https://github.com/ajtribick/twofloat

A double-double arithmetic library for Rust
https://github.com/ajtribick/twofloat

extended-precision floating-point numerics rust-library

Last synced: 9 months ago
JSON representation

A double-double arithmetic library for Rust

• Host: GitHub
• URL: https://github.com/ajtribick/twofloat
• Owner: ajtribick
• License: bsd-3-clause
• Created: 2020-03-01T11:25:56.000Z (almost 6 years ago)
• Default Branch: main
• Last Pushed: 2025-02-02T18:52:51.000Z (11 months ago)
• Last Synced: 2025-04-09T15:07:39.519Z (9 months ago)
• Topics: extended-precision, floating-point, numerics, rust-library
• Language: Rust
• Homepage:
• Size: 310 KB
• Stars: 17
• Watchers: 3
• Forks: 8
• Open Issues: 9
• Metadata Files:
  • Readme: README.md
  • Changelog: CHANGELOG.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# twofloat

[![Crate](https://img.shields.io/crates/v/twofloat)](https://crates.io/crates/twofloat)

**A double-double arithmetic library for Rust.**

This library provides an implementation of double-double arithmetic for the

Rust language. Note that this is not the same as the IEEE quadruple-precision

floating-point format. Instead, higher precision is obtained by representing

the value as the sum of two non-overlapping `f64` values.

Operator traits are implemented both for reference and value types where

appropriate. The code indicates the source of the algorithms used.

Mathematical constants are available in the `twofloat::consts` module, which

provides the same set of constants as `std::f64::consts`.

Please note that the implementation of the mathematical functions (`exp`,

`powf`, etc.) is very preliminary. In particular, they are calculated using

operations at the same precision as the result, so they will not return values

which are correct to the full precision of the `TwoFloat` type. This may be

addressed in future releases.

## Optional features

* `math_funcs` - include mathematical functions (enabled by default)

* `serde` - enable serialization/deserialization with Serde.

* `std` - use std mathematical functions instead of libm.

## Known issues

* The MinGW `fma` implementation appears to give incorrect results in some

  cases, so the libm implementation is always used on this platform.

## References

* Mioara Joldeș, Jean-Michel Muller, Valentina Popescu. Tight and rigourous

  error bounds for basic building blocks of double-word arithmetic. ACM

  Transactions on Mathematical Software, Association for Computing Machinery,

  2017, 44 (2), pp. 1-27. 10.1145/3121432. hal-01351529v3

* Alan H. Karp, Peter Markstein. High Precision Division and Square Root. ACM

  Transactions on Mathematical Software, Association for Computing Machinery,

  1997, 23 (4), pp. 561-589. 10.1145/279232.279237.

* S. Chevillard, M. Joldeș and C. Lauter. Sollya: an environment for the

  development of numerical codes. Mathematical Software - ICMS 2010, pp.

  28–31.