The floating-point, its presentation, IEEE 754 standards, special values, rounding modes, and some problems with its implementation are presented: The Table Maker's Dilemma and Round-trip conversions. Arbitrary fixed precision floating-point is an extension of floating-point implemented in Multiple Precision Floating-Point Reliable Library (MPFR). Working towards an interface for a purely functional language, Haskell as a representative is introduced. For an efficient implementation good knowledge of one of its implementation -- GHC -- is needed, especially how memory management is done and how low-level stages of the compiler are implemented. Last, the interface for the GHC and its benchmark with C implementation is presented.
|
