Release 1.5 (2017-05-26):
- Install all .cmx files, improving performance of clients and
  avoiding a warning from OCaml 4.03 and up.
- Z.of_float: fix a bug in the fast path [Richard Jones]
  (See https://bugzilla.redhat.com/show_bug.cgi?id=1392247)
- Improve compatibility with OCaml 4.03 and up
  [Bernhard Schommer]
- Overflow issue in Z.pow and Z.root with very large exponents (GPR#5)
  [Andre Maroneze]
- Added function Q.to_float.

Release 1.4.1 (2015-11-09):
- Fixed ml_z_of_substring_base and Z.of_substring [Thomas Braibant]
- Integrated Opam fix for Perl scripts [Thomas Braibant]

Release 1.4 (2015-11-02):
- Improvements to Q (using divexact) [Bertrand Jeannet]
- Fixed div_2exp bug [Bertrand Jeannet]
- Improvements for divexact [Bertrand Jeannet]
- Added of_substring, with fast path for native integers [Thomas Braibant]
- Added Z.powm_sec (constant-time modular exponentiation)
- Reimplemented Z.to_float, now produces correctly rounded FP numbers
- Added Z.trailing_zeros.
- Added Z.testbit, Z.is_even, Z.is_odd.
- Added Z.numbits, Z.log2 and Z.log2up.
- PR#1467: Z.hash is declared as "noalloc" [François Bobot]
- PR#1451: configure fix [Spiros Eliopoulos]
- PR#1436: disable "(void)" trick for unused variables on Windows [Bernhard Schommer]
- PR#1434: removed dependencies on printf & co when Z_PERFORM_CHECK is 0 [Hannes Mehnert]
- PR#1462: issues with Z.to_float and large numbers.

Release 1.3 (2014-09-03):
- Fixed inefficiencies in asm fast path for ARM.
- Revised detection of NaNs and infinities in Z.of_float
- Suppress the redundant fast paths written in C if a corresponding
  fast path exists in asm.
- Use <stdint.h> to ensure compatibility with OCaml 4.02.
- More prudent implementation of Z.of_int, avoids GC problem
  with OCaml < 4.02 (PR#6501 in the OCaml bug tracker).
- PR#1429: of_string accepts 'a' in base 10.
- Macro change to avoid compiler warnings on unused variables.

Release 1.2.1 (2013-06-12):
- Install fixes

Release 1.2 (2013-05-19):
- Added fast asm path for ARMv7 processors.
- PR#1192: incorrect behavior of div_2exp
- Issue with aggressive C compiler optimization in the fast path for multiply
- Better support for Windows/Mingw32

Release 1.1 (2012-03-24):
- Various improvements in the asm fast path for i686 and x86_64
- PR#1034: support for static linking of GMP/MPIR
- PR#1046: autodetection of ocamlopt and dynlink
- PR#1048: autodetection of more platforms that we support
- PR#1051: support architectures with strict alignment constraints for 
    64-bit integers (e.g. Sparc)
- Fixed 1-bit precision loss when converting doubles to rationals
- Improved support for the forthcoming release 4.00 of OCaml

Release 1.0 (2011-08-18):
- First public release
