Mailman 3 December 2006 - SciPy-User

Re: [SciPy-user] Building SciPy on OS X: Universal binary
by Bill Northcott 11 Dec '06

11 Dec '06

On 12/12/2006, at 5:00 AM, Josh Marshall wrote: > However, as a related question, if I have both g77 and gfortran > installed, how do I select gfortran as the compiler? I have tried > various combinations of lines similar to > > python setup.py config --fcompiler=gnu95 build > > but g77 is always selected for the build. > The usual procedure is to define the F77 environment variable: export F77=/path/to/fortrancompiler (on bash) or setenv F77 /path/to/fortrancompiler (for tcsh) Cheers Bill

1 0

looking for Monte Carlo Method for generating normal distribition
by Xiaojian Wang 11 Dec '06

11 Dec '06

*Hi, I can generate uniform random numbers, using random(),* *I would like to know if there is any library in python can generate * *normal distribution numbers.* thanks, Xiaojian

2 1

Re: [SciPy-user] feed-forward neural network for python
by Marek Wojciechowski 11 Dec '06

11 Dec '06

On Wed, 06 Dec 2006 20:16:05 -0000, <scipy-user-request(a)scipy.org> wrote: > I find that the rprop algorithm is the best for real-world-noisy data > (besides GAs), and this could be an easy add. I have a python > implementation that uses scipy to run rprop, there is definite room > for improvement, I'm mainly using scipy.where() and wrote the function > directly from the original rprop paper (i can send you this code if > you'd like it). However, it may be faster to use fortran here as > well. The original c version or rprop is here > http://www.koders.com/c/fid1B403F9BA6B2BAA1C7CEDD1C680F8AAB2819B44E.aspx Yes, adding rprop would be quite easy, and it can be done in fortran for speed. I think it can appear in next release. If you like, just send me your implementation of rprop in python, I'll see if it can be useful... Generally, in future, I planned to have all fortran parts also in python. This is for all these poor people without numpy and scipy. How do you think? Another solution would be to try to include ffnet to scipy distribution. There is actually no support of ANNs in scipy. I saw scipy.sandbox.ann modules, but these, in opposite to ffnet, don't calculate even gradients for minimalization yet. Besides, ffnet has pikaia genetic optimizer bindings (which could be included to scipy.optimize) and many improvements, like automatic normalization of data which makes using ANNs much easier... > I also think that it would be great to have two versions of this. One > for general usage and one for altivec and sse 64bit ibm and intel > chips, which at this stage of the project would be easy to modify > http://developer.apple.com/documentation/Performance/Conceptual/Accelerate_… > gcc 4.0.3 does some auto optimization with the right flags and I've > found you can increase the speed by factors of 10 or more in many > cases. What type of platforms do you have access to? The CHUD tools > on OS X are very effective and might be useful > (http://developer.apple.com/tools/performance/) to find optimizations > relevant to all platforms. Generally I was never interested in optimizing my code for platforms other than x86. I have access to sourceforge compile farm which includes several architectures (but I don't suspect there are altivec sorts of things). I, personally, have no time to investigate special coding for altivec/sse. If you are interested in this and you think it could be really useful, I may consider creating a branch of ffnet for this purpose. And just to remind, ffnet is at https://sourceforge.net/projects/ffnet Greetings -- Marek Wojciechowski

2 1

Re: [SciPy-user] Building SciPy on OS X: Universal binary issues
by Josh Marshall 11 Dec '06

