%
%   what hfloat is (good for)
%   -------------------------
%

hfloat (for huge floats) is a library package for doing
calculations with floating point numbers of extreme precision.
It is optimised for computations with 1000...several million digits. 
The computations can be done in (almost) arbitrary radix.

The library contains routines for 
add, subtract, multiply, divide,
n-th power, square root, n-th root, 
logarithm, exp, polynom, poly-root and many more.

There are implementations of several
superlinear converging algorithms for
the computation of pi=3.14159265... (in src/pi/).
On a PC or workstation the computation
of 1 million decimal pi-digits takes less than 1h
(1h on a i486/100, about 20min on an AMD K6/233).

Code examples for the usage of the lib are in simpleex/.

Code for a binary that collects the pi algorithms
can be found in examples/.

Included is the fxt-library, containing many FFT-implementations, 
code for convolution, correlation, spectrum and much more.
It has its own documentation in src/fxt/.

High precision computations test your systems reliability
(hardware and compiler): every little error results in 
garbage digits (or coredump)

Digits of appropriate constants can be used as high quality
random numbers eg. for cryptographic stuff.

You may (and are encouraged to) use the fast multiplication
in your own noncommercial bignum software.

hfloat (and the accompanying texts) may help you to
learn about the things mentioned above.
