Mailman 3 September 2008 - SciPy-User

can't build scipy on mac os 10.5.5 intel
by gdahl＠cs.toronto.edu Sept. 19, 2008

Sept. 19, 2008

Hi everyone. I am not sure if my other email account is working so I apologize if this already hit the list. I get the error below when I run python setup.py build with the latest SVN version of scipy. I have numpy (latest svn) and it works in python2.5, which is the version of python I want to use scipy from. I have gcc 4.0.1 and gfortran 4.2.1. This is the command I ran: sudo python setup.py build_src build_clib --fcompiler=gnu95 build_ext --fcompiler=gnu95 build I have pasted some of the output of python setup.py build below. I don't really know what I am doing, but I would really like to get scipy working and I would appreciate any help anyone can give me! I looked around in the archives with google a bit, but nothing seemed to deal with my situation exactly, hopefully I didn't miss anything. - George $ sudo python setup.py build_src build_clib --fcompiler=gnu95 build_ext --fcompiler=gnu95 build . . . building extension "scipy.sparse.linalg.dsolve.umfpack.__umfpack" sources adding 'scipy/sparse/linalg/dsolve/umfpack/umfpack.i' to sources. swig: scipy/sparse/linalg/dsolve/umfpack/umfpack.i swig -python -o build/src.macosx-10.3-i386-2.5/scipy/sparse/linalg/dsolve/umfpack/_umfpack_wrap.c -outdir build/src.macosx-10.3-i386-2.5/scipy/sparse/linalg/dsolve/umfpack scipy/sparse/linalg/dsolve/umfpack/umfpack.i scipy/sparse/linalg/dsolve/umfpack/umfpack.i:192: Error: Unable to find 'umfpack.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:193: Error: Unable to find 'umfpack_solve.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:194: Error: Unable to find 'umfpack_defaults.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:195: Error: Unable to find 'umfpack_triplet_to_col.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:196: Error: Unable to find 'umfpack_col_to_triplet.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:197: Error: Unable to find 'umfpack_transpose.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:198: Error: Unable to find 'umfpack_scale.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:200: Error: Unable to find 'umfpack_report_symbolic.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:201: Error: Unable to find 'umfpack_report_numeric.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:202: Error: Unable to find 'umfpack_report_info.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:203: Error: Unable to find 'umfpack_report_control.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:215: Error: Unable to find 'umfpack_symbolic.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:216: Error: Unable to find 'umfpack_numeric.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:225: Error: Unable to find 'umfpack_free_symbolic.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:226: Error: Unable to find 'umfpack_free_numeric.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:248: Error: Unable to find 'umfpack_get_lunz.h' scipy/sparse/linalg/dsolve/umfpack/umfpack.i:272: Error: Unable to find 'umfpack_get_numeric.h' error: command 'swig' failed with exit status 1

1 0

scipy build fails on Linux-x86-64 with custom ATLAS
by David Warde-Farley Sept. 18, 2008

