simde ☆

« Back to VersTracker

Description:
Implementations of SIMD intrinsics for systems which don't natively support them

Type: Formula | Tracked Since: Dec 28, 2025

Links: Homepage | @simd_everywhere | formulae.brew.sh

Category: Developer tools

Tags: simd intrinsics performance portability vectorization cryptography

Install: brew install simde

About:
SIMD Everywhere (SIMDe) provides header-only implementations of SIMD intrinsics for architectures that lack native support. It enables developers to write portable SIMD code once and compile it across various platforms without rewriting architecture-specific implementations. This library significantly lowers the barrier to entry for high-performance computing on non-standard hardware.

Key Features:

Header-only library requiring no compilation or linking
Supports over 100 instruction sets including SSE, AVX, NEON, and SVE
Zero-overhead abstractions with runtime CPU feature detection
Seamless fallback to scalar operations when SIMD is unavailable

Use Cases:

Porting high-performance multimedia applications to embedded systems
Optimizing machine learning inference on diverse hardware architectures
Cross-platform game engine development requiring consistent SIMD behavior

Alternatives:

Agner Fog's VCL – VCL is C++ specific and focused on x86, while SIMDe is C-compatible and supports broader architectures
Standard C++ Parallel Algorithms – Standard algorithms provide higher-level abstractions but lack fine-grained SIMD control

Version History

Detected	Version	Rev	Change	Commit