gmp-d 0.1.0
High-level wrapper for GNU MP (GMP)
To use this package, put the following dependency into your project's dependencies section:
gmp-d
D-language high-level wrapper for GNU MP (GMP) library
that aims to
be mostly compatible
with std.bigint.BigInt
(copy construction excluded) and @safe pure nothrow
@nogc
except when converting to string
.
Basics
- Integers (GMP's
mpz_t
) are wrapped ingmp.z.MpZ
- Rationals (GMP's
mpq_t
) are wrapped ingmp.q.MpQ
along with most of their common operations implemented either as member functions, free functions or both.
Sample Usage
MpZ
import gmp.z;
@safe pure nothrow @nogc:
unittest
{
alias Z = MpZ;
assert(Z.mersennePrime(15) == 2^^15 - 1);
const a = 42.Z;
const b = a.dup; // explicit copying required
}
MpQ
import gmp.q;
@safe pure nothrow @nogc:
unittest
{
alias Q = MpQ;
Q x = Q(-4, 6);
assert(x.numerator == -4);
assert(x.denominator == 6);
x.canonicalize();
assert(x.numerator == -2);
assert(x.denominator == 3);
assert(inverse(x) == Q(-3, 2));
assert(Q(1, 2) + Q(1, 3) == Q(5, 6));
}
Features
Copy construction is disabled by default (for now) to prevent inadvertent
copying. Instead use f(move(z))
or f(z.move)
(from std.algorithm.mutation
)
to pass by move or f(z.dup)
to pass by value (via .dup
member function).
If you really need to have copy construction you can use CopyableMpZ
.
Implementation is optimized through
mapping of GMP's C macros into D inline functions that operate directly on the internal C-representations
__mpz_struct
and__mpq_struct
,passing of
MpZ
-typed parameters asauto ref const
in combination with conditional compilation for r-valueMpZ
passed parameters via__traits(isRef)
. This enables clever reuse ofmpz_t
instances when passed to__gmpz
-functions. For instance,x + 42.Z
can be lowered to
__gmpz_add(rhs._ptr, this._ptr, rhs._ptr);
return move(rhs);
in
opBinary!"+"()(auto ref const MpZ rhs)
Compilation Performance
Compilation is very fast; DMD compiles gmp-d
in 0.04 seconds on my machine and
test-build in 0.22 seconds.
Run-time Performance
Wrapper is as light-weight as possible; Some D-functions directly access the internal C-datastructure exposed by the structs and macros in the C-headers of GNU GMP.
Limitations
Note that D's __traits(isRef)
currently cannot be used to distinguish l-value
from r-value passing of this
(it should). This severly limits the
possibilities of using
C++-style
expression templates to
realize lazy evaluation in operator overloading. If this limitation were to be
fixed, this library could implement lowering of expressions such
Z result = base^^exp % modulo
to the builtin
__gmpz_powm(result._ptr, base._ptr, expr._ptr, modulo._ptr)
See the unittests for MpzAddExpr
, MpzMulExpr
, etc for details on how this
currently can be implemented and verified (in ccc
-version) with free
functions such add
and sub
.
Future
There are more mpz_t
- and mpq_t
-functions that should be wrapped but these are good start.
- Registered by Per Nordlöw
- nordlow/gmp-d
- Boost
- Per Nordlöw