use criterion::criterion_group;
use criterion::criterion_main;
use criterion::Benchmark;
use criterion::Criterion;
use criterion::Throughput;

use rand::Rng;
use cyfs_raptorq::SourceBlockDecoder;
use cyfs_raptorq::SourceBlockEncoder;
use cyfs_raptorq::Symbol;
use cyfs_raptorq::{ObjectTransmissionInformation, Octet};

fn criterion_benchmark(c: &mut Criterion) {
    let octet1 = Octet::new(rand::thread_rng().gen_range(1, 255));
    let symbol_size = 512;
    let mut data1: Vec<u8> = vec![0; symbol_size];
    let mut data2: Vec<u8> = vec![0; symbol_size];
    for i in 0..symbol_size {
        data1[i] = rand::thread_rng().gen_range(1, 255);
        data2[i] = rand::thread_rng().gen_range(1, 255);
    }
    let symbol1 = Symbol::new(data1);
    let symbol2 = Symbol::new(data2);

    let symbol1_mul_scalar = symbol1.clone();
    let octet1_mul_scalar = octet1.clone();
    c.bench(
        "Symbol mulassign_scalar()",
        Benchmark::new("", move |b| {
            b.iter(|| {
                let mut temp = symbol1_mul_scalar.clone();
                temp.mulassign_scalar(&octet1_mul_scalar);
                temp
            })
        })
        .throughput(Throughput::Bytes(symbol1.len() as u64)),
    );

    let symbol1_addassign = symbol1.clone();
    let symbol2_addassign = symbol2.clone();
    c.bench(
        "Symbol +=",
        Benchmark::new("", move |b| {
            b.iter(|| {
                let mut temp = symbol1_addassign.clone();
                temp += &symbol2_addassign;
                temp
            })
        })
        .throughput(Throughput::Bytes(symbol1.len() as u64)),
    );

    let symbol1_fma = symbol1.clone();
    let symbol2_fma = symbol2.clone();
    let octet1_fma = octet1.clone();
    c.bench(
        "Symbol FMA",
        Benchmark::new("", move |b| {
            b.iter(|| {
                let mut temp = symbol1_fma.clone();
                temp.fused_addassign_mul_scalar(&symbol2_fma, &octet1_fma);
                temp
            })
        })
        .throughput(Throughput::Bytes(symbol1.len() as u64)),
    );

    let elements = 10 * 1024;
    let symbol_size = 512;
    let mut data: Vec<u8> = vec![0; elements];
    for i in 0..elements {
        data[i] = rand::thread_rng().gen();
    }

    let encode_data = data.clone();
    c.bench(
        "encode 10KB",
        Benchmark::new("", move |b| {
            b.iter(|| {
                let config = ObjectTransmissionInformation::new(0, symbol_size, 0, 1, 1);
                let encoder = SourceBlockEncoder::new2(1, &config, &encode_data);
                return encoder.source_packets();
            })
        })
        .throughput(Throughput::Bytes(data.len() as u64)),
    );

    let roundtrip_data = data.clone();
    c.bench(
        "roundtrip 10KB",
        Benchmark::new("", move |b| {
            b.iter(|| {
                let config = ObjectTransmissionInformation::new(0, symbol_size, 0, 1, 1);
                let encoder = SourceBlockEncoder::new2(1, &config, &roundtrip_data);
                let mut decoder = SourceBlockDecoder::new2(1, &config, elements as u64);
                return decoder.decode(encoder.source_packets());
            })
        })
        .throughput(Throughput::Bytes(data.len() as u64)),
    );

    let repair_data = data.clone();
    c.bench(
        "roundtrip repair 10KB",
        Benchmark::new("", move |b| {
            b.iter(|| {
                let config = ObjectTransmissionInformation::new(0, symbol_size, 0, 1, 1);
                let encoder = SourceBlockEncoder::new2(1, &config, &repair_data);
                let repair_packets = (elements / symbol_size as usize) as u32;
                let mut decoder = SourceBlockDecoder::new2(1, &config, elements as u64);
                return decoder.decode(encoder.repair_packets(0, repair_packets));
            })
        })
        .throughput(Throughput::Bytes(data.len() as u64)),
    );
}

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