# google/highway

Performance-portable, length-agnostic SIMD with runtime dispatch

Repository: https://github.com/google/highway
Canonical: https://ross.abutalabs.com/products/highway
Language: C++
License: NOASSERTION
License Family: other
Topics: simd, simd-instructions, simd-programming, intrinsics, avx2, avx512, neon, wasm, avx, avx-512, avx-instructions, sse42, simd-library, simd-parallelism, simd-intrinsics
Last push: 2026-08-26T16:26:30+00:00

## Health v2 (maintenance only)
Score: 81/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 48, longevity 100
- inputs: {"age_days": 2553, "days_push": 7, "days_rel": 132, "gap_med": 252, "n_releases_24m": 2}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5779, forks 463 (observed 2026-08-28T04:09:29.544356+00:00)

## What it is
Highway is a C++ library providing performance-portable SIMD/vector intrinsics that work across seven CPU architectures. It supports runtime dispatch to select the best available instruction set, or static targeting of a single instruction set with the same application code.

## Use cases
- write SIMD code that runs on x86, ARM, and WASM without rewriting
- speed up image processing kernels with vector instructions
- accelerate compression or cryptography routines with SIMD
- avoid fragile compiler autovectorization with explicit vector intrinsics
- dispatch to AVX-512 or NEON at runtime on heterogeneous hardware
- implement fast sorting or linear algebra with data-parallel code

## When to choose
- you need predictable SIMD performance across multiple CPU architectures
- you want runtime CPU feature dispatch without duplicating code
- you need fine-grained control over vector instructions instead of relying on autovectorization
- you are building performance-critical C++ software like codecs, image processing, or math libraries

## When to avoid
- your workload is not data-parallel and gains little from SIMD
- you want a higher-level parallelism framework like OpenMP or TBB for task parallelism
- you are not writing C++ or cannot use C++17
- you only target a single CPU architecture and compiler intrinsics suffice

## Facets
- artifact type: library
- maturity: stable
- function: math, image-processing, compression, cryptography
- domain: performance, graphics, developer-tools
- platform: cpp, cross-platform, wasm
- tags: simd, vectorization, runtime-dispatch, intrinsics, avx2, avx512, neon, sse, algorithms, gpu

## Member repositories
- google/highway (main) score 81

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:29.544356+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:52:50.437080+00:00, confidence not recorded.
  - readme: https://github.com/google/highway (fetched 2026-08-28T04:09:29.544356+00:00, sha c6bd01a0f33c)
- Data as of 2026-08-30T08:39:29.467469+00:00.