f32: Wasm type

Syntax

wat

;; Function returning an f32 constant
(func (result f32)
  f32.const 3.14)

;; f32 parameter and local
(func (param $p f32) (local $tmp f32)
  ;; ...
)

;; Mutable f32 global
(global $ratio (mut f32) (f32.const 1.0))

f32 values follow the IEEE 754 binary32 (single-precision) format: one sign bit, eight exponent bits, and twenty-three significand bits. They cover roughly seven decimal digits of precision over the range ±3.4 × 10³⁸, plus the special values ±0, ±∞, and NaN.

Arithmetic instructions (f32.add, f32.mul, and so on) follow IEEE 754 round-to-nearest-ties-to-even semantics. The exact bit pattern of a NaN produced by arithmetic is non-deterministic, so do not rely on specific NaN payloads being the same across implementations or architectures.

f32 is transparent: its bit pattern is observable, and f32 values may be stored in linear memory. f32.reinterpret_i32 and its inverse let you move bits between f32 and i32 without conversion.