halide/Halide

a language for fast, portable data-parallel computation

What it solves

Halide is a programming language designed to simplify the creation of high-performance image and array processing code. It allows developers to write efficient code that can run across a variety of modern hardware architectures without needing to write platform-specific optimizations manually for every target.

How it works

Halide is an embedded language, meaning it is not standalone. It is primarily embedded in C++ (requiring C++17 or later), where developers use a Halide C++ API to build an in-memory representation of a processing pipeline. This pipeline can then be either JIT-compiled (Just-In-Time) or compiled to an object file (Ahead-Of-Time/AOT).

Halide also provides Python bindings, allowing the language to be used within Python without requiring C++.

Who it’s for

It is designed for developers who need to perform high-performance array and image processing on a variety of platforms, including CPUs (X86, ARM, RISC-V, etc.) and GPUs (CUDA, OpenCL, Metal, Vulkan, etc.) across multiple operating systems.

Highlights

  • Broad Hardware Support: Targets a wide range of CPU architectures and GPU Compute APIs.
  • Embedded Nature: Integrates directly into C++ and Python.
  • Flexible Compilation: Supports both JIT and AOT compilation modes.
  • Cross-Platform: Works across Linux, Windows, macOS, Android, and iOS.

Related

  • Project
    ROCm/hip

    HIP is a C++ runtime API and kernel language for AMD GPUs that enables developers to port NVIDIA CUDA code to AMD hardware for portable GPU acceleration.

  • Project
    ermig1979/Simd

    A high-performance C/C++ image processing and machine learning library that uses SIMD CPU extensions to accelerate image operations and neural networks.

  • Project
    facebookincubator/AITemplate

    A Python framework that transforms deep neural networks into optimized CUDA/HIP C++ code for high-performance inference serving on NVIDIA and AMD GPUs.

  • Project
    CVCUDA/CV-CUDA

    CV‑CUDA is NVIDIA’s open‑source GPU‑accelerated computer‑vision library. It provides fast, scalable image/video operators (resize, filter, color conversion, etc.) that run entirely on the GPU and integrate with C/C++ and Python AI frameworks. Install via pre‑built wheels (`cvcuda-cu12` or `cvcuda-cu13`) for Linux, decode images directly to GPU with `nvimgcodec`, and chain CV‑CUDA operators before feeding data to deep‑learning models. The library targets modern NVIDIA GPUs (Turing and newer) and is used in production systems such as Microsoft Bing visual search and Tencent Cloud video processing.

  • Project
    google/highway

    A C++ library providing portable SIMD/vector intrinsics to enable high-performance, architecture-independent software across various CPU platforms.