Crate simd_popcnt 

simd-popcnt

Count the number of 1 bits (bit population count, a.k.a. Hamming weight) in an array as quickly as possible using specialized CPU instructions: POPCNT, AVX2 and AVX512 on x86/x86-64, and NEON and SVE on AArch64. The fastest instruction set the CPU supports is detected once at runtime and cached; on every other architecture the count falls back to u64::count_ones, which the compiler lowers to a hardware popcount instruction wherever one exists.

The crate is portable by default and thread-safe. It has no external crate dependencies and needs the Rust standard library only for runtime SIMD dispatch (CPU feature detection); it is otherwise no_std.

This is an AI-assisted Rust port of the libpopcnt C/C++ library.

Usage

popcnt counts the 1 bits in a byte slice; the PopcntExt trait adds a .popcnt() method to slices, arrays and Vecs of every built-in integer type.

use simd_popcnt::{popcnt, PopcntExt};

assert_eq!(popcnt(&[0xFF, 0x0F]), 12);
assert_eq!([u64::MAX, 0x0F0F_0F0F_0F0F_0F0F].popcnt(), 96);

Performance

For the fastest possible code, compile with RUSTFLAGS="-C target-cpu=native". This selects the best SIMD path at compile time and removes the runtime dispatch entirely.

Traits

PopcntExt

Adds a popcnt method to slices of the built-in integer types, counting their bits without a manual byte cast. Implemented for slices, arrays and Vecs of u8/u16/u32/u64/u128/usize and their signed counterparts; bring it into scope with use simd_popcnt::PopcntExt;.

Functions

popcnt

Counts the number of one bits (population count) in bytes.