Improved performance in lexical write (~5%)

[ritchie46]

It might be interesting to explore a slightly adapted version of BufStreamingIterator.

If I am correct it currently writes to an internal buffer, and then copies that buffer to the W: Write buffer. If we would accept that final buffer in the next() method as argument, we could directly write to that.

Will explore that idea later.

[codecov]

Codecov Report

Merging #1067 (8b84f3d) into main (6919cc3) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##             main    #1067   +/-   ##
=======================================
  Coverage   81.30%   81.31%           
=======================================
  Files         363      363           
  Lines       34689    34700   +11     
=======================================
+ Hits        28205    28217   +12     
+ Misses       6484     6483    -1     

Impacted Files	Coverage Δ
src/util/lexical.rs	76.47% <100.00%> (+11.25%)	⬆️
src/io/ipc/read/stream_async.rs	74.38% <0.00%> (+0.82%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6919cc3...8b84f3d. Read the comment docs.

[jorgecarleitao]

Looks reasonable! Do you have a bench on performance delta? Curious how it impacts.

[ritchie46]

Yup

bench_1                 time:   [1.0596 ms 1.0599 ms 1.0601 ms]                     
                        change: [-5.8026% -5.3279% -4.9843%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 6 outliers among 100 measurements (6.00%)
  4 (4.00%) low mild
  2 (2.00%) high mild

pub fn compute(vals: &[i32], buf: &mut Vec<u8>)  {
    for v in vals {
        memcheck::lexical_to_bytes_mut_unchecked(*v, buf)
    }
}

fn add_benchmark(c: &mut Criterion) {

    let mut rng = thread_rng();
    let values = (0..100000)
        .map(|_| rng.sample(Standard))
        .collect::<Vec<i32>>();

    let mut buf = vec![];

    c.bench_function("bench_1", |b| {
        b.iter(|| black_box(compute(&values, &mut buf)))
    });
}