v0.8.1

• Re-enable use of std::mul_add if the std feature is enabled (except on
  MinGW).

v0.8.0

• Always use libm functions.
• math_funcs feature no longer requires std.
• TwoFloat is now #[repr(C)].

v0.7.0

• Support no_std builds. (Thanks: joseluis)

v0.6.1

• Re-enable MinGW support via libm.

v0.6.0

• Add const constructor from_f64.
• Breaking change: remove MinGW support due to incorrect fma implementation
• Breaking change: fix definition of no_overlap to match definition 1.4 in Joldes et al. (2017).
• Breaking change: Correct values of MIN and MAX constants.

v0.5.0

• Add operator overloads for &f64.
• Breaking change: use serde as the feature flag name.
• Breaking change: use helper struct for Serde serialization.
• Integrate with num_traits crate.
• Internal: use hexf to specify constants.

v0.4.1

• Internals now checked with clippy.

v0.4.0

• Breaking change: update error handling to make it more future-proof.

v0.3.1

• Support Default trait.

v0.3.0

• Breaking change: add specific implementations for exp2 and log2 functions.
• Breaking change: update formatting, support exponential formats and precision specifiers.
• Add optional support for serde (feature "serde_support").
• Add conversions to/from i128 and u128.
• Mathematical functions are optional but enabled by default (feature "math_funcs").