Win32 installer: please test it

classic Classic list List threaded Threaded
21 messages Options
12
Reply | Threaded
Open this post in threaded view
|

Win32 installer: please test it

Jarrod Millman
Hello,

David Cournapeau has prepared a new win32 installer, which is aimed at
solving the recurring problem of non working atlas on different sets
of CPU. This installer simply checks which cpu you have, and installs
the appropriate numpy accordingly (without atlas if your cpu is not
supported). Windows users, please test the installer, and report
problems with it; we hope to use this scheme for all numpy and scipy
installers in the future.

Download from here:
https://cirl.berkeley.edu/numpy/numpy-superpack-python2.5.exe

Technical details:

 - ATLAS is 3.8.0, and built with cygwin. Built on Core 2 duo for
SSE3, pentium 4 xeon for SSE2.
 - BLAS/LAPACK are built with mingw.
 - only SSE3 and SSE2 are supported. If you don't have at least sse2,
it will not use ATLAS, but netlib blas/lapack instead.
 - the numpy installers are prepared with build_wininst for each
different supported architecture
 - the numpy installers are then bundled in another installer, based on nsis.
 - Nsis is an open source installer, based on a small scripting
language, and is extensible by plug-ins. I wrote a plug-in to detect
the cpu type, and the installer will simply execute a different numpy
installer depending on the cpu.

Thanks,

--
Jarrod Millman
Computational Infrastructure for Research Labs
10 Giannini Hall, UC Berkeley
phone: 510.643.4014
http://cirl.berkeley.edu/
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

cdavid
Jarrod Millman wrote:
> Hello,
>
> David Cournapeau has prepared a new win32 installer, which is aimed at
> solving the recurring problem of non working atlas on different sets
> of CPU. This installer simply checks which cpu you have, and installs
> the appropriate numpy accordingly (without atlas if your cpu is not
> supported). Windows users, please test the installer, and report
> problems with it; we hope to use this scheme for all numpy and scipy
> installers in the future.

Sorry for being insistent, but I would really like to get rid of those
windows/sse* problems once for all for numpy/scipy, and this means
having people to test the installer.

cheers,

David
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Bill Baxter-2
On Mon, Apr 14, 2008 at 10:07 AM, David Cournapeau
<[hidden email]> wrote:

> Jarrod Millman wrote:
>  > Hello,
>  >
>  > David Cournapeau has prepared a new win32 installer, which is aimed at
>  > solving the recurring problem of non working atlas on different sets
>  > of CPU. This installer simply checks which cpu you have, and installs
>  > the appropriate numpy accordingly (without atlas if your cpu is not
>  > supported). Windows users, please test the installer, and report
>  > problems with it; we hope to use this scheme for all numpy and scipy
>  > installers in the future.
>
>  Sorry for being insistent, but I would really like to get rid of those
>  windows/sse* problems once for all for numpy/scipy, and this means
>  having people to test the installer.

Maybe your insistence might be more effective if it included the link
part of the message. :-)

"""
Download from here:
https://cirl.berkeley.edu/numpy/numpy-superpack-python2.5.exe
"""

I'll give it a try.  I've been meaning to upgrade my numpy for a while now.

--bb
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Damian Eads-2
In reply to this post by cdavid
David Cournapeau wrote:

> Jarrod Millman wrote:
>> Hello,
>>
>> David Cournapeau has prepared a new win32 installer, which is aimed at
>> solving the recurring problem of non working atlas on different sets
>> of CPU. This installer simply checks which cpu you have, and installs
>> the appropriate numpy accordingly (without atlas if your cpu is not
>> supported). Windows users, please test the installer, and report
>> problems with it; we hope to use this scheme for all numpy and scipy
>> installers in the future.
>
> Sorry for being insistent, but I would really like to get rid of those
> windows/sse* problems once for all for numpy/scipy, and this means
> having people to test the installer.
>
> cheers,
>
> David

Hi David,

I'd be happy to help you test the installer (and let you know what my
processor supports). Is it just the installer you want tested or do you
also want us to try some computation with numpy?

