Why use it

The highest set bit determines an unsigned integer's magnitude. A leading-zero count turns that position into a direct calculation for normalization, bit-width selection and radix bucketing, including a defined result for zero.

Operations

import native.x86.lzcnt; provides native::lzcnt<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 16, 32 or 64; a value with its highest bit set returns zero. All calls are noexcept.

Caveats

Runtime calls need x86_feature::lzcnt, a "lzcnt" caller target and CPU admission. LZCNT is independent of BMI and needs no vector OS state. On a CPU without the feature, its encoding can execute as BSR, with different semantics.

Arch is an isa<x86>. The scalar default is the provider's NATIVE_BASELINE; a caller target attribute does not change that default. Feature-bearing overloads are constexpr with native runtime paths. Tags without LZCNT have consteval overloads only. The wrappers return the count, without an instruction-flags contract.

Use std::countl_zero when the algorithm should work across architectures; lzcnt is the explicit x86 feature-gated spelling.

See Intel's LZCNT instruction entry.