Benchmarking osu!lazer against rosu-pp

main.rs

use std::{
    fs::read_to_string,
    hint::black_box,
    path::Path,
    time::{Duration, Instant},
};

use rosu_pp::{
    any::{DifficultyAttributes, PerformanceAttributes},
    Beatmap, Difficulty, Performance,
};

fn main() {
    // 100 random map ids
    let map_ids_string = read_to_string("/path/to/map_ids.csv").unwrap();

    let map_ids: Vec<u32> = map_ids_string
        .split(',')
        .map(str::parse)
        .map(Result::unwrap)
        .collect();

    print!("Decoding maps:            ");
    let paths: Vec<_> = map_ids
        .iter()
        .map(|id| format!("F:/osu!/beatmaps/{id}.osu"))
        .collect();
    bench(decode_map, &paths);

    print!("Calculating difficulties: ");
    let maps: Vec<_> = paths.iter().map(decode_map).collect();
    bench(calculate_difficulty, &maps);

    print!("Calculating performances: ");
    let attrs: Vec<_> = maps.iter().map(calculate_difficulty).collect();
    bench(calculate_performance, &attrs);
}

fn decode_map(path: impl AsRef<Path>) -> Beatmap {
    Beatmap::from_path(path).unwrap()
}

fn calculate_difficulty(map: &Beatmap) -> DifficultyAttributes {
    Difficulty::new().calculate(map)
}

fn calculate_performance(diff_attrs: &DifficultyAttributes) -> PerformanceAttributes {
    Performance::from(diff_attrs.clone()).calculate()
}

fn bench<'a, I, O>(f: impl Fn(&'a I) -> O, inputs: &'a [I]) {
    const CHUNK_SIZE: usize = 10;

    // Warmup
    for _ in 0..CHUNK_SIZE {
        for input in inputs {
            let _ = black_box(f(input));
        }
    }

    let mut chunk_times = Vec::new();

    let start = Instant::now();

    while start.elapsed().as_secs() < 10 {
        let curr = Instant::now();

        for _ in 0..CHUNK_SIZE {
            for input in inputs {
                let _ = black_box(f(input));
            }
        }

        let elapsed = curr.elapsed();
        chunk_times.push(elapsed / CHUNK_SIZE as u32);
    }

    chunk_times.sort_unstable();

    let mean = chunk_times.iter().copied().sum::<Duration>() / chunk_times.len() as u32;
    let median = chunk_times[chunk_times.len() / 2];

    println!("Median: {median:.2?} | Mean: {mean:.2?}");
}

Program.cs

using osu.Game.Beatmaps;
using osu.Game.Beatmaps.Formats;
using osu.Game.Rulesets;
using osu.Game.Rulesets.Catch;
using osu.Game.Rulesets.Difficulty;
using osu.Game.Rulesets.Mania;
using osu.Game.Rulesets.Osu;
using osu.Game.Rulesets.Taiko;
using osu.Game.Scoring;
using System.Diagnostics;

namespace Bench
{
    public class Program
    {
        static void Main()
        {
            Decoder.RegisterDependencies(new AssemblyRulesetStore());

            // 100 random map ids
            string mapIdsString = File.ReadAllText("/path/to/map_ids.csv");

            var mapIds = mapIdsString.Split(',').Select(int.Parse).ToList();

            Console.Write($"Decoding maps:            ");
            var paths = mapIds.Select(mapId => $"F:/osu!/beatmaps/{mapId}.osu").ToList();
            Bench(DecodeMap, paths);

            Console.Write($"Calculating difficulties: ");
            var maps = paths.Select(DecodeMap).ToList();
            Bench(CalculateDifficulty, maps);

            Console.Write($"Calculating performances: ");
            var attrs = maps.Select(map => (CreateRuleset(map.Beatmap.BeatmapInfo.Ruleset), CalculateDifficulty(map))).ToList();
            Bench(CalculatePerformance, attrs);
        }

        static FlatFileWorkingBeatmap DecodeMap(string path) => new FlatFileWorkingBeatmap(path);

        static DifficultyAttributes CalculateDifficulty(IWorkingBeatmap map)
            => CreateRuleset(map.Beatmap.BeatmapInfo.Ruleset).CreateDifficultyCalculator(map).Calculate();

        static PerformanceAttributes? CalculatePerformance((Ruleset ruleset, DifficultyAttributes diffAttrs) input)
            => input.ruleset.CreatePerformanceCalculator()?.Calculate(new ScoreInfo(), input.diffAttrs);

        static void Bench<I, O>(Func<I, O> f, List<I> inputs)
        {
            const int chunk_size = 10;

            // Warmup
            for (int i = 0; i < chunk_size; i++)
            {
                foreach (I input in inputs)
                {
                    var _ = f(input);
                }
            }

            var chunkTimes = new List<TimeSpan>();

            var start = Stopwatch.StartNew();

            while (start.Elapsed.Seconds < 10)
            {
                var curr = Stopwatch.StartNew();

                for (int i = 0; i < chunk_size;i++)
                {
                    foreach (I input in inputs)
                    {
                        var _ = f(input);
                    }
                }

                var elapsed = curr.Elapsed;
                chunkTimes.Add(elapsed / chunk_size);
            }

            chunkTimes.Sort();

            var mean = new TimeSpan(chunkTimes.Aggregate(0L, (sum, next) => sum + next.Ticks) / chunkTimes.Count);
            var median = chunkTimes[chunkTimes.Count / 2];

            Console.WriteLine($"Median: {FormatTimeSpan(mean)} | Mean: {FormatTimeSpan(median)}");
        }

        static Ruleset CreateRuleset(RulesetInfo info)
        {
            switch (info.OnlineID)
            {
                case 0:
                    return new OsuRuleset();
                case 1:
                    return new TaikoRuleset();
                case 2:
                    return new CatchRuleset();
                case 3:
                    return new ManiaRuleset();
                default:
                    throw new InvalidDataException($"invalid ruleset id {info.OnlineID}");
            }
        }

        static string FormatTimeSpan(TimeSpan time)
        {
            const long ns_per_tick = 100L;
            const long ticks_per_milli = 1000000L / ns_per_tick;
            const long ticks_per_micro = 1000L / ns_per_tick;

            if (time.Seconds > 0)
                return $"{time:ss\\.ff}s";
            else if (time.Ticks >= ticks_per_milli)
                return $"{time.Ticks / ticks_per_milli}.{(time.Ticks % ticks_per_milli).ToString().PadRight(2, '0')[..2]}ms";
            else if (time.Ticks >= ticks_per_micro)
                return $"{time.Ticks / ticks_per_micro}.{(time.Ticks % ticks_per_micro).ToString().PadRight(2, '0')[..2]}µs";
            else
                return $"{time.Ticks * ns_per_tick}ns";
        }
    }
}