Nice work btw!

Cheers,

Damian

_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

cdavid
Damian Eads wrote:
> Hi David,
>
> I'd be happy to help you test the installer (and let you know what my
> processor supports). Is it just the installer you want tested or do you
> also want us to try some computation with numpy?
>  

Doing numpy.test() should be fine. The actual numpy installers are built
the exact same way than before, so if no crash happens, it should mean
it is fine, hopefully.

cheers,

David
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Bill Baxter-2
In reply to this post by Bill Baxter-2
What's the installer supposed to do?
It appears to have decided I have SSE3 extracted an sse3 exe to
somewhere, and then quit without doing anything else.  I can't seem to
find where it put the numpy-1.0.5-sse3.exe it extracted, either.

The details, say "Execute: numpy-1.0.5-sse3.exe" but no executing
seems to have happened.

--bb

On Mon, Apr 14, 2008 at 10:37 AM, Bill Baxter <[hidden email]> wrote:

> On Mon, Apr 14, 2008 at 10:07 AM, David Cournapeau
>  <[hidden email]> wrote:
>  > Jarrod Millman wrote:
>  >  > Hello,
>  >  >
>  >  > David Cournapeau has prepared a new win32 installer, which is aimed at
>  >  > solving the recurring problem of non working atlas on different sets
>  >  > of CPU. This installer simply checks which cpu you have, and installs
>  >  > the appropriate numpy accordingly (without atlas if your cpu is not
>  >  > supported). Windows users, please test the installer, and report
>  >  > problems with it; we hope to use this scheme for all numpy and scipy
>  >  > installers in the future.
>  >
>  >  Sorry for being insistent, but I would really like to get rid of those
>  >  windows/sse* problems once for all for numpy/scipy, and this means
>  >  having people to test the installer.
>
>  Maybe your insistence might be more effective if it included the link
>  part of the message. :-)
>
>  """
>  Download from here:
>  https://cirl.berkeley.edu/numpy/numpy-superpack-python2.5.exe
>  """
>
>  I'll give it a try.  I've been meaning to upgrade my numpy for a while now.
>
>  --bb
>
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

cdavid
Bill Baxter wrote:
> What's the installer supposed to do?

It is supposed to install numpy :) The fact that it is not clear is not
good, obviously. Suggestions to make it more obvious are welcomed.

> It appears to have decided I have SSE3 extracted an sse3 exe to
> somewhere, and then quit without doing anything else.  I can't seem to
> find where it put the numpy-1.0.5-sse3.exe it extracted, either.

You did not have the usual serie of screens for numpy installation
(python version detection, etc...) ? If not, this means the installer
does not work at all.

If working, you are not supposed to find the numpy-1.0.5-sse3.exe, the
whole idea of the installer is to choose the right numpy installer, and
run it. This is meant to simplify windows' users life, the ones who do
not know nor care about what SSE or ATLAS is and whether their computer
support it.

> The details, say "Execute: numpy-1.0.5-sse3.exe" but no executing
> seems to have happened.

No executing and no error ? That's the worse thing which could happen,
and the exact thing I was fearing (this and vista support).

thanks for testing this out,

cheers,

David
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

cdavid
In reply to this post by Bill Baxter-2
Bill Baxter wrote:
> What's the installer supposed to do?
> It appears to have decided I have SSE3 extracted an sse3 exe to
> somewhere, and then quit without doing anything else.  I can't seem to
> find where it put the numpy-1.0.5-sse3.exe it extracted, either.
>
> The details, say "Execute: numpy-1.0.5-sse3.exe" but no executing
> seems to have happened.
>  

Ok, I've just tested it on my vmware installed windows, and got the
exact same behaviour. Maybe I screwed up and uploaded the wrong file, I
am looking at it now.

cheers,

David
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Bill Baxter-2
In reply to this post by cdavid
On Mon, Apr 14, 2008 at 10:57 AM, David Cournapeau
<[hidden email]> wrote:

