@hackage / fp-ieee

IEEE 754-2019 compliant operations

Latest0.1.0.6

About

Metadata

  • Last updated , by aratamizuki
  • License BSD-3-Clause
  • Categories Mathematics

  • Maintained by: minorinoki@gmail.com

  • Lottery factor: 1

Links

Installation

Package Flags

Use the -f option with cabal commands to enable flags

    pure-hs (off by default)

    Disable FFI

    sse4_1 (off by default)

    Use SSE4.1 instructions on x86

    fma3 (off by default)

    Use FMA3 instructions on x86. On GHC 9.8 or later, this flag enables use of FMA primitives.

    f16c (off by default)

    Use F16C instructions on x86

    float128 (off by default)

    Support Float128 via float128 package

    half (off by default)

    Support Half (float16) via half package

    integer-gmp (on by default)

    Use integer-gmp package on GHC 8.x

    ghc-bignum (on by default)

    Use ghc-bignum package on GHC 9.x

Readme

fp-ieee: IEEE 754 operations for floating-point types

This library provides IEEE 754-compliant operations, including

  • fusedMultiplyAdd.
  • correctly-rounding versions of fromInteger.
  • realFloatToFrac, which correctly handles signed zeros, infinities, and NaNs (unlike realToFrac).

Some operations (e.g. fusedMultiplyAdd) can make use of the native instruction in the architecture via C FFI, or GHC 9.8's FMA primitives.

For non-native targets, "Pure Haskell" mode is supported via a package flag.

Most operations require only RealFloat constraint, but RealFloatNaN is needed by some operations that access the sign and payload of NaNs.