Introduction to gmpy2

gmpy2 is a C-coded Python extension module that supports multiple-precision arithmetic. gmpy2 is the successor to the original gmpy module. The gmpy module only supported the GMP multiple-precision library. gmpy2 adds support for the MPFR (correctly rounded real floating-point arithmetic) and MPC (correctly rounded complex floating-point arithmetic) libraries. gmpy2 also updates the API and naming conventions to be more consistent and support the additional functionality.

The following libraries are supported:

Changes in gmpy2 2.1.0a1

  • Thread-safe contexts are now supported. Properly integrating thread-safe contexts required an extensive rewrite of almost all internal functions.
  • MPFR and MPC are now required. It is no longer possible to build a version of gmpy2 that only supports the GMP library.
  • The function inverse() now raises an exception if the inverse does not exist.
  • Context methods have been added for MPFR/MPC related functions.

Changes in gmpy2 2.0.4

  • Fix bit_scan0() for negative values.
  • Changes to to allow static linking.
  • Fix performance regression with mpmath and Python 3.

Changes in gmpy2 2.0.3

  • Fix lucas2() and atanh(); they were returning incorrect values.

Changes in gmpy2 2.0.2

  • Rebuild Windows binary installers due to MPIR 2.6.0 bug in next_prime().
  • Another fix for is_extra_strong_lucas_prp().

Changes in gmpy2 2.0.1

  • Updated to work in more situations.
  • Corrected exception handling in basic operations with mpfr type.
  • Correct InvalidOperation exception not raised in certain circumstances.
  • invert() now raises an exception if the modular inverse does not exist.
  • Fixed internal exception in is_bpsw_prp() and is_strong_bpsw_prp().
  • Updated is_extra_strong_lucas_prp() to latest version.

Changes in gmpy2 2.0.0

  • Fix segmentation fault in _mpmath_normalize (an undocumented helper function specifically for mpmath.)
  • Improved See below for documentation on the changes.
  • Fix issues when compiled without support for MPFR.
  • Conversion of too large an mpz to float now raises OverflowError instead of returning inf.
  • Renamed min2()/max2() to minnum()/maxnum()
  • The build and install process (i.e. has been completely rewritten. See the Installation section for more information.
  • get_context() no longer accepts keyword arguments.

Known issues in gmpy2 2.0.0

  • The test suite is still incomplete.

Changes in gmpy2 2.0.0b4

  • Added __ceil__, __floor__, __trunc__, and __round__ methods to mpz and mpq types.
  • Added __complex__ to mpc type.
  • round(mpfr) now correctly returns an mpz type.
  • If no arguments are given to mpz, mpq, mpfr, mpc, and xmpz, return 0 of the appropriate type.
  • Fix broken comparison between mpz and mpq when mpz is on the left.
  • Added __sizeof__ to all types. Note: sys.getsizeof() calls __sizeof__ to get the memory size of a gmpy2 object. The returned value reflects the size of the allocated memory which may be larger than the actual minimum memory required by the object.

Known issues in gmpy2 2.0.0b4

  • The new test suite (test/ is incomplete and some tests fail on Python 2.x due to formating issues.

Changes in gmpy2 2.0.0b3

  • mp_version(), mpc_version(), and mpfr_version() now return normal strings on Python 2.x instead of Unicode strings.
  • Faster conversion of the standard library Fraction type to mpq.
  • Improved conversion of the Decimal type to mpfr.
  • Consistently return OverflowError when converting “inf”.
  • Fix mpz.__format__() when the format code includes “#”.
  • Add is_infinite() and deprecate is_inf().
  • Add is_finite() and deprecate is_number().
  • Fixed the various is_XXX() tests when used with mpc.
  • Added caching for mpc objects.
  • Faster code path for basic operation is both operands are mpfr or mpc.
  • Fix mpfr + float segmentation fault.

Changes in gmpy2 2.0.0b2

  • Allow xmpz slice assignment to increase length of xmpz instance by specifying a value for stop.
  • Fixed reference counting bug in several is_xxx_prp() tests.
  • Added iter_bits(), iter_clear(), iter_set() methods to xmpz.
  • Added powmod() for easy access to three argument pow().
  • Removed addmul() and submul() which were added in 2.0.0b1 since they are slower than just using Python code.
  • Bug fix in gcd_ext when both arguments are not mpz.
  • Added ieee() to create contexts for 32, 64, or 128 bit floats.
  • Bug fix in context() not setting emax/emin correctly if they had been changed earlier.
  • Contexts can be directly used in with statement without requiring set_context()/local_context() sequence.
  • local_context() now accepts an optional context.