> Bill Baxter wrote:
>  > What's the installer supposed to do?
>
>  It is supposed to install numpy :) The fact that it is not clear is not
>  good, obviously. Suggestions to make it more obvious are welcomed.
>
>
>  > It appears to have decided I have SSE3 extracted an sse3 exe to
>  > somewhere, and then quit without doing anything else.  I can't seem to
>  > find where it put the numpy-1.0.5-sse3.exe it extracted, either.
>
>  You did not have the usual serie of screens for numpy installation
>  (python version detection, etc...) ? If not, this means the installer
>  does not work at all.
>
>  If working, you are not supposed to find the numpy-1.0.5-sse3.exe, the
>  whole idea of the installer is to choose the right numpy installer, and
>  run it. This is meant to simplify windows' users life, the ones who do
>  not know nor care about what SSE or ATLAS is and whether their computer
>  support it.
>
>
>  > The details, say "Execute: numpy-1.0.5-sse3.exe" but no executing
>  > seems to have happened.
>
>  No executing and no error ? That's the worse thing which could happen,
>  and the exact thing I was fearing (this and vista support).

That's right.  No execution and no error.  The installer finishes
(quite quickly!) as if it did everything according to plan.

Anything I can do to help you track down the problem?  It's a Win XP system.

Some of my environment vars that might be relevant:
PROCESSOR_ARCHITECTURE=x86
PROCESSOR_IDENTIFIER=x86 Family 15 Model 4 Stepping 1, GenuineIntel
PROCESSOR_LEVEL=15
PROCESSOR_REVISION=0401
NUMBER_OF_PROCESSORS=2
HOME=c:\bax
HOMEDRIVE=C:
HOMEPATH=\Documents and Settings\baxter
PYTHONPATH=c:\bax\code\python;f:\bax\code\python
PYTHONSTARTUP=f:\bax\code\python\mystartup.py


The default encoding of the machine is set to Japanese (CP931), but
the langauge is English.

That's about all I can think of that might be unusual.

--bb
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

cdavid
Bill Baxter wrote:
>
> That's right.  No execution and no error.  The installer finishes
> (quite quickly!)

Doing nothing is quick, even on windows :)

>  as if it did everything according to plan.
>
> Anything I can do to help you track down the problem?  It's a Win XP system.

I think I got the problem. Now, I got it working on my two vmware
windows at hand. Could you try this:

http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-superpack-python2.5.exe

The problem was that I thought I did set the install directory to a
windows temp directory, but I actually only set up its default value,
and as such the actual numpy installer was put in C:\ (which is a really
crappy default ...). Because C:\ is not in the PATH, it does not work.

I set up the directory correctly, and there should be a message if the
execution of the actual numpy installer fails (there is also more
details in the installation log). I still do not understand why it
worked on one of my windows and not the other, though.

> The default encoding of the machine is set to Japanese (CP931), but
> the langauge is English.

Well, I happend to have the same thing: my university's laptop windows
is in Japanese, and it works on this one too.

cheers,

David
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Bill Baxter-2
Seems to work here now, too!

It doesn't tell you in an easy to see place what version of SSE it
decides to use.  Do you think that's ok?  (You can tell by looking at
the "details" at then end of installation, though.

Is there some way to tell this info from inside NumPy itself?  If so
then maybe it doesn't matter.

