m4rs-mt/ILGPU

ILGPU JIT Compiler for high-performance .Net GPU programs

What it solves

ILGPU provides a way for .NET developers to write high-performance GPU programs without needing to leave the .NET ecosystem or rely on native dependencies. It bridges the gap between the ease of use found in C++ AMP and the raw performance of CUDA programs.

How it works

It acts as a JIT (just-in-time) compiler that translates .NET-based language code into GPU-executable instructions. It includes a multi-threaded CPU accelerator, allowing developers to execute and debug kernels on the CPU before deploying to the GPU.

Who it’s for

.NET developers who need to leverage GPU acceleration for computationally intensive tasks but want to maintain the use of C# and the .NET 6.0+ toolchain.

Highlights

  • Pure C# Implementation: Written entirely in C# with no native dependencies.
  • Integrated CPU Accelerator: Allows for full kernel debugging and execution on the CPU.
  • Algorithm Library: Includes a standard library of high-level algorithms like sorting and prefix sum that work across all supported accelerators.
  • Hardware Feature Support: Supports advanced GPU features such as shared memory and atomics.

Related

  • Project
    mitsuba-renderer/drjit

    A just-in-time compiler for ordinary and differentiable computation designed to handle the massive computation graphs typical of differentiable rendering.

  • Project
    brainpy/BrainPy

    A JIT-compiled framework for computational neuroscience and brain-inspired computation that enables efficient simulation, training, and analysis of brain dynamics models.

  • Project
    tracel-ai/cubecl

    A Rust-based JIT compiler and language extension for writing high-performance compute kernels that run across CUDA, HIP, Metal, WebGPU, and CPU from a single codebase.

  • Project
    Jittor/jittor

    Jittor is a high‑performance deep‑learning framework that JIT‑compiles the entire computation graph and uses meta‑operators to generate specialised CUDA/C++ kernels. It offers a PyTorch‑like Python API, supports CPU and GPU back‑ends (CUDA, ROCm, Hygon), and includes a model zoo for vision, rendering, geometric learning, and reinforcement learning.

  • Project
    beehive-lab/TornadoVM

    A GPU programming framework for Java that JIT-compiles bytecode to CUDA, OpenCL, and Metal, enabling high-performance GPU acceleration for Java applications.