Misaki.HighPerformance/README.md

Misaki.HighPerformance

A collection of high-performance C# libraries focused on low allocation, low overhead, and systems-level control.

The solution is organized around a few core goals:

• fast collections and reusable memory primitives
• unsafe and low-level building blocks when performance matters
• SIMD/SPMD-friendly mathematics
• a zero-allocation job system
• STB-based image loading and translation helpers
• analyzers and code fixes that help avoid accidental performance regressions

Most runtime projects target .NET 10 and enable unsafe code where needed. Analyzer packages target .NET Standard 2.0.

Packages

Each major package has its own README for project-specific guidance.

Project	Purpose	Package
Misaki.HighPerformance	Core collection utilities, pools, slot maps, sparse sets, and shared helpers.
Misaki.HighPerformance.LowLevel	Unsafe collections, allocators, arenas, fixed strings, memory utilities, and other low-level primitives.
Misaki.HighPerformance.Mathematics	Math helpers, geometry types, SIMD-friendly helpers, numeric constants, and vector operations.
Misaki.HighPerformance.Mathematics.SPMD	SPMD-oriented math helpers, lane abstractions, and vector template infrastructure.
Misaki.HighPerformance.Jobs	Job scheduling, worker threads, handles, dependency management, and temporary job allocation.

Highlights

Low-level collections and memory management

The low-level layer includes building blocks such as:

• UnsafeArray<T>
• UnsafeList<T>
• UnsafeQueue<T>
• UnsafeStack<T>
• UnsafeHashMap<TKey, TValue>
• UnsafeHashSet<T>
• UnsafeSparseSet<T>
• UnsafeSlotMap<T>
• MemoryPool
• VirtualArena
• DynamicArena
• AllocationManager
• FreeList
• FixedString and FixedText

These types are designed for scenarios where allocation patterns, data locality, and explicit lifetime management matter.

Mathematics for performance-sensitive code

The mathematics layer provides:

• math constants and helpers
• custom numeric types and vector extensions
• geometry primitives such as AABB, OBB, Plane, and SphereBounds
• SIMD-friendly operations through System.Runtime.Intrinsics

Job system

The job system is designed for work scheduling with minimal overhead and dependency tracking. It supports:

• single jobs
• parallel-for style jobs
• parallel jobs with batching
• job dependencies
• worker threads
• temporary frame-based allocation

Image loading

The image package wraps STB-based decoding workflows and includes support for image metadata, animated frames, and runtime statistics.

Getting started

Install the packages

dotnet add package Misaki.HighPerformance
dotnet add package Misaki.HighPerformance.LowLevel
dotnet add package Misaki.HighPerformance.Mathematics
dotnet add package Misaki.HighPerformance.Jobs
dotnet add package Misaki.HighPerformance.Image

Reference only what you need

The solution is split into smaller packages so you can bring in only the pieces required by your application. For example:

• use Misaki.HighPerformance for lightweight collection utilities
• add Misaki.HighPerformance.LowLevel when you need unsafe memory primitives
• use Misaki.HighPerformance.Mathematics for fast numeric and geometry work
• use Misaki.HighPerformance.Jobs for background work scheduling
• use Misaki.HighPerformance.Image for image decoding

Quick examples

Dynamic array

using Misaki.HighPerformance.Collections;

var values = new DynamicArray<int>();
values.Add(10);
values.Add(20);

Span<int> span = values.AsSpan();

Math helpers

using Misaki.HighPerformance.Mathematics;

float angle = math.radians(90f);
float tau = math.TAU;

Low-level memory usage

The low-level project is intended for advanced scenarios where explicit allocation and lifetime control are required. Prefer the safer higher-level projects when they already meet your needs.

Building the solution

Open the solution in Visual Studio or build from the command line:

dotnet build

Some projects use unsafe code and source generation. Make sure you are building with a compatible .NET SDK and that your environment allows those features.

Design goals

• keep allocations under control
• favor cache-friendly data structures
• expose low-level primitives without unnecessary abstraction
• support high-performance math and job execution paths
• provide analyzers to help keep code fast

Repository notes

• Most runtime projects target net10.0
• Analyzer projects target netstandard2.0
• Several projects enable packaging on build
• Unsafe code is used intentionally in the low-level, math, job, and image layers

License

See the repository license if one is provided.