--bb


--------------------
Numpy is installed in c:\python25\lib\site-packages\numpy
Numpy version 1.0.5.dev5008
Python version 2.5.1 (r251:54863, Apr 18 2007, 08:51:08) [MSC v.1310
32 bit (Intel)]
  Found 10/10 tests for numpy.core.defmatrix
  Found 3/3 tests for numpy.core.memmap
  Found 266/266 tests for numpy.core.multiarray
  Found 69/69 tests for numpy.core.numeric
  Found 31/31 tests for numpy.core.numerictypes
  Found 12/12 tests for numpy.core.records
  Found 7/7 tests for numpy.core.scalarmath
  Found 16/16 tests for numpy.core.umath
  Found 5/5 tests for numpy.ctypeslib
  Found 5/5 tests for numpy.distutils.misc_util
  Found 2/2 tests for numpy.fft.fftpack
  Found 3/3 tests for numpy.fft.helper
  Found 20/20 tests for numpy.lib._datasource
  Found 10/10 tests for numpy.lib.arraysetops
  Found 1/1 tests for numpy.lib.financial
  Found 0/0 tests for numpy.lib.format
  Found 48/48 tests for numpy.lib.function_base
  Found 5/5 tests for numpy.lib.getlimits
  Found 4/4 tests for numpy.lib.index_tricks
  Found 7/7 tests for numpy.lib.io
  Found 4/4 tests for numpy.lib.polynomial
  Found 49/49 tests for numpy.lib.shape_base
  Found 15/15 tests for numpy.lib.twodim_base
  Found 43/43 tests for numpy.lib.type_check
  Found 1/1 tests for numpy.lib.ufunclike
  Found 59/59 tests for numpy.linalg
  Found 92/92 tests for numpy.ma.core
  Found 14/14 tests for numpy.ma.extras
  Found 7/7 tests for numpy.random
  Found 0/0 tests for numpy.testing.utils
  Found 0/0 tests for __main__
.......................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
----------------------------------------------------------------------
Ran 887 tests in 1.750s

OK

On Mon, Apr 14, 2008 at 12:24 PM, David Cournapeau
<[hidden email]> wrote:

> Bill Baxter wrote:
>  >
>  > That's right.  No execution and no error.  The installer finishes
>  > (quite quickly!)
>
>  Doing nothing is quick, even on windows :)
>
>
>  >  as if it did everything according to plan.
>  >
>  > Anything I can do to help you track down the problem?  It's a Win XP system.
>
>  I think I got the problem. Now, I got it working on my two vmware
>  windows at hand. Could you try this:
>
>  http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-superpack-python2.5.exe
>
>  The problem was that I thought I did set the install directory to a
>  windows temp directory, but I actually only set up its default value,
>  and as such the actual numpy installer was put in C:\ (which is a really
>  crappy default ...). Because C:\ is not in the PATH, it does not work.
>
>  I set up the directory correctly, and there should be a message if the
>  execution of the actual numpy installer fails (there is also more
>  details in the installation log). I still do not understand why it
>  worked on one of my windows and not the other, though.
>
>
>  > The default encoding of the machine is set to Japanese (CP931), but
>  > the langauge is English.
>
>  Well, I happend to have the same thing: my university's laptop windows
>  is in Japanese, and it works on this one too.
>
>
>
>  cheers,
>
>  David
>  _______________________________________________
>  Numpy-discussion mailing list
>  [hidden email]
>  http://projects.scipy.org/mailman/listinfo/numpy-discussion
>
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

cdavid
Bill Baxter wrote:
> Seems to work here now, too!
>
> It doesn't tell you in an easy to see place what version of SSE it
> decides to use.  Do you think that's ok?

I simply did not think it was useful information. How would you like to
get the information ? To be honest, I do not want to spend too much time
on this (I do not use windows at all): I was just really fed up with all
the messages about numpy crashes on windows, and wanted a quick, working
solution.

I intend to put everything necessary to create those installers in some
place on the scipy.org website of somewhere else in a near future, so
that other people can contribute and enhance (the big part is actually
another installer, which I annouced a few weeks ago, and which contain
blas/lapack and ATLAS for both mingw and vs-based builds, as well as a
bunch of scripts to regenerate all of them automatically).

cheers,

David
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Bill Baxter-2
On Mon, Apr 14, 2008 at 1:08 PM, David Cournapeau
<[hidden email]> wrote:
> Bill Baxter wrote:
>  > Seems to work here now, too!
>  >
>  > It doesn't tell you in an easy to see place what version of SSE it
>  > decides to use.  Do you think that's ok?
>
>  I simply did not think it was useful information. How would you like to
>  get the information ?

