| AArch64 | |
| ppc64le | |
| s390x | |
| x86-64 |
- Update to version 0.12.0:
+ scipy.fft interface This interface operates like the existing
scipy.fftpack interface, but matches the API of the newer
scipy.fft module introduced in SciPy 1.4.
+ test suite was updated to be compatible with more recent dask
(gh#pyFFTW/pyFFTW#278).
+ Cython variable _N was renamed to avoid a name conflict with a
preprocessor token on some platforms (gh#pyFFTW/pyFFTW#259).
+ Cython code has been updated to explicitly use
`language_level=3str` for compatibility with a future Cython
3.0 release.
- Drop 265.patch: Incorporated upstream.
- Update URL and Source URL to point to new github repository.
- Add patch:
* 265.patch (Fix Factory builds)
https://github.com/pyFFTW/pyFFTW/pull/265
- Update to version 0.11.1
+ New features
* Dask interface
* Fast transform planning utility
* Expanded support for norm keyword argument in the numpy interfaces
* Support for norm keyword argument in FFTW builders
* Dynamic library detection at build and run time
* OpenMP threading support
* Custom Configuration of Planners and Interfaces
+ Bugs Fixed
* A platform-dependent bug that results in potentially overwriting a previously
computed output upon repeated calls to the numpy interfaces was fixed (#199).
* Fix to potential platform-dependent integer overflow in empty_aligned (#192).
* rfftfreq is now present in the numpy fft interfaces for numpy >= 1.8 (#207)
+ Other changes
* float16 inputs are now transformed using single rather than double precision.
* The default planning for the numpy and scipy interfaces has changed from
FFTW_MEASURE to FFTW_ESTIMATE. This results in faster planning. In cases
where the same transform is to be repeated many times, it is likely
advantageous to manually specify FFTW_MEASURE instead (or use the FFTW builders
to pre-plan the FFT).
* FutureWarnings related to NumPy multiindexing in NumPy 1.15 are avoided by
using more modern indexing conventions.
* version number handling is now automatically handled by versioneer
* All documentation is now built and hosted at Read the Docs (
http://pyfftw.readthedocs.io).
- Replace future aims in description.
- Implement single-spec version - Update to version 0.10.4 + Fixed bugs: * Numpy interface can fail with non-writeable arrays. * undefined symbol: simd\_alignment with gcc 5.3.0 * FTBFS: TypeError: can't pickle Cython.Compiler.FlowControl.NameAssignment objects + Closed issues: * pyfftw fails with ImportError/undefined symbol fftwl\_plan\_with\_nthreads * pyfftw does not appear to be using available wisdom files * Accuracy of non-power 2 data * Cannot find fftw3l when installing from pip * Incorrect links to docs * PyFFTW returning all zero array after transform... * Move to separate pyFFTW project page * Release GIL during planning? * Merging/sharing wisdom * scipy interface patching. * very slow test suite * setup: some targets should not require numpy to be installed * 2x margin on the planning timelimit test is inadequate in windows on Complex64
- update to version 0.10.1:
* it seems that pypi has an annoying feature that minor tweaks can't
be pushed without a new release.
- changes from version 0.10.0:
* Closed issues:
+ Conda downloads are failing
+ Python 3.4 and WinPython
+ Installing pyfftw on Anaconda3 on Windows 7
+ is python 3.5 supported?
+ deadlock of cache handler at interpreter shutdown
+ pyFFTW breaks when forked
+ build with mingw
+ Striding in n\_byte\_align not on a uniform standard
+ No exception on wrong arguments of function call of
pyfftw.FFTW\(...\)
+ pyfftw vs numpy.fft: faster and slower
+ simple transposes?
+ `Datatype not supported` with scipy
+ Update tutorial with new byte align functions
+ OS X Installation errors:
+ Wrong results for pyfftw's ifft?
+ Installing on OS X Mavericks
+ Install error. Ld cannot find -lfftw3f
+ new source release with updated licensing
+ Crash during initialization of FFTW plan for r2c and c2r with
Intel compiler
+ Move FFTW class definition to pyfftw.pxd
+ Provide transform metadata such as axes and direction
+ Provide shape and dtype properties
+ problem with very large arrays: OverflowError: value too large
to convert to int
+ add support for in-place multidimensional r2c transform
+ Add numpy interface for hfft
+ add cython as a build dependency
+ Potential memory leak in caching
+ Allow GIL to be released with threads=1
+ Building for 64bit windows
+ Test failure using numpy 1.6.2
+ Remove the requirement for users to specify alignment
+ pyfftw.interfaces can only handle numpy arrays
* Merged pull requests:
+ Release GIL during both single- and multi-thread execution
+ Support FFTW\_WISDOM\_ONLY
+ Release GIL around FFTW planning
+ Updated the tutorial to reflect changes with new aligned array
+ Add description for installing on OS X
+ close issue \#8 (More numpy like aligned array creation)
+ Discrete sine transform imports
- cleanup package
- initial package