11 Dec '06

Thanks for that line Robert. Executing that command line definitely chooses gfortran. I am now running into this error: compiling C sources C compiler: gcc -arch ppc -arch i386 -isysroot /Developer/SDKs/ MacOSX10.4u.sdk -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -fno-common -dynamic -DNDEBUG -g -O3 creating build/temp.macosx-10.3-fat-2.4/Lib/linsolve creating build/temp.macosx-10.3-fat-2.4/Lib/linsolve/SuperLU creating build/temp.macosx-10.3-fat-2.4/Lib/linsolve/SuperLU/SRC compile options: '-DUSE_VENDOR_BLAS=1 -c' gcc: Lib/linsolve/SuperLU/SRC/zgscon.c gcc: cannot specify -o with -c or -S and multiple compilations gcc: cannot specify -o with -c or -S and multiple compilations error: Command "gcc -arch ppc -arch i386 -isysroot /Developer/SDKs/ MacOSX10.4u.sdk -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -fno-common -dynamic -DNDEBUG -g -O3 - DUSE_VENDOR_BLAS=1 -c Lib/linsolve/SuperLU/SRC/zgscon.c -o build/ temp.macosx-10.3-fat-2.4/Lib/linsolve/SuperLU/SRC/zgscon.o" failed with exit status 1 Any ideas on how to resolve this? It goes away if I don't try to compile as universal binary. Regards, Josh On 12/12/2006, at 5:00 AM, scipy-user-request(a)scipy.org wrote: > Message: 3 > Date: Sun, 10 Dec 2006 17:33:20 -0600 > From: Robert Kern <robert.kern(a)gmail.com> > Subject: Re: [SciPy-user] Building SciPy on OS X: Universal binary > issues > To: SciPy Users List <scipy-user(a)scipy.org> > Message-ID: <457C9940.90004(a)gmail.com> > Content-Type: text/plain; charset=UTF-8 > > Josh Marshall wrote: > >> However, as a related question, if I have both g77 and gfortran >> installed, how do I select gfortran as the compiler? I have tried >> various combinations of lines similar to >> >> python setup.py config --fcompiler=gnu95 build >> >> but g77 is always selected for the build. > > As far as I know, "config --fcompiler=gnu95" only sets the compiler > for the > configuration steps (i.e. finding ATLAS and such). I don't think it > sets the > FORTRAN compiler for any of the subsequent steps. > > python setup.py config --fcompiler=gnu95 build_src build_clib -- > fcompiler=gnu95 > build_ext --fcompiler=gnu95 build > > -- > Robert Kern > > "I have come to believe that the whole world is an enigma, a > harmless enigma > that is made terrible by our own mad attempt to interpret it as > though it had > an underlying truth." > -- Umberto Eco

1 0

Re: [SciPy-user] Building SciPy on OS X: Universal binary issues
by Josh Marshall 10 Dec '06

10 Dec '06

> From: Bill Northcott <w.northcott(a)unsw.edu.au> > Subject: Re: [SciPy-user] Building SciPy on OS X: Universal binary > issues > > On 08/12/2006, at 6:34 PM, Josh Marshall wrote: >> I get an error when importing scipy, as follows: >> >> ImportError: Failure linking new module: /Library/Frameworks/ >> Python.framework/Versions/2.4/lib/python2.4/site-packages/scipy/ >> special/_cephes.so: Symbol not found: _printf$LDBLStub >> Referenced from: /Library/Frameworks/Python.framework/Versions/ >> 2.4/ >> lib/python2.4/site-packages/scipy/special/_cephes.so >> Expected in: dynamic lookup > > These errors are caused by mixing compiler versions. You cannot link > object code files/static libraries if some of the code is built with > gcc3.x.x and some with gcc4.x.x. gcc3.x.x must be used with g77 and > gcc4.x.x with gfortran. You can always link to a dynamic library > (.dylib) regardless of the compiler that built it or even the > language of the source code. I am aware of this inability to mix compilers. What I am stating is that the downloadable, binary versions on the MacPython universal binaries page have this problem, and they should not. I would like to know who built these versions, and how I go about building my own universal binary versions. However, as a related question, if I have both g77 and gfortran installed, how do I select gfortran as the compiler? I have tried various combinations of lines similar to python setup.py config --fcompiler=gnu95 build but g77 is always selected for the build. Cheers, Josh

