xoofx

var config = ManualConfig.Create(DefaultConfig.Instance);
var gcDiagnoser = new MemoryDiagnoser();
config.Add(new Job { Mode = Mode.SingleRun, LaunchCount = 2, WarmupCount = 2, TargetCount = 10 });
config.Add(gcDiagnoser);

_________________________________ Patch for BenchMarkDotNet to have precise GC:


diff --git a/BenchmarkDotNet/Running/MethodInvoker.cs b/BenchmarkDotNet/Running/MethodInvoker.cs
index e8d8de5..6a46aa2 100644

xoofx

using System;
using System.Runtime.InteropServices;

namespace EvilArray
{
    /// <summary>
    /// Cast an array of structs to an array of byte[]
    /// </summary>
    class Program
    {

xoofx

General

• Dataflow (Task Parallel Library)
• Document: Introduction to TPL Dataflow (Microsoft - Stephen Toub)
• Video: Beginner: TPL Dataflow Tour (Stephen Toub), 15min
• Video: Advanced: Stephen Toub: Inside TPL Dataflow (Stephen Toub), 1h

Repos

• System.Threading.Tasks.Dataflow on github
• Gridsum.DataflowEx : An Extension to DataFlow (@gridsum)

xoofx

// Related to this benchmark article https://stebet.net/benchmarking-and-performance-optimizations-in-c-using-benchmarkdotnet/

public static class IP2Locationxoofx
{
	public static uint IPAddressToInteger(string input)
	{
		uint ipAddress = 0;
		uint acc = 0;
		// Note: we assume that the string is well formed
		foreach (var c in input)

xoofx

// Follow up of the discussion https://twitter.com/stebets/status/739563007202758656

public static unsafe string IntegerToIPAddress(uint input)
{
	var text = stackalloc char[15];
	var ptext = text + 15;
	for (int i = 0; i < 4; i++)
	{
		if (i > 0)
		{

xoofx

// Follow up of https://twitter.com/stebets/status/739563007202758656
private static readonly string[] ByteToString = Enumerable.Range(0, 256).Select(x => x.ToString()).ToArray();

public unsafe static string FastIntegerToIPAddress2(uint input)
{
	// t3.t2.t1.t0
	var t0 = ByteToString[input & 0xFF];
	input = input >> 8;
	var t1 = ByteToString[input & 0xFF];
	input = input >> 8;

xoofx

// Check the performance of using an Array (preallocated) vs a List<T>, vs a Dictionary<string, T>
// both in terms of CPU and memory
// * Summary *
// 
// BenchmarkDotNet=v0.9.7.0
// OS=Microsoft Windows NT 6.2.9200.0
// Processor=Intel(R) Core(TM) i7-4770 CPU 3.40GHz, ProcessorCount=8
// Frequency=3319351 ticks, Resolution=301.2637 ns, Timer=TSC
// HostCLR=MS.NET 4.0.30319.42000, Arch=32-bit RELEASE
// JitModules=clrjit-v4.6.1080.0

xoofx

// Ugly pure ASM version of https://github.com/aspnet/KestrelHttpServer/pull/1138 and gist https://gist.github.com/benaadams/2dd3f99230757111e91915f638067a09
// Not performing better though...

[Benchmark]
public int TestFindFirstByteAVX()
{
	var vector = _vectors[ByteSet];
	return FindFirstByteAVX(ref vector);
}

xoofx

/*
Q: At which size is it preferrable to use binary search over a simple linear search for a small ordered set?

R: Under 5 elements, linear search is slightly faster (from 200% to 10% faster)
But in practice, not sure a switch case to select the best method is really worth it
Unless main usecase is having most of the time only 1-2 elements (so it could be optimized manually without a loop and switch to binary otherwise)
At 3-4 elements, linear is only 10-5% faster

Relative performance between linear and binary search:
size   x86     x64

xoofx

// You need to configure your C# project with x86 or x64 platform (Tools\Configuration Manager\Create new Platform on the project)
// otherwise the native libSkiaSharp.dll will not get copied
using System;
using System.IO;
using SkiaSharp;

namespace TestSkia
{
    class Program
    {