[SciPy-dev] Thank you, Intel compilers, MKL+fft

[John Travers]

Hello SciPy developers,

I've started to use SciPy in my work (nonlinear fibre optics) and I
just wanted to say a big thank you to Travis, Pearu, Eric and everyone
else for developing SciPy. I have recently moved a fairly large amount
of numerical code over from Matlab and got an immediate speedup of 30%
Which is very nice. Especially as I can now use the whole of python
rather than just Matlab to write my code with now. (And it's much much
cheaper!!). So thanks very much! Another main advantage I now have
with SciPy is that I can use exactly the same code on my linux
cluster.. but that brings a question:

Have the problems reported by Arnd Baecker in December/January with
the intel c compiler (to do with cephes and more) been solved yet. At
the moment, I can't get SciPy to compile on the cluster system with
the intel c compiler. (It is working very well with gcc + intel ifort
+ MKL though).

Another question:
Has anything been done to use the MKL fft library for the ffts in
SciPy? I'm am currently half way through getting this working (have
complex transforms in 1D working I think), but I don't want to
duplicate. If not, I'll submit a patch as soon as I get it going.

Thanks again for all the hard work!!
John Travers

----------------------------------
Femtosecond Optics Group
Physics Department
Imperial College
Prince Consort Road
London, SW7 2BW