Well, reason for wanting it were two-fold
1) I think my computer has SSE3, so I would be surprised if it came up
saying it was going to install the SSE2 version.
2) If there are still reports about Numpy crashing even using this,
then the first thing you will want to ask people is "which version did
the installer install for you".  They're more likely to know the
answer if the installer tells them.

I could imagine a couple of ways to show the information:
1) after determining the version to use pop up a message box saying
something like "This will install the SSE3 enabled version of Numpy
1.0.5.   OK?"
2) Put the version used in some text on the final panel (the panel
with the "Details button").   There's a big blank space there now.  It
could say "successfully installed Numpy version Blah.Blah-SSE3"

> To be honest, I do not want to spend too much time
>  on this (I do not use windows at all): I was just really fed up with all
>  the messages about numpy crashes on windows, and wanted a quick, working
>  solution.

Ok.  Didn't realize that.

>  I intend to put everything necessary to create those installers in some
>  place on the scipy.org website of somewhere else in a near future, so
>  that other people can contribute and enhance (the big part is actually
>  another installer, which I annouced a few weeks ago, and which contain
>  blas/lapack and ATLAS for both mingw and vs-based builds, as well as a
>  bunch of scripts to regenerate all of them automatically).

That sounds good.  From there it probably wouldn't take someone else
very long to make the suggested change.  I've used NSIS before, so I
could probably figure out how.

--bb
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Gary Ruben-2
In reply to this post by Jarrod Millman
Tested fine on my old Classic Athlon 500 (no SSE) under Win98. It
correctly reported installing the non-SSE version when I clicked on the
details button on the last page of the install wizard. Whereas
previously numpy.test() would bring up an illegal operation dialog box,
now all tests pass. Nice!. Gary R.

 >>> N.test()
Numpy is installed in C:\PYTHON25\lib\site-packages\numpy
Numpy version 1.0.5.dev5008
Python version 2.5.2a0 (release25-maint:59715M, Jan  3 2008, 22:45:43)
[MSC v.13
10 32 bit (Intel)]
   Found 10/10 tests for numpy.core.defmatrix
   Found 3/3 tests for numpy.core.memmap
   Found 266/266 tests for numpy.core.multiarray
   Found 69/69 tests for numpy.core.numeric
   Found 31/31 tests for numpy.core.numerictypes
   Found 12/12 tests for numpy.core.records
   Found 7/7 tests for numpy.core.scalarmath
   Found 16/16 tests for numpy.core.umath
   Found 5/5 tests for numpy.ctypeslib
   Found 5/5 tests for numpy.distutils.misc_util
   Found 2/2 tests for numpy.fft.fftpack
   Found 3/3 tests for numpy.fft.helper
   Found 20/20 tests for numpy.lib._datasource
   Found 10/10 tests for numpy.lib.arraysetops
   Found 1/1 tests for numpy.lib.financial
   Found 0/0 tests for numpy.lib.format
   Found 48/48 tests for numpy.lib.function_base
   Found 5/5 tests for numpy.lib.getlimits
   Found 4/4 tests for numpy.lib.index_tricks
   Found 7/7 tests for numpy.lib.io
   Found 4/4 tests for numpy.lib.polynomial
   Found 49/49 tests for numpy.lib.shape_base
   Found 15/15 tests for numpy.lib.twodim_base
   Found 43/43 tests for numpy.lib.type_check
   Found 1/1 tests for numpy.lib.ufunclike
   Found 59/59 tests for numpy.linalg
   Found 92/92 tests for numpy.ma.core
   Found 14/14 tests for numpy.ma.extras
   Found 7/7 tests for numpy.random
   Found 0/0 tests for numpy.testing.utils
   Found 0/0 tests for __main__
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
.......
----------------------------------------------------------------------
Ran 887 tests in 7.580s

OK
<unittest._TextTestResult run=887 errors=0 failures=0>


Jarrod Millman wrote:
> Hello,
>
> David Cournapeau has prepared a new win32 installer, which is aimed at
> solving the recurring problem of non working atlas on different sets
> of CPU. This installer simply checks which cpu you have, and installs
> the appropriate numpy accordingly (without atlas if your cpu is not
> supported). Windows users, please test the installer, and report

_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

josef.pktd
Everything installed without problem on Intel Pentium M on my notebook
recognized as SSE2 capable.

