.72 us 562.61 us]
thrpt: [888.71 MiB/s 890.13 MiB/s 891.44 MiB/s]
change:
time: [+116.82% +117.38% +117.92%] (p = 0.00 <
0.05)
thrpt: [-54.113% -53.997% -53.879%]
Performance has regressed.
Found 4 outliers among 100 measurements (4.00%)
1 (1.00%) low severe
1 (1.00%) low mild
2 (2.00%) high mild
```
I have also tried to `inline(always)` the
`BooleanBufferBuilder::append_slice` but it had no effect.
cc @tustvold @viirya @alamb could you reproduce the above result? And any
idea about it?
--
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: github-unsubscr...@arrow.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
