halide/Halide

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

6590 stars · 1105 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Halide is an embedded DSL (in C++ and Python) for writing high-performance, data-parallel image and array processing pipelines. It separates the algorithm from its schedule, compiling pipelines ahead-of-time or via JIT to CPUs (x86, ARM, Hexagon, RISC-V, WebAssembly) and GPUs (CUDA, OpenCL, Metal, DirectX 12, Vulkan).

Use cases

• write fast image processing pipelines
• optimize code for cpu and gpu without rewriting
• schedule separable image filters like blurs
• compile image processing for mobile dsp (hexagon)
• accelerate tensor and array computations
• cross-compile performance kernels for multiple platforms
• auto-schedule image processing code

When to choose

• you need high-performance image or array processing across cpu and gpu targets
• you want to tune performance (tiling, vectorization, parallelism) separately from algorithm code
• you target heterogeneous hardware including DSPs and mobile GPUs
• you want JIT or ahead-of-time compilation of processing pipelines

When to avoid

• you need a general-purpose programming language for arbitrary applications
• your workload is scalar control-flow-heavy rather than dense array computation
• you want a simple drop-in library without learning scheduling concepts
• you need small, dependency-free image utilities

Facets

library · maturity stable

compiler image-processing gpu-computing programming-language math image-processing compilers performance computer-vision gpu-computing programming-languages cpp python windows wasm cross-platform dsl data-parallel-computation scheduling aot-compilation jit-compilation auto-scheduler tensor-processing embedded-language linux macos android ios gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="halide/Halide")