Changes in 0.8.0: * Daniel Mendler added minDenormal. Changes in 0.7.9: * Rename nextup/nextdown to ieeesucc/ieeepred to avoid clashes with libm functions (fixes issue #15, reported by @clinty) Changes in 0.7.8: * Alain O'Dea fixed broken illumos build. * Simplified endianness detection in C code. Changes in 0.7.7: * Johan Kiviniemi added tests for minNormal, maxFinite, epsilon, maxNaNPayload. Changes in 0.7.6: * Fix broken windows build. Changes in 0.7.5: * Added more robust check for endianness (related to issue #5, reported by @juricast). Changes in 0.7.4: * Add C implementation of copysign, copysignf * Remove libm dependency * Anders Kaseorg fixed documentation for epsilon. Changes in 0.7.3: * Bugfix from Björn Buckwalter: any two negative values were approxEqIEEE Changes in 0.7.2: * Fix aliasing warnings in C code Changes in 0.7.1: * Rename package from "ieee" to "ieee754" Changes in 0.7: * Add IEEE type class with instances for Double, Float, CDouble, and CFloat * Add functions for getting/setting NaN payloads * Add succIEEE/predIEEE for advancing up and down the IEEE number line (ported from Tango's nextUp and nextDown) * Add bisectIEEE for midpoints of two numbers on the IEEE number line (ported from Tango's ieeeMean) * Add identicalIEEE for exact (bitwise) equality of IEEE numbers * Add copySign for setting the sign bit of an IEEE number * Add sameSignificandBits for seeing how many significand bits of two IEEE numbers agree, ported from Tango's feqrel * Add nan, infinity, maxFinite, minNormal constants for IEEE numbers * Add maxNum and minNum * Rename maxF and minF to maxNaN and minNaN * Switch to a simpler "~==" comparison for complex numbers * Make "~==" comparison use sameSignificandBits for IEEE types * Make "===" comparison use bitwise equality for IEEE types * Remove old "eqRel" comparisons. * Remove old epsilon' and delta constants * Remove (RealFloat a) => AEq (Complex a) instance in favor of explicit instances for Complex {Double,Float,CDouble,CFloat} Changes in 0.6.1: * Remove AEq instance for CLDouble (thanks to Bjorn Buckwalter)