Performance-portable, length-agnostic SIMD with runtime dispatch
Implementations of SIMD instruction sets for systems which don't natively support them.
C++ wrappers for SIMD intrinsics and parallelized, optimized mathematical functions (SSE, AVX, AVX512, NEON, SVE, WebAssembly, VSX, RISC-V))