f64: Wasm type
The f64 value type holds a 64-bit double-precision floating-point value.
Syntax
;; Function returning an f64 constant
(func (result f64)
f64.const 3.141592653589793)
;; f64 parameter and local
(func (param $p f64) (local $tmp f64)
;; ...
)
;; Mutable f64 global
(global $ratio (mut f64) (f64.const 1.0))
Description
f64 values follow the IEEE 754 binary64 (double-precision) format: one sign bit, eleven exponent bits, and fifty-two significand bits. They cover roughly fifteen to seventeen decimal digits of precision over the range ±1.8 × 10³⁰⁸, plus the special values ±0, ±∞, and NaN.
Arithmetic instructions (f64.add, f64.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.
f64 is transparent: its bit pattern is observable, and f64 values may be stored in linear memory. f64.reinterpret_i64 and its inverse let you move bits between f64 and i64 without conversion.
JavaScript boundary
f64 matches the representation of JavaScript Number exactly, so values cross the boundary without loss of precision (with the one exception that JavaScript has only a single NaN value).
Specifications
| Specification |
|---|
| Unknown specification # syntax-numtype |