Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Improve parsing performance #29

Merged
merged 1 commit into from
Jun 11, 2023
Merged

Improve parsing performance #29

merged 1 commit into from
Jun 11, 2023

Conversation

RoosterDragon
Copy link
Contributor

@RoosterDragon RoosterDragon commented Jun 10, 2023

  • In LinguiniParser, ZeroCopyReader, ZeroCopyUtil: change helpers from dealing in ReadOnlyMemory to char. In inner loops this removes a lot of redundant checks for a span of length 1 and inlines more readily.
  • In ZeroCopyReader add helpers for SeekEol, IndexOfAnyChar. These use existing helper methods of Span that are SIMD accelerated to search for indexes quickly. These provide significant speedups compared to incrementing the index in a loop.
  • In FluentBundle.AddResourceOverriding, provide a specialized 'override' method to avoid two dictionary lookups and allocating an empty list.

This makes parsing performance 1.5x faster. On my machine, the following test project runs in 2.9 seconds, compared to 4.5 seconds previously.

Test Project
using System;
using System.Diagnostics;
using System.Globalization;
using System.IO;
using System.Linq;
using Linguini.Bundle.Builder;
using Linguini.Syntax.Parser;

namespace ConsoleApp
{
    internal class Program
    {
        static void Main()
        {
            var sw = new Stopwatch();
            for (var i = 0; i < 10000; i++)
            {
                Setup(out var culture, out var streams);

                sw.Start();
                var bundle = LinguiniBuilder.Builder()
                    .CultureInfo(culture)
                    .SkipResources()
                    .SetUseIsolating(false)
                    .UseConcurrent()
                    .UncheckedBuild();

                foreach (var stream in streams)
                {
                    using var reader = new StreamReader(stream);
                    var parser = new LinguiniParser(reader);
                    var resource = parser.Parse();

                    bundle.AddResourceOverriding(resource);
                }
                sw.Stop();
            }

            sw.Stop();

            Console.WriteLine($"Done in {sw.Elapsed.TotalSeconds:0.00s}");
        }

        static void Setup(out CultureInfo culture, out MemoryStream[] streams)
        {
            culture = new CultureInfo("en");
            var paths = new[]
            {
                @"your-test-files-here.ftl",
            };
            streams = paths.Select(p => new MemoryStream(File.ReadAllBytes(p))).ToArray();
        }
    }
}

I used the following test files, but feel free to try with whatever you have to hand.

@Ygg01
Copy link
Owner

Ygg01 commented Jun 10, 2023

Seems to be failing tests.

Also, some benchmarking would be nice. I can add some Benchmarks later (although benchmarking is a bit of a dark science - easy to get wrong).

@Ygg01
Copy link
Owner

Ygg01 commented Jun 10, 2023

Here are results for current master:

BenchmarkDotNet=v0.13.5, OS=Windows 10 (10.0.19044.2965/21H2/November2021Update)
AMD Ryzen 9 5900X, 1 CPU, 24 logical and 12 physical cores
.NET SDK=7.0.102
[Host] : .NET 5.0.12 (5.0.1221.52207), X64 RyuJIT AVX2 DEBUG

Toolchain=InProcessEmitToolchain

Master:
Method N Mean Error StdDev
BenchmarkParser 1 3.009 ms 0.0236 ms 0.0197 ms
BenchmarkParser 25 74.498 ms 0.3450 ms 0.3227 ms
BenchmarkParser 50 150.409 ms 0.7432 ms 0.6952 ms
Perf branch:
Method N Mean Error StdDev
BenchmarkParser 1 1.337 ms 0.0124 ms 0.0110 ms
BenchmarkParser 25 32.971 ms 0.4621 ms 0.4096 ms
BenchmarkParser 50 65.290 ms 0.7167 ms 0.6354 ms

All in all those are some nice speedups.

- In LinguiniParser, ZeroCopyReader, ZeroCopyUtil: change helpers from dealing in ReadOnlyMemory<char> to char. In inner loops this removes a lot of redundant checks for a span of length 1 and inlines more readily.
- In ZeroCopyReader add helpers for SeekEol, IndexOfAnyChar. These use existing helper methods of Span that are SIMD accelerated to search for indexes quickly. These provide significant speedups compared to incrementing the index in a loop.
- In FluentBundle.AddResourceOverriding, provide a specialized 'override' method to avoid two dictionary lookups and allocating an empty list.
@Ygg01
Copy link
Owner

Ygg01 commented Jun 11, 2023

Ok, expect a release soonish. Need to write changelog.

@RoosterDragon
Copy link
Contributor Author

@Ygg01 looks like the action failed partway through pushing to nuget, just a ping in case you hadn't noticed: https://github.com/Ygg01/Linguini/actions/runs/5237132620

@Ygg01
Copy link
Owner

Ygg01 commented Jun 12, 2023

@Ygg01 looks like the action failed partway through pushing to nuget, just a ping in case you hadn't noticed: https://github.com/Ygg01/Linguini/actions/runs/5237132620

Yeah, I manually deployed them. I need to sort that deployment script. Look at repo readme , the versions of Linguini.Shared and Linguini.Syntax are updated to v0.5.0.

@RoosterDragon
Copy link
Contributor Author

May I request Linguini.Bundle as well?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants