Crate rten_simd

Portable SIMD types for implementing vectorized functions that work across different architectures.

Compared to std::simd it offers the following benefits:

• Works on stable Rust
• Includes infrastructure for dispatching vectorized operations using the optimal instruction set as determined at runtime.
• Includes higher order functions for vectorized maps, folds etc.

Supported architectures

SIMD wrappers are provided for the following architectures:

• Arm Neon
• AVX 2 / FMA
• AVX-512 (requires avx512 feature and nightly Rust)
• WebAssembly SIMD

There is also a scalar fallback that works on all platforms, but provides no performance benefit over non-SIMD code.

Modules

• arch
  Architecture-specific functionality.
• dispatch
  Dispatch SIMD operations using the preferred SIMD instruction set for the current system, as determined at runtime.
• functional
  Higher order functions (map, fold etc.) that use vectorized operations.
• isa_detection
  Functions for testing the availability of instruction sets at runtime.
• span
  Slice-like types without the restrictions on aliasing.

Traits

• SimdFloat
  Trait for SIMD vectors containing single-precision floats.
• SimdInt
  Trait for SIMD vectors containing 32-bit integers.
• SimdMask
  Trait implemented by SIMD masks.
• SimdVal
  Base trait for SIMD vectors.

Functions

• vec_count
  Return the number of SIMD vectors required to hold count elements.