fpmath

fpmath is a pure-Rust floating point library that implements math functions for f32 and f64.

Features

• Less than 1 ULP error in most functions.
• f32 versions do not internally use f64 arithmetic.
• All functions are also implemented for soft-float single-precision and double precision.
• no_std
• All functions are tested for accuracy (MPFR is used to calculate reference values).
• The included generator crate can generate all magic constants used in the algorithms.

Included functions

• Absolute value
• Copy sign
• Rounding (to nearest, towards zero, towards infinity, towards negative infinity)
• Exponential in base e, 2 and 10
• Logarithm in base e, 2 and 10
• Trigonometric (sine, cosine, tangent) in radians, degrees and half-revolutions
• Inverse trigonometric (arcsine, arccosine, arctangent) in radians, degrees and half-revolutions
• Hyperbolic (sine, cosine, tangent)
• Inverse hyperbolic (arcsine, arccosine, arctangent)
• Square and cube root
• Hypotenuse
• Power (floating point and integer exponent)

Development

The code contains some generated constants. They can be identified with a comment that starts with // GENERATE:.

They can be regenerated with:

cargo run -p generator -- src

The generator needs Sollya and Julia to be installed.

License

Licensed under either of

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.