Build Python, Numpy and Scipy source with Visual Studio 6.0 forwindows

[Richard_Martineau]

Hello All

I wonder if anyone can advise me or has done similar to the following?

Basically I've downloaded the Python 2.5.2 source code that builds
with Visual Studio 6.0. I've built Python for windows. This was easy
(it even came with the pcbuild.dsw workspace file). Great!

Now comes the troubled bit...I now look for similar source code for
Python extensions Numpy and Scipy but the source code and directories
are not all obvious. Looks like these are normally built via other
compilers. However I need to do all my builds in VS 6.0.

For Numpy (I haven't got as far as Scipy yet) I've organised the
source code into the following projects and lumped them all into my
new workspace numpy.dsw:

dotblas
fft
lib
linalg
multiarray
numarray
random
scalarmath
sort
umath

I'm obviously not sure if this is correct. As you can see from above,
I've re-arranged the source code directory named core into the
following projects:

dotblas
sort
multiarray
scalarmath
umath


Unbelievably I am having some luck and nearly all the projects compile
and link (untested though).
I created the auto-generated files using the python scripts. Then I
included the headers where necessary.

However the remaining problems are:
multiarray (wont compile)
dotblas (wont link)

dotblas seems to have missing dependencies on these:
_cblas_sdot
_cblas_ddot
_cblas_cdotu_sub
_cblas_zdotu_sub
_cblas_cgemm
_cblas_zgemm
_cblas_sgemm
_cblas_dgemm
_cblas_cgemv
_cblas_zgemv
_cblas_sgemv
_cblas_dgemv
_cblas_caxpy
_cblas_saxpy
_cblas_zaxpy
_cblas_daxpy
_cblas_cdotc_sub
_cblas_zdotc_sub

Well that's a snapshot of where I am this morning. What I'm really
after is some advice from anybody who has done similar. I'd really
appreciate hearing from you.

Thanks for your help and very happy to have joined this group


Richard

 

[Gabriel Genellina]

I wonder if anyone can advise me or has done similar to the following?

Basically I've downloaded the Python 2.5.2 source code that builds
with Visual Studio 6.0. I've built Python for windows. This was easy
(it even came with the pcbuild.dsw workspace file). Great!

Now comes the troubled bit...I now look for similar source code for
Python extensions Numpy and Scipy but the source code and directories
are not all obvious. Looks like these are normally built via other
compilers. However I need to do all my builds in VS 6.0.

For Numpy (I haven't got as far as Scipy yet) I've organised the
source code into the following projects and lumped them all into my
new workspace numpy.dsw:

Don't try to roll your own projects, compile it using distutils instead.
Distutils takes care of defining the right symbols and compiler options,
and should detect VS.
I'm pretty sure the Numpy README file (or similar) describes how to build
it; usually you install the required dependencies and then run:
python setup.py build

 

[Richard_Martineau]

Don't try to roll your own projects, compile it using distutils instead.  
Distutils takes care of defining the right symbols and compiler options,  
and should detect VS.
I'm pretty sure the Numpy README file (or similar) describes how to build  
it; usually you install the required dependencies and then run:
python setup.py build

Thanks for your message.

Unfortunately I need as much control and visibility as possible as I
eventually want to port from windows os to pharlap rtos.

I appreciate your advice and I understand it's not a great idea to
roll your own, but if it's possible, I really need to do so.

Any further help/advice greatly appreciated


Richard

 

[Terry Reedy]

Now comes the troubled bit...I now look for similar source code for
Python extensions Numpy and Scipy but the source code and directories
are not all obvious. Looks like these are normally built via other
compilers. However I need to do all my builds in VS 6.0.

Numpy/scipy specific questions tend to get more specific answers on the
numpy/scipy mailing lists. Or see gmane.comp.python.numeric.general at
news.gmane.org

 

[Richie]

Numpy/scipy specific questions tend to get more specific answers on the
numpy/scipy mailing lists.  Or see gmane.comp.python.numeric.general at
news.gmane.org

Thanks very much for your advice Terry. I've just signed up with Numpy
discussion and Scipy dev. Just taking a look at gmane now!

Richard