2 1

ANN: An alternative to numpy.core.ma
by Pierre GM 10 Dec '06

10 Dec '06

All, I just posted on the DeveloperZone of the wiki the latest version of maskedarray, an alternative to numpy.core.ma. You can download it here: http://projects.scipy.org/scipy/numpy/attachment/wiki/MaskedArray/maskedarr… The package has three modules: core (with the basic functions of numpy.core.ma), extras (which adds some functions, such as apply_along_axis, or the concatenator mr_), and testutils (which adds support for maskedarray for the tests functions). It also comes with its test suite (available in the tests subdirectory). For those of you who were not aware of it, the new MaskedArray is a subclass of ndarray, and it accepts any subclass of ndarray as data. You can use it as you would with numpy.core.ma.MaskedArray. For those of you who already tested the package, the main modifications are: - the reorganization of the initial module in core+extras. - Data are now shared by default (in other terms, the copy flag defaults to False in MaskedArray.__new__), for consistency with the rest of numpy. - An additional boolean flag has been introduced: keep_mask (with a default of True). This flag is useful when trying to mask a mask array: it tells __new__ whether to keep the initial mask (in that case, the new mask will be combined with the old mask) or not (in that case, the new mask replaces the old one). - Some functions/routines that were missing have been added (any/all...) As always, this is a work in progress. In particular, I should really check for the bottlenecks: would anybody have some pointers ? If you wanna be on the safe, optimized side, stick to numpy.core.ma. Otherwise, please try this new implementation, and don't forget to give me some feedback! PS: Technical question: how can I delete some files in the DeveloperZone wiki ? The maskedarray.py, test_maskedarray.py, test_subclasses.py are out of date, and should be replaced by the package. Thanks a lot in advance !

1 0

Re: [SciPy-user] Building SciPy on OS X: Universal binary issues
by Bill Northcott 09 Dec '06

09 Dec '06

On 08/12/2006, at 6:34 PM, Josh Marshall wrote: > I get an error when importing scipy, as follows: > > ImportError: Failure linking new module: /Library/Frameworks/ > Python.framework/Versions/2.4/lib/python2.4/site-packages/scipy/ > special/_cephes.so: Symbol not found: _printf$LDBLStub > Referenced from: /Library/Frameworks/Python.framework/Versions/2.4/ > lib/python2.4/site-packages/scipy/special/_cephes.so > Expected in: dynamic lookup These errors are caused by mixing compiler versions. You cannot link object code files/static libraries if some of the code is built with gcc3.x.x and some with gcc4.x.x. gcc3.x.x must be used with g77 and gcc4.x.x with gfortran. You can always link to a dynamic library (.dylib) regardless of the compiler that built it or even the language of the source code. Bill Northcott

1 0

Advantages of Intel's MKL
by Nils Wagner 09 Dec '06

09 Dec '06

Hi all, What are the advantages of Intel's MKL library in connection with numpy/scipy ? Nils

4 7

Re: [SciPy-user] Windows Binary Problem
by R. Padraic Springuel 08 Dec '06

08 Dec '06

I'm still having the compatibility issues between the numpy and scipy Windows binaries, even with the installations that Gary sent me. Does anyone have any other suggestions as to how to get scipy up and running on a Windows machine (numpy is working fine)? -- R. Padraic Springuel Teaching Assistant Department of Physics and Astronomy University of Maine Bennett 309 Office Hours: Wednesday 2-3pm

4 3

Adding a page to the SciPy Wiki??
by Lou Pecora 08 Dec '06

08 Dec '06

I would like to add a page on my C extensions to the Wiki and a link to the tar file of the code, but I am having problems getting started. Several people have pointed me to various SciPy.org pages. I have added a new link to a non-existent page by adding a /MyNewPage on the URL and it says it automatically will add a page, but when I tell it what to add (a blank page) it just says I need to log in to do that. When I look for log in on the SciPy site I don't see it anywhere. I have no idea what the templates mean and whether I should choose one of them. All the pages I go to to try to add a new page are marked Immutable Page in the left column. Clicking on the balloon (Edit) tells me I can't do that. Sorry, gang, this newbie is still lost. -- Lou Pecora My views are my own. ____________________________________________________________________________________ Any questions? Get answers on any topic at www.Answers.yahoo.com. Try it now.

3 6