Changes in gmpy2 2.0.0b1 and earlier

  • Renamed functions that manipulate individual bits to bit_XXX() to align with bit_length().
  • Added caching for mpq.
  • Added rootrem(), fib2(), lucas(), lucas2().
  • Support changed hash function in Python 3.2.
  • Added is_even(), is_odd().
  • Add caching of the calculated hash value.
  • Add xmpz (mutable mpz) type.
  • Fix mpq formatting issue.
  • Add read/write bit access using slices to xmpz.
  • Add read-only bit access using slices to mpz.
  • Add pack()/unpack() methods to split/join an integer into n-bit chunks.
  • Add support for MPFR (casevh)
  • Removed fcoform float conversion modifier.
  • Add support for MPC.
  • Added context manager.
  • Allow building with just GMP/MPIR if MPFR not available.
  • Allow building with GMP/MPIR and MPFR if MPC not available.
  • Removed most instance methods in favor of gmpy2.function. The general guideline is that properties of an instance can be done via instance methods but functions that return a new result are done using gmpy2.function.
  • Added __ceil__, __floor__, and __trunc__ methods since they are called by math.ceil(), math.floor(), and math.trunc().
  • Removed gmpy2.pow() to avoid conflicts.
  • Removed gmpy2._copy and added xmpz.copy.
  • Added support for __format__.
  • Added as_integer_ratio, as_mantissa_exp, as_simple_fraction.
  • Updated rich_compare.
  • Require MPFR 3.1.0+ to get divby0 support.
  • Added fsum(), degrees(), radians().
  • Updated random number generation support.
  • Changed license to LGPL 3+.
  • Added lucasu, lucasu_mod, lucasv, and lucasv_mod. Based on code contributed by David Cleaver.
  • Added probable-prime tests. Based on code contributed by David Cleaver.
  • Added to_binary()/from_binary.
  • Renamed numdigits() to num_digits().
  • Added keyword precision to constants.
  • Added addmul() and submul().
  • Added __round__(), round2(), round_away() for mpfr.
  • round() is no longer a module level function.
  • Renamed module functions min()/max() to min2()/max2().
  • No longer conflicts with builtin min() and max()
  • Removed set_debug() and related functionality.


Installing gmpy2 on Windows

Pre-compiled versions of gmpy2 are available at Downloads . Please select the installer that corresponds to the version of Python installed on your computer. Note that either a 32 or 64-bit version of Python can be installed on a 64-bit version of Windows. If you get an error message stating that Python could not be found in the registry, you have the wrong version of the gmpy2 installer.

Installing gmpy2 on Unix/Linux


gmpy2 has only been tested with the most recent versions of GMP, MPFR and MPC. Specifically, for integer and rational support, gmpy2 requires GMP 5.0.x or later. To support multiple-precision floating point arithmetic, MPFR 3.1.x or later is required. MPC 1.0.1 or later is required for complex arithmetic.

Short Instructions

If your system includes sufficiently recent versions of GMP, MPFR and MPC, and you have the development libraries installed, compiling should be as simple as:

cd <gmpy2 source directory>
python install

If this fails, read on.

Detailed Instructions

If your Linux distribution does not support recent versions of GMP, MPFR and MPC, you will need to compile your own versions. To avoid any possible conflict with existing libraries on your system, it is recommended to use a directory not normally used by your distribution. will automatically search the following directories for the required libraries:

  1. /opt/local
  2. /opt
  3. /usr/local
  4. /usr
  5. /sw

If you can’t use one of these directories, you can use a directory located in your home directory. The examples will use /home/case/local. If you use one of standard directories (say /opt/local), then you won’t need to specify –prefix=/home/case/local to but you will need to specify the prefix when compiling GMP, MPFR, and MPC.

Create the desired destination directory for GMP, MPFR, and MPC.

$ mkdir /home/case/local

Download and un-tar the GMP source code. Change to the GMP source directory and compile GMP.

$ cd /home/case/local/src/gmp-5.1.0
$ ./configure --prefix=/home/case/local
$ make
$ make check
$ make install

Download and un-tar the MPFR source code. Change to the MPFR source directory and compile MPFR.

$ cd /home/case/local/src/mpfr-3.1.1
$ ./configure --prefix=/home/case/local --with-gmp=/home/case/local
$ make
$ make check
$ make install

Download and un-tar the MPC source code. Change to the MPC source directory and compile MPC.

$ cd /home/case/local/src/mpc-1.0.1
$ ./configure --prefix=/home/case/local --with-gmp=/home/case/local --with-mpfr=/home/case/local
$ make
$ make check
$ make install

Compile gmpy2 and specify the location of GMP, MPFR and MPC. The location of the GMP, MPFR, and MPC libraries is embedded into the gmpy2 library so the new versions of GMP, MPFR, and MPC do not need to be installed the system library directories. The prefix directory is added to the beginning of the directories that are checked so it will be found first.

$ python install --prefix=/home/case/local

If you get a “permission denied” error message, you may need to use:

$ python build --prefix=/home/case/local
$ sudo python install --prefix=/home/case/local

Options for

Ignore the timestamps on all files and recompile. Normally, the results of a previous compile are cached. To force gmpy2 to recognize external changes (updated version of GMP, etc.), you will need to use this option.
Force the use of MPIR instead of GMP. GMP is the default library on non-Windows operating systems.
Force the use of GMP instead of MPIR. MPIR is the default library on Windows operating systems.
Specify the directory prefix where GMP/MPIR, MPFR, and MPC are located. For example, –prefix=/opt/local instructs to search /opt/local/include for header files and /opt/local/lib for libraries.
Disables support for MPFR and MPC. This option is intended for testing purposes and is not offically supported.
Disables support MPC. This option is intended for testing purposes and is not officially supported.