Installer found Python 2.5. immediately, which I just installed and
all my windows environment settings are still setup for python 2.4

Thanks,
Josef

>>> numpy.test()
Numpy is installed in C:\Programs\Python25\lib\site-packages\numpy
Numpy version 1.0.5.dev5008
Python version 2.5.2 (r252:60911, Feb 21 2008, 13:11:45) [MSC v.1310
32 bit (Intel)]
  Found 10/10 tests for numpy.core.defmatrix
  Found
  <...>
----------------------------------------------------------------------
Ran 887 tests in 5.922s

OK
<unittest._TextTestResult run=887 errors=0 failures=0>





>>> numpy.show_config()
atlas_threads_info:
  NOT AVAILABLE

blas_opt_info:
    libraries = ['f77blas', 'cblas', 'atlas']
    library_dirs = ['C:\\local\\lib\\sse2']
    define_macros = [('NO_ATLAS_INFO', 2)]
    language = c

atlas_blas_threads_info:
  NOT AVAILABLE

lapack_opt_info:
    libraries = ['lapack', 'f77blas', 'cblas', 'atlas']
    library_dirs = ['C:\\local\\lib\\sse2']
    define_macros = [('NO_ATLAS_INFO', 2)]
    language = f77

atlas_info:
    libraries = ['lapack', 'f77blas', 'cblas', 'atlas']
    library_dirs = ['C:\\local\\lib\\sse2']
    language = f77

lapack_mkl_info:
  NOT AVAILABLE

blas_mkl_info:
  NOT AVAILABLE

atlas_blas_info:
    libraries = ['f77blas', 'cblas', 'atlas']
    library_dirs = ['C:\\local\\lib\\sse2']
    language = c

mkl_info:
  NOT AVAILABLE
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Jarrod Millman
In reply to this post by cdavid
On Sun, Apr 13, 2008 at 8:24 PM, David Cournapeau
<[hidden email]> wrote:
>  I think I got the problem. Now, I got it working on my two vmware
>  windows at hand. Could you try this:
>
>  http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-superpack-python2.5.exe

Did you have a problem putting it on the cirl site?

--
Jarrod Millman
Computational Infrastructure for Research Labs
10 Giannini Hall, UC Berkeley
phone: 510.643.4014
http://cirl.berkeley.edu/
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Bruce Southey
In reply to this post by Jarrod Millman
Hi,
The installer requires 2.5 but that is not explicitly stated. Given
that numpy and scipy technically require Python 2.3 as a minimum, the
installer should support Python 2.3 and Python 2.4. Otherwise, it
needs to clearly state that Python 2.5 is a must.

I did install Python 2.5.2 on my WIndows XP AMD Athlon XP 2100+ system
and all the tests passed.

Regards
Bruce

On Thu, Apr 10, 2008 at 12:18 PM, Jarrod Millman <[hidden email]> wrote:

