# mercury_gmp_int

Multi-precision integers for Mercury based on GMP

This provides a binding to a subset of the functions from the `mpz` data-type of
[GMP](http://gmplib.org).

## Usage

To use, you'll need a compiled version of GMP, both shared and static libraries
should work.

Compile your program and link to the library:

```
$ mmc -l gmp $OPTIONS $PROGRAM_NAME
```

or as follows if you want to use static linking (slightly faster):

```
$ mmc --link-object $FULL_PATH_TO_STATIC_GMP_LIB $OPTIONS $PROGRAM_NAME
```

## Benchmark

For heavy number crunching code on bignums, first benchmarks show that
Mercury-GMP is 2.15 -- 2.4x faster than SWI and 1.2 -- 1.3x faster than YAP
(both using GMP) and around 1.35x -- 1.15x slower than Haskell (depending on
dynamic or static linking).

For a Miller-Rabin test, Mercury-GMP was more than two orders of magnitude
faster than using `integer.m` from the standard library. It is between 2 to 6
times faster than Mercury-MP which is based on [libtommath](http://libtom.net).

## Remarks

This only works in the C grades (hlc / asm_fast) with conservative garbage
collection.

Currently, there is basically no error checking.