[SciPy-Dev] Removal of fortran FFTPACK

Ralf Gommers ralf.gommers at gmail.com
Thu Jul 25 14:37:54 EDT 2019


On Thu, Jul 25, 2019 at 11:08 AM Eric Larson <larson.eric.d at gmail.com>
wrote:

> I plan to merge Peter's PR this coming Monday unless there are objections.
>

Sounds good to me.

Cheers,
Ralf


> Eric
>
>
> On Thu, Jul 25, 2019 at 11:00 AM Peter Bell <PeterBell10 at live.co.uk>
> wrote:
>
>> Hello all,
>>
>>
>>
>> As of gh-10493 <https://github.com/scipy/scipy/pull/10493>, pypocketfft
>> now implements the discrete sine and cosine transforms meaning it is a
>> complete replacement for the fortran FFTPACK library. In gh-10507
>> <https://github.com/scipy/scipy/pull/10507>, I have proposed removing
>> the fortran code from scipy completely and having scipy.fftpack depend on
>> scipy.fft.
>>
>>
>>
>> This means that scipy.fftpack will now see some of the additional benefits
>> of scipy.fft
>> <https://github.com/scipy/scipy/wiki/Release-note-entries-for-SciPy-1.4.0#scipyfft-added>
>> such as: long double transforms, GIL release, bluestein’s algorithm,
>> improved accuracy and improved multi-dimensional transform performance.
>>
>>
>>
>> However, there may be some cases where this affects users. For example,
>> it will break anyone importing the private scipy.fftpack._fftpack submodule
>> and the DCT and DST type-I are now about two times slower (but orders of
>> magnitude more accurate).
>>
>>
>>
>> Regards,
>>
>> Peter
>>
>>
>>
>>
>> _______________________________________________
>> SciPy-Dev mailing list
>> SciPy-Dev at python.org
>> https://mail.python.org/mailman/listinfo/scipy-dev
>>
> _______________________________________________
> SciPy-Dev mailing list
> SciPy-Dev at python.org
> https://mail.python.org/mailman/listinfo/scipy-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scipy-dev/attachments/20190725/d968bbf2/attachment.html>


More information about the SciPy-Dev mailing list