Why use it

A population count gives the size of a bitset without visiting each bit. Applied to XOR it gives Hamming distance; applied to AND it counts an intersection. POPCNT exposes that operation on a single unsigned integer.

Operations

import native.x86.popcnt; provides native::popcnt<Arch>(value). native.x86 and native re-export it. The overloads accept std::uint16_t, std::uint32_t or std::uint64_t and return the same unsigned type.

Zero returns zero. An all-one value returns its width. Calls are noexcept.

Caveats

Runtime calls need x86_feature::popcnt, a "popcnt" caller target and CPU admission. POPCNT is independent of SSE and BMI and needs no vector OS state. The 16-bit overload counts a zero-extended value through Clang's 32-bit intrinsic; the public result remains a 16-bit count.

Arch is an isa<x86>. Its scalar default is the provider's NATIVE_BASELINE, unaffected by target attributes in the importer. Feature-bearing overloads are constexpr with native runtime paths; tags without POPCNT have consteval overloads only. No instruction-flags or exact-encoding promise follows from the spelling.

Generic popcount covers portable scalar and vector code. Use popcnt when the x86 feature is part of the algorithm's target contract.

See Intel's POPCNT instruction entry.