Sept. 18, 2008

Hi folks, I am having an issue building scipy from SVN. I rolled my own ATLAS and installed it under /usr/local/atlas, and passed the ATLAS environment variable. I also told it to use gfortran as opposed to g77 as that's what was used for ATLAS. The compilation error is as follows: --------------- /usr/bin/gfortran -Wall -Wall -shared build/temp.linux-x86_64-2.5/ scipy/integrate/_odepackmodule.o -L/usr/local/atlas/lib -Lbuild/ temp.linux-x86_64-2.5 -lodepack -llinpack_lite -lmach -lptf77blas - lptcblas -latlas -lgfortran -o build/lib.linux-x86_64-2.5/scipy/ integrate/_odepack.so /usr/bin/ld: /usr/local/atlas/lib/libptf77blas.a(dscal.o): relocation R_X86_64_32 against `a local symbol' can not be used when making a shared object; recompile with -fPIC /usr/local/atlas/lib/libptf77blas.a: could not read symbols: Bad value collect2: ld returned 1 exit status /usr/bin/ld: /usr/local/atlas/lib/libptf77blas.a(dscal.o): relocation R_X86_64_32 against `a local symbol' can not be used when making a shared object; recompile with -fPIC /usr/local/atlas/lib/libptf77blas.a: could not read symbols: Bad value collect2: ld returned 1 exit status error: Command "/usr/bin/gfortran -Wall -Wall -shared build/ temp.linux-x86_64-2.5/scipy/integrate/_odepackmodule.o -L/usr/local/ atlas/lib -Lbuild/temp.linux-x86_64-2.5 -lodepack -llinpack_lite - lmach -lptf77blas -lptcblas -latlas -lgfortran -o build/lib.linux- x86_64-2.5/scipy/integrate/_odepack.so" failed with exit status 1 --------------- I notice it's complaining about -fPIC, except that when I configured ATLAS, I specifically told it that every compiler should use -fPIC: ../configure -m 2400 -b 64 -D c -DPentiumCPS=2400 --prefix=/usr/ local/atlas --with-netlib-lapack=../../lapack-3.1.1/lapack_LINUX.a - Fa acg '-fPIC' That last option, if I am reading the docs correctly, should take care of it, yet SciPy's build still fails. Does anyone have any idea why? Thanks, David

2 2

having problems with fortran compiler on Mac OS X 10.5.2
by Kiana Toufighi Sept. 18, 2008

Sept. 18, 2008

Hi all. I am trying to install SciPy on my PowerMac with OS X 10.5.2 (Leopard) by following the instructions here http://www.scipy.org/Installing_SciPy/Mac_OS_X and I'm getting errors with respect to the fortran compiler. I have personally installed the newest version of the fortran compiler v. 4.2.3 and I obviously have a C compiler. gcc --version: i686-apple-darwin9-gcc-4.0.1 (GCC) 4.0.1 (Apple Inc. build 5465) Copyright (C) 2005 Free Software Foundation, Inc. gfortran --version: GNU Fortran (GCC) 4.2.3 Copyright (C) 2007 Free Software Foundation, Inc. Now I've tried building scipy by the following commands and here are the errors I get. python setup.py build Couldn't match compiler version for 'GNU Fortran (GCC) 4.2.3\nCopyright (C) 2007 Free Software Foundation, Inc. python setup.py config_fc --fcompiler=gfortran build error: don't know how to compile Fortran code on platform 'posix' with 'gfortran' compiler. Supported compilers are: compaq,none,absoft,intel,f,gnu,sun,nag,vast,ibm,gnu95,intelv,g95,intele,pg,lahey,compaqv,mips,hpux,intelev,intelem) So then I try one of these listend supported compilers like gnu95 but it's not installed!! I wish Apple would just include a fortran compiler with their dev tools! Any help/advice is apprecaited. Thanks, Kiana _________________________________________________________________

2 3

Error in nonlinear least squares fit analysis
by Roger Fearick Sept. 18, 2008

Sept. 18, 2008

Hi all, I'm an occasional user of scipy but this is one of the things I have looked at, so I thought I'd better delurk and comment. > This sounds like what I'm looking for -- just to make sure I'm looking > at this right, the diagonal terms of the covariance matrix represent > the variance in the parameters? I.e., for the error in a exponential > fit y = coef[0]*e**(coef[1]*x), the error for coef[1] would be the s> qrt of leastsq's cov_x[1][1], and I can return the best value of the p> arameter as: > coef[1] +/- sqrt(cov_x[1][1]) ? > This may be a simple question, but I'm having great difficulty finding > a source that explains the meaning of all of these various matricies > :) It's not a simple question, actually. The gory details are in something like Numerical Recipes -- in latest (3rd) edition Sect 15.6 (15.6.5). There is a scipy example at www.phy.uct.ac.za/courses/python/examples/moreexamples.html (Non-linear least squares fit) which shows how to process the output of leastsq. Roger. -- Roger Fearick Department of Physics University of Cape Town