> Hello,
>
>  David Cournapeau has prepared a new win32 installer, which is aimed at
>  solving the recurring problem of non working atlas on different sets
>  of CPU. This installer simply checks which cpu you have, and installs
>  the appropriate numpy accordingly (without atlas if your cpu is not
>  supported). Windows users, please test the installer, and report
>  problems with it; we hope to use this scheme for all numpy and scipy
>  installers in the future.
>
>  Download from here:
>  https://cirl.berkeley.edu/numpy/numpy-superpack-python2.5.exe
>
>  Technical details:
>
>   - ATLAS is 3.8.0, and built with cygwin. Built on Core 2 duo for
>  SSE3, pentium 4 xeon for SSE2.
>   - BLAS/LAPACK are built with mingw.
>   - only SSE3 and SSE2 are supported. If you don't have at least sse2,
>  it will not use ATLAS, but netlib blas/lapack instead.
>   - the numpy installers are prepared with build_wininst for each
>  different supported architecture
>   - the numpy installers are then bundled in another installer, based on nsis.
>   - Nsis is an open source installer, based on a small scripting
>  language, and is extensible by plug-ins. I wrote a plug-in to detect
>  the cpu type, and the installer will simply execute a different numpy
>  installer depending on the cpu.
>
>  Thanks,
>
>  --
>  Jarrod Millman
>  Computational Infrastructure for Research Labs
>  10 Giannini Hall, UC Berkeley
>  phone: 510.643.4014
>  http://cirl.berkeley.edu/
>  _______________________________________________
>  Numpy-discussion mailing list
>  [hidden email]
>  http://projects.scipy.org/mailman/listinfo/numpy-discussion
>
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

cdavid
Bruce Southey wrote:
> Hi,
> The installer requires 2.5 but that is not explicitly stated. Given
> that numpy and scipy technically require Python 2.3 as a minimum, the
> installer should support Python 2.3 and Python 2.4. Otherwise, it
> needs to clearly state that Python 2.5 is a must.
>  

Stricly speaking, when numpy says it supports all python starting at
2.3, this does not mean we have to provide binaries for all versions. It
means that the code itself is compatible.

That being said, the posted installer was just a test version. We will
do just as before, that is provide a separate installer for python 2.3,
2.4 and 2.5.

cheers,

David
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

pec27
In reply to this post by Jarrod Millman
Yes, I am one of those users with crashes in numpy 1.4. Your build
seems to pass for me. For reference my machine is Windows XP, Intel
Xeon 5140

-----------------
Numpy is installed in C:\Python25\lib\site-packages\numpy
Numpy version 1.0.5.dev5008
Python version 2.5.2 (r252:60911, Feb 21 2008, 13:11:45) [MSC v.1310 32 bit (Int
el)]
  Found 10/10 tests for numpy.core.defmatrix
  Found 3/3 tests for numpy.core.memmap
  Found 266/266 tests for numpy.core.multiarray
  Found 69/69 tests for numpy.core.numeric
  Found 31/31 tests for numpy.core.numerictypes
  Found 12/12 tests for numpy.core.records
  Found 7/7 tests for numpy.core.scalarmath
  Found 16/16 tests for numpy.core.umath
  Found 5/5 tests for numpy.ctypeslib
  Found 5/5 tests for numpy.distutils.misc_util
  Found 2/2 tests for numpy.fft.fftpack
  Found 3/3 tests for numpy.fft.helper
  Found 20/20 tests for numpy.lib._datasource
  Found 10/10 tests for numpy.lib.arraysetops
  Found 1/1 tests for numpy.lib.financial
  Found 0/0 tests for numpy.lib.format
  Found 48/48 tests for numpy.lib.function_base
  Found 5/5 tests for numpy.lib.getlimits
  Found 4/4 tests for numpy.lib.index_tricks
  Found 7/7 tests for numpy.lib.io
  Found 4/4 tests for numpy.lib.polynomial
  Found 49/49 tests for numpy.lib.shape_base
  Found 15/15 tests for numpy.lib.twodim_base
  Found 43/43 tests for numpy.lib.type_check
  Found 1/1 tests for numpy.lib.ufunclike
  Found 59/59 tests for numpy.linalg
  Found 92/92 tests for numpy.ma.core
  Found 14/14 tests for numpy.ma.extras
  Found 7/7 tests for numpy.random
  Found 0/0 tests for numpy.testing.utils
  Found 0/0 tests for __main__
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
.......
----------------------------------------------------------------------
Ran 887 tests in 0.890s

