Mailman 3 January 2008 - NumPy-Discussion

Error during scipy.fftpack._fftpack compilation ( AIX/cc_r)
by Hans-Joachim Ehlers 29 Jan '08

29 Jan '08

Given: AIX 5.3 xlc v8 xlf v10 During the build scipy the following error shows up: ..... building 'scipy.fftpack._fftpack' extension compiling C sources C compiler: cc_r -DNDEBUG -O2 -qmaxmem=-1 ... "scipy/fftpack/src/zfftnd_fftpack.c", line 92.5: 1506-046 (S) Syntax error. "scipy/fftpack/src/zfftnd_fftpack.c", line 93.5: 1506-046 (S) Syntax error. "scipy/fftpack/src/zfftnd_fftpack.c", line 115.13: 1506-046 (S) Syntax error. error: Command "cc_r -DNDEBUG -O2 -qmaxmem=-1 -Ibuild/src.aix-5.3-2.5 -I/usr/local/lib/python2.5/site-packages/numpy/core/include -I/usr/local/include/python2.5 -c scipy/fftpack/src/zfftnd.c -o build/temp.aix-5.3-2.5/scipy/fftpack/src/zfftnd.o" failed with exit status 1 .... The problem comes from C++ comment lines within c-code. The IBM Compiler v8 invoked as cc_r will complain and stop compilation. Solution: 1) Remove C++ comments from source code or 2) add the option "-qcpluscmt" for the cc_r compiler into the /etc/vac.cfg file or build python with this option. ( -qcpluscmt should be the default for cc_r but it does not afaics ) >From xlc ... -qcpluscmt | -qnocpluscmt Permits the usage of "//" to introduce a comment that lasts until the end of the current source line, as in C++. The default is -qcpluscmt when you invoke the compiler with xlc, xlc_r, cc, or cc_r, or when -qlanglvl is set to stdc99 or extc99. Otherwise, the default setting is -qnocpluscmt. .... regards Hajo BTW: Is there any way to add C-compiler options to the scipy build process. I could figure this out for the fortran compiler but for the c compiler i did not.

2 1

tensordot and axes argument
by lorenzo bolla 29 Jan '08

29 Jan '08

Shouldn't the "axes" argument in tensordot be named "axis"? L. -- Lorenzo Bolla lbolla(a)gmail.com http://lorenzobolla.emurse.com/

2 1

scipy build error - ./numpy/distutils/ccompiler.py bug ? - Was (Numpy-discussion Digest, Vol 16, Issue 59 )
by Hans-Joachim Ehlers 29 Jan '08

29 Jan '08

Given: AIX 5.3 xlc v8 xlf v10 essl v4.2 Build and installed: numpy 1.0.4 Build and installed: lapack 3.1.1 with CCI and essl support Task: Trying to build scipy 0.6.0 Error: Get the following error which seems to be numpy related: ... ... File "/usr/local/lib/python2.5/site-packages/numpy/distutils/ccompiler.py", line 303, in CCompiler_cxx_compiler + cxx.linker_so[2:] TypeError: can only concatenate list (not "str") to list ... Current Solution: Changed line 303 in /usr/local/lib/python2.5/site-packages/numpy/distutils/ccompiler.py from: ... cxx.linker_so = [cxx.linker_so[0]] + cxx.compiler_cxx[0] \ + cxx.linker_so[2:] ... to ... cxx.linker_so = [cxx.linker_so[0]] + [cxx.compiler_cxx[0]] \ + cxx.linker_so[2:] ... Is that change OK for the logic ? Must be a bug report opened for that ? BTW: Has somebody build scipy for AIX 5.3 with the xlc & xlf compilers ? If yes - Could she/he provide some information regarding the requirements tia Hajo

1 0

Which fortran compilers should numpy/scipy support ?
by David Cournapeau 27 Jan '08

27 Jan '08

Hi, For fortran support in numscons, I would like to know which fortran compiler we support ? For now, I successfully used the following: - g77 on linux. - mingw g77 on windows. - gfortran on linux and mac os X. - intel compiler on linux. - sun compiler on linux and solaris. I suspect most unix compilers to work ok wo explicit support (the annoying ones are really windows ones), but I would prefer having a list of compilers explicitely supported for my unit tests. cheers, David

2 2

Re: [Numpy-discussion] [ANN] numscons 0.3.0 release
by David Cournapeau 26 Jan '08

