//! Single-threaded benchmarks, writing and reading chunks.
//!
//! Single-threaded usage is *not* a typical use case,
//! but the used chunk size and capacity could be.

use std::io::{Read, Write};

use criterion::{black_box, criterion_group, criterion_main};
use criterion::{AxisScale, PlotConfiguration};

use basedrop::Collector;
use rtrb_basedrop::{CopyToUninit, RingBuffer};

const CHUNK_SIZE: usize = 4096;

fn add_function<F, M>(group: &mut criterion::BenchmarkGroup<M>, id: impl Into<String>, mut f: F)
where
    F: FnMut(&[u8]) -> [u8; CHUNK_SIZE],
    M: criterion::measurement::Measurement,
{
    group.bench_function(id, |b| {
        let mut data = [0; CHUNK_SIZE];
        let mut i: u8 = 0;
        for dst in data.iter_mut() {
            *dst = i;
            i = i.wrapping_add(1);
        }
        b.iter(|| {
            assert_eq!(f(black_box(&data)), data);
        });
    });
}

pub fn criterion_benchmark(criterion: &mut criterion::Criterion) {
    let collector = Collector::new();

    let mut group = criterion.benchmark_group("single-thread-with-chunks");
    group.throughput(criterion::Throughput::Bytes(CHUNK_SIZE as u64));
    group.plot_config(PlotConfiguration::default().summary_scale(AxisScale::Logarithmic));

    let (mut p, mut c) = RingBuffer::<u8>::new(1000 * CHUNK_SIZE, &collector.handle());

    add_function(&mut group, "1-pop", |data| {
        let mut result = [0; CHUNK_SIZE];
        let _ = p.write(data).unwrap();
        for i in result.iter_mut() {
            *i = c.pop().unwrap();
        }
        result
    });

    add_function(&mut group, "1-push", |data| {
        let mut result = [0; CHUNK_SIZE];
        for &i in data.iter() {
            p.push(i).unwrap();
        }
        let _ = c.read(&mut result).unwrap();
        result
    });

    add_function(&mut group, "2-slice-read", |data| {
        let mut result = [0; CHUNK_SIZE];
        let _ = p.write(data).unwrap();
        let chunk = c.read_chunk(data.len()).unwrap();
        let (first, second) = chunk.as_slices();
        result.copy_from_slice(first);
        debug_assert!(second.is_empty());
        chunk.commit_all();
        result
    });

    add_function(&mut group, "2-slice-write", |data| {
        let mut result = [0; CHUNK_SIZE];
        let mut chunk = p.write_chunk(data.len()).unwrap();
        let (first, second) = chunk.as_mut_slices();
        first.copy_from_slice(data);
        debug_assert!(second.is_empty());
        chunk.commit_all();
        let _ = c.read(&mut result).unwrap();
        result
    });

    add_function(&mut group, "2-slice-write-uninit", |data| {
        let mut result = [0; CHUNK_SIZE];
        let mut chunk = p.write_chunk_uninit(data.len()).unwrap();
        let (first, second) = chunk.as_mut_slices();
        data.copy_to_uninit(first);
        debug_assert!(second.is_empty());
        unsafe {
            chunk.commit_all();
        }
        let _ = c.read(&mut result).unwrap();
        result
    });

    add_function(&mut group, "3-iterate-read", |data| {
        let mut result = [0; CHUNK_SIZE];
        let _ = p.write(data).unwrap();
        let chunk = c.read_chunk(data.len()).unwrap();
        for (dst, src) in result.iter_mut().zip(chunk) {
            *dst = src;
        }
        result
    });

    add_function(&mut group, "3-iterate-write", |data| {
        let mut result = [0; CHUNK_SIZE];
        let chunk = p.write_chunk_uninit(data.len()).unwrap();
        chunk.fill_from_iter(&mut data.iter().copied());
        let _ = c.read(&mut result).unwrap();
        result
    });

    add_function(&mut group, "4-write-read", |data| {
        let mut result = [0; CHUNK_SIZE];
        let _ = p.write(data).unwrap();
        let _ = c.read(&mut result).unwrap();
        result
    });

    group.finish();
}

criterion_group!(benches, criterion_benchmark);
criterion_main!(benches);