OK


On 14/04/2008, Bill Baxter wrote:
>  Seems to work here now, too!
>
>  It doesn't tell you in an easy to see place what version of SSE it
>  decides to use.  Do you think that's ok?  (You can tell by looking at
>  the "details" at then end of installation, though.
>
>  Is there some way to tell this info from inside NumPy itself?  If so
>  then maybe it doesn't matter.
>
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Win32 installer: please test it

Damian Eads-2
In reply to this post by cdavid
David Cournapeau wrote:

> Jarrod Millman wrote:
>> Hello,
>>
>> David Cournapeau has prepared a new win32 installer, which is aimed at
>> solving the recurring problem of non working atlas on different sets
>> of CPU. This installer simply checks which cpu you have, and installs
>> the appropriate numpy accordingly (without atlas if your cpu is not
>> supported). Windows users, please test the installer, and report
>> problems with it; we hope to use this scheme for all numpy and scipy
>> installers in the future.
>
> Sorry for being insistent, but I would really like to get rid of those
> windows/sse* problems once for all for numpy/scipy, and this means
> having people to test the installer.
>
> cheers,
>
> David
> _______________________________________________
> Numpy-discussion mailing list
> [hidden email]
> http://projects.scipy.org/mailman/listinfo/numpy-discussion

Sorry for getting back to you so late. I was out of range of wireless
networks for a day and a half. Shown below are my /proc/cpu-info and
result of running numpy.test(). SSE and SSE2 are supported but SSE3 is not.

Nice work David!

Damian

-------------------------------------------
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 13
model name      : Intel(R) Pentium(R) M processor 1.60GHz
stepping        : 6
cpu MHz         : 1600.000
cache size      : 2048 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 2
wp              : yes
flags           : fpu vme de pse tsc msr mce cx8 mtrr pge mca cmov pat
clflush dts acpi mmx fxsr sse sse2 ss tm pbe up bts est tm2
bogomips        : 3198.55
clflush size    : 64


Numpy version 1.0.4
Python version 2.5.1 (r251:54863, Apr 18 2007, 08:51:08) [MSC v.1310 32
bit (Int
el)]
   Found 10/10 tests for numpy.core.defmatrix
   Found 36/36 tests for numpy.core.ma
   Found 223/223 tests for numpy.core.multiarray
   Found 65/65 tests for numpy.core.numeric
   Found 31/31 tests for numpy.core.numerictypes
   Found 12/12 tests for numpy.core.records
   Found 6/6 tests for numpy.core.scalarmath
   Found 14/14 tests for numpy.core.umath
   Found 4/4 tests for numpy.ctypeslib
   Found 5/5 tests for numpy.distutils.misc_util
   Found 1/1 tests for numpy.fft.fftpack
   Found 3/3 tests for numpy.fft.helper
   Found 9/9 tests for numpy.lib.arraysetops
   Found 46/46 tests for numpy.lib.function_base
   Found 5/5 tests for numpy.lib.getlimits
   Found 4/4 tests for numpy.lib.index_tricks
   Found 3/3 tests for numpy.lib.polynomial
   Found 49/49 tests for numpy.lib.shape_base
   Found 15/15 tests for numpy.lib.twodim_base
   Found 43/43 tests for numpy.lib.type_check
   Found 1/1 tests for numpy.lib.ufunclike
   Found 40/40 tests for numpy.linalg
   Found 2/2 tests for numpy.random
   Found 0/0 tests for __main__
................................................................................
................................................................................
................................................................................
................................................................................
....................................................................Warning:
inv
alid value encountered in absolute
Warning: invalid value encountered in absolute
Warning: invalid value encountered in less_equal
................................................................................
................................................................................
................................................................................
..........................................................
----------------------------------------------------------------------
Ran 686 tests in 1.250s

OK
<unittest._TextTestResult run=686 errors=0 failures=0>
------------------------
_______________________________________________
Numpy-discussion mailing list
[hidden email]
http://projects.scipy.org/mailman/listinfo/numpy-discussion
12