26 Jan '08

Matthew Brett wrote: > Hi David, > > Thanks again for doing this. I've been playing with the compilation - > should I email you direct with questions, to the lists? I don't think it would be inappropriate to use the current lists for that. > > Basically, I'm trying to understand the library discovery, linking > steps - ATLAS in particular. Don't trust the included doc: it is not up-to-date, and that's the part which I totally redesigned since I wrote the initial doc. For things like FFT, BLAS, LAPACK, things are a bit complicated. Fortunately, for "normal" libraries, things are much easier: you use the NumpyCheckLibraryAndHeader function http://projects.scipy.org/scipy/numpy/wiki/NumpySconsExtExamples#Checkingfo… This function can check for a library, headers, and check for symbol in the library. site.cfg customization is also automatically taken care of. I also intend to add the possibility to check for libraries which use pkgconfig. For BLAS/LAPACK/FFT, the basic principles are: - I make a distinction between "API libraries" (called meta libraries in numscons) and "implementation libraries" (called perflib in numscons). - Meta libraries uses perflibs: FFT/BLAS/LAPACK may use Sunperf/MKL/ATLAS. - For each meta library and each perflib, a Check function is defined: CheckMKL, CheckCBLAS, etc... - For a perflib check to succeed, I try to compile, link and run code snippets. Any failure means the Check* function will return failure. - When a perflib Check* succeeds, the meta library check checks that another code snippet can be compiled/linked/run. For example, if CheckMKL succeeds in CheckCBLAS, I will compile/link/run a cblas program using MKL configuration. - Both perflib and meta libraries are customizable through site.cfg (I tried to keep backward compatibility at this level with numpy.distutils). I still have the feeling that it is over-designed, but BLAS/LAPACK libraries are really a PITA. Every perflib has slightly different conventions, and I wanted to use common code for all of them as much as possible [1]. Also, the configuration is not inside python code anymore: libraries, frameworks, link flags are all defined in perflib.cfg. cheers, David [1] This design also allow some funky stuff: for example, when using sun performance libraries with sun tools, because of something which look like a bug to me, the sunperf are not linked when building a shared library. So I link a code snippet, and parse the linker output to get the necessary libraries; this is totally transparent to BLAS/LAPACK checkers.

3 8

tensordot bug?
by Charles R Harris 26 Jan '08

26 Jan '08

Is this a bug? In [21]: a = ones((2,2)) In [22]: b = ones((2,2,2,2)) In [23]: tensordot(a, tensordot(b, a, 2), 2) Out[23]: array(16.0) It seems to me that consistency with the dot product would require a scalar result, not a 0-dim array. Also, do we have a plain old tensor product? Outer flattens the indices. Chuck

1 0

NPY_FORCECAST
by Bill Spotz 26 Jan '08

26 Jan '08

Hi, I am currently using PyArray_FromObject() to convert an input argument to a 32-bit integer array. On a 64-bit architecture, it is easy to create an integer array whose default type is 64-bit. When this is sent to PyArray_FromObject(), it raises an error, saying "array cannot be safely cast to required type", even though all of its elements are representable by 32 bits. I see from the "Guide to NumPy" that I could call PyArray_FromAny() instead, providing the flag NPY_FORCECAST to force the cast to be made. However, this code is within numpy.i, will be used by others, and could lead to unexpected behavior if I were to make the change. How hard would it be to implement a new option (NPY_ATTEMPTCAST?) that attempts to make the cast, but raises an OverflowError if any of the source data is too large for the target array? (Or, alternatively, make this the default behavior if NPY_FORCECAST is false?) Thanks ** Bill Spotz ** ** Sandia National Laboratories Voice: (505)845-0170 ** ** P.O. Box 5800 Fax: (505)284-0154 ** ** Albuquerque, NM 87185-0370 Email: wfspotz(a)sandia.gov **

2 1

extending a recarray
by Kevin Christman 26 Jan '08

26 Jan '08

I am using recarray to store experiment information that is read from a file: self.records = numpy.rec.fromrecords(self.allData,formats='f4,S30,S30,f4,f4,f4,f4,f4,f4,f4,f4,f4,f4',names='MotorNum,MotorType,MotorName,MotorHP, NSync,NShaftFL,NShaft,pLoad,pLoadSlipMethod, pLoadORMEL,etaMeasured,etaORMEL,etaNameplate') Now I would like to add additional "columns" to this array for my calculations/analysis of the data. How is this done? Thanks, Kevin ____________________________________________________________________________________ Looking for last minute shopping deals? Find them fast with Yahoo! Search. http://tools.search.yahoo.com/newsearch/category.php?category=shopping

2 1

Doc-days are today! Come join on irc.freenode.net (#scipy)
by Travis E. Oliphant 26 Jan '08

26 Jan '08

1 0

[ANN] numscons 0.3.0 release
by David Cournapeau 25 Jan '08

25 Jan '08

Hi, I've just released the 0.3.0 release of numscons, an alternative build system for numpy. The tarballs are available on launchpad. https://launchpad.net/numpy.scons.support/0.3/0.3.0 To use it, you need to get the build_with_scons numpy branch: see http://projects.scipy.org/scipy/numpy/wiki/NumpyScons for more details. This release is an important milestone: - all regressions because of the split from numpy are fixed. - it can build numpy on linux (gcc/intel), mac os X (gcc), windows (mingw) and solaris (Sun compilers). - mkl, sunperf, accelerate/vecLib frameworks and ATLAS should work on the platforms where it makes sense. - a lot of internal changes: some basic unittest, a total revamp of the code to check for performance libraries. - almost all changes necessary to scons code are now included upstream, or pending review. If you test it and has problems building numpy, please submit a bug to launchpad: https://bugs.launchpad.net/numpy.scons.support/0.3/ Thanks, David

6 8