2 1

optimize.fmin_bfgs, vectorial epsilon
by Lars Friedrich Sept. 18, 2008

Sept. 18, 2008

Hello, the docstring of scipy.optimze.fmin_bfgs says that the epsilon parameter can be an ndarray: epsilon : int or ndarray I assume, that it should be possible to pass different epsilons for all the different elements of x. But, if I do so, this causes an error, since in optimize.approx_fprime there are the lines: ei[k] = epsilon grad[k] = (f(*((xk+ei,)+args)) - f0)/epsilon that assume that epsilon is scalar. Is this intended? Why does the docstring say 'int', shouldn't this be 'float'? Besides that: I get 'Warning: Desired error not necessarily achieved due to precision loss' on my optimization problem. Could this be due to the fact that my parameters differ by several orders of magnitude? Do I have to rescale my error function to cope with that? Or should a 'per-parameter-epsilon' be enough? Thanks! Lars -- Dipl.-Ing. Lars Friedrich Bio- and Nano-Photonics Department of Microsystems Engineering -- IMTEK University of Freiburg Georges-Köhler-Allee 102 D-79110 Freiburg Germany phone: +49-761-203-7531 fax: +49-761-203-7537 room: 01 088 email: lfriedri(a)imtek.de

3 2

Error in nonlinear least squares fit analysis
by David Lonie Sept. 17, 2008

Sept. 17, 2008

I got some help here earlier about finding a function to fit a function to some exponentially increasing data. I have a few questions: a) fmin vs. leastsq: The method I wrote ended up using the fmin() function to minimize the error vector. What is the difference between fmin and leastsq? Is there an advantage to using either? b) Error in the parameters: I'd like to know the precision that the fitted parameters are good to. Basically, I'd like to know that b = 3.456 +/- 0.003 instead of just b = 3.456. leastsq can return a Jacobian matrix -- will pulling out the diagonal elements of this matrix give me the results I want? Or is there a better way? TIA, Dave

5 5

singular value decomposition
by Chris Lee Sept. 17, 2008

Sept. 17, 2008

Hi All, I have code that needs to repeatedly perform an svd on a matrix (2304, 2304) shape. It seems to take forever to do (I haven't timed it, but approximately two hours passed on the first trial (without completion) and I am now running a second to make sure nothing went wrong. Is this because I am not linked to an optimized BLAS/LAPACK library or something else? I have access to mkl. Is it possible to link an already installed scipy to mkl, or would I have to reinstall? Cheers Chris *************************************************** Chris Lee Laser Physics and Nonlinear Optics Group MESA+ Research Institute for Nanotechnology University of Twente Phone: ++31 (0)53 489 3968 fax: ++31 (0)53 489 1102 ***************************************************

3 4

butterworth filter
by cyril giraudon Sept. 17, 2008

Sept. 17, 2008

Hi, I use scipy 0.6.0 and i try to reproduce the plot of the matlab butter function web documentation (first response for a google request "matlab butter example"). The matlab code is : [z,p,k] = butter(9,300/500,'high'); [sos,g] = zp2sos(z,p,k); % Convert to SOS form Hd = dfilt.df2tsos(sos,g); % Create a dfilt object h = fvtool(Hd); % Plot magnitude response set(h,'Analysis','freq') % Display frequency response In scipy, I write : from scipy.signal import butter, freqz from pylab import show, grid, log, plot b, a = butter(9, 300./500., 'high') fi = freqz(b, a) plot(fi[0], 20*log(abs(fi[1]))) grid() show() Why are the two filters not the same ? Thanks a lot, Cyril.

6 7

Rombouts algorithm for calculating the characteristic polynomial
by Nils Wagner Sept. 16, 2008

Sept. 16, 2008

Hi all, Has someone implemented Rombouts algorithm in Python ? Details are available at http://arxiv.org/pdf/math/9804133v1 Cheers, Nils

1 0

SciPy 08 conference videos?
by Stephen McInerney Sept. 15, 2008

Sept. 15, 2008

Enthought folks, the slides are up at http://conference.scipy.org/ but can you also put up the videos (unedited is fine, for starters) Thanks, Stephen _________________________________________________________________ Get more out of the Web. Learn 10 hidden secrets of Windows Live. http://windowslive.com/connect/post/jamiethomson.spaces.live.com-Blog-cns!5…

1 0