Four dedicated native MingW-W64 based toolchains for both 32 bit and 64 bit architectures. The main objective of these toolchains is the usage as a convinient build tool for Python extensions for the standard MSVC based Windows Python distributions. Supported Python versions are: 2.6, 2.7, 3.2, 3.3, 3.4.
History
2015-01
- use of
mingwpy
as package name
- upgrade to OpenBLAS-0.2.13
- added
-DMS_WIN64
in the specs for the 64bit builds
- added
libgcc_eh.a
stub library to prevent link errors under certain circumstances
2014-11
2014-07
python2.6-3.2: mingwpy_amd64_vc90.tar.xz or mingwpy_amd64_vc90.7z
python3.3-3.4: mingwpy_amd64_vc100.tar.xz or mingwpy_amd64_vc100.7z
python2.6-3.2: mingwpy_win32_vc90.tar.xz or mingwpy_win32_vc90.7z
python3.3-3.4: mingwpy_win32_vc100.tar.xz or mingwpy_win32_vc100.7z
features
- gcc-4.9.2 build with patched mingw-build scripts
- language support: gfortran, gcc, g++, lto
- statically build gcc runtimes
- uses MingW-W64 trunk
- win32 thread model configuration
- pthread support with winpthreads library (MIT licanse)
- seh structured exception handling model for C++ (64bit)
- sjlj exception handling model for C++ (32 bit)
- ensure correct exclusive MSVCR runtime linkage
- automated linkage of manifest resources to binaries
ftime
patches for OpenMP support with CRT higher than MSVCRT
- patches for MSVC compatible default FPU control word
- OpenBLAS-0.2.13 BLAS/LAPACK library included
OpenBLAS-0.2.13
- OpenBLAS-0.2.13 development release 07ff001
- BLAS and Lapack routines with optimized kernels and thread support
- compiled with
DYNAMIC_ARCH=1
and CPU runtime detection
- the 32 bit OpenBLAS binary is optimized for modern CPU’s and doesn’t work on legacy hardware without SSE2 support
Download and Install
- download and install the appropriate toochain (s. above) for your python installation and make sure it is added to your
PATH
environmanr variable.
- download the appropriate import libraries and copy the import libraries into the
libs
folder of your Python installation: