numpy 1.1.rc2: win32 binaries

Hi, After some delay, here are the win32 binaries for numpy 1.1.1rc2: http://www.ar.media.kyoto-u.ac.jp/members/david/numpy-1.1.1.dev5559-win32-su... Notes on those binaries: - Based on Atlas 3.8.2 (the 1.1.0 was built against 3.8.0, which had a serious bug wrt dgemm, which is used for numpy.dot. It should solve #844 (problem with numpy.inner) - It is not stricly based on 1.1.rc2, but on 1.1.x trunk. The only difference, though, is a small fix in MANIFEST.in which was broken wrt sdist target and the version of course. If those work out, I will also prepare 2.4 binaries. I am sorry for the delay, cheers, David

David Cournapeau wrote:
Hi,
After some delay, here are the win32 binaries for numpy 1.1.1rc2:
http://www.ar.media.kyoto-u.ac.jp/members/david/numpy-1.1.1.dev5559-win32-su...
I managed to screw up the link: http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-1.1.1.dev5559... cheers, David

On Sun, Jul 27, 2008 at 10:30 PM, David Cournapeau <david@ar.media.kyoto-u.ac.jp> wrote:
http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-1.1.1.dev5559...
I want to get the final 1.1.1 release out ASAP, but I need some feedback on the windows binaries. Could someone please try them out and let us know if you run into any problems. In particular, could someone verify that this has been fixed: http://projects.scipy.org/scipy/numpy/ticket/844 Thanks, -- Jarrod Millman Computational Infrastructure for Research Labs 10 Giannini Hall, UC Berkeley phone: 510.643.4014 http://cirl.berkeley.edu/

Hi, The installation worked on my old Athlon XP running Windows XP and 'numpy.test(level=1)' gave no errors. I did not get an error for the code provided for ticket 844 so I presume this ticket is fixed: 'numpy.inner(F,F)' results in 'array([[ 0.]])' Also, the installer gives this information: Author: Travis E. Oliphant, et.al. Author_email: oliphant@ee.byu.edu Description: NumPy: array processing for numbers, strings, records, and objects. Maintainer: NumPy Developers Maintainer_email: numpy-discussion@lists.sourceforge.net Name: numpy Url: http://numpy.scipy.org Version: 1.1.1.dev5559 I think that at least the ' Author_email' and 'Maintainer_email' should be updated. Thanks Bruce On Thu, Jul 31, 2008 at 4:08 PM, Jarrod Millman <millman@berkeley.edu> wrote:
On Sun, Jul 27, 2008 at 10:30 PM, David Cournapeau <david@ar.media.kyoto-u.ac.jp> wrote:
http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-1.1.1.dev5559...
I want to get the final 1.1.1 release out ASAP, but I need some feedback on the windows binaries. Could someone please try them out and let us know if you run into any problems.
In particular, could someone verify that this has been fixed: http://projects.scipy.org/scipy/numpy/ticket/844
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 Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion

Bruce Southey wrote:
Hi, The installation worked on my old Athlon XP running Windows XP and 'numpy.test(level=1)' gave no errors.
By old, do you mean it does not have at least SSE2 ? If so, the problem would not have happened anyway because the numpy installer does not use ATLAS there, Thanks for the report, David

David Cournapeau wrote:
Bruce Southey wrote:
Hi, The installation worked on my old Athlon XP running Windows XP and 'numpy.test(level=1)' gave no errors.
By old, do you mean it does not have at least SSE2 ? If so, the problem would not have happened anyway because the numpy installer does not use ATLAS there,
Thanks for the report,
David _______________________________________________ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Okay, I was not aware that the ticket was SSE2 as the system is only SSE. I can get access to a system that does. Can you please me the link again because I am getting a 404 error for this: http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-1.1.1.dev5559... Thanks Bruce

http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-1.1.1.dev5559...
Still dead. Cheers, Alan Isaac

Alan G Isaac wrote:
http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-1.1.1.dev5559...
Yes, sorry, it should read: http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy/numpy-1.1.1.d... cheers, David

On Fri, Aug 1, 2008 at 10:03 PM, Alan G Isaac <aisaac@american.edu> wrote:
http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-1.1.1.dev5559...
You can get the official binaries from here: http://sourceforge.net/project/showfiles.php?group_id=1369&package_id=175103&release_id=617279 Please let us know if they work for you or not. Thanks, -- Jarrod Millman Computational Infrastructure for Research Labs 10 Giannini Hall, UC Berkeley phone: 510.643.4014 http://cirl.berkeley.edu/

Hi, I installed the 'official binaries' on a Intel Celeron M 530 that supports SSE2 and SSE3 running MS Vista. All tests passed and with regards to the ticket: numpy.inner(F,F) resulted in 'array([[ Inf]])' Regards Bruce On Sat, Aug 2, 2008 at 7:36 PM, Jarrod Millman <millman@berkeley.edu> wrote:
On Fri, Aug 1, 2008 at 10:03 PM, Alan G Isaac <aisaac@american.edu> wrote:
http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-1.1.1.dev5559...
You can get the official binaries from here: http://sourceforge.net/project/showfiles.php?group_id=1369&package_id=175103&release_id=617279
Please let us know if they work for you or not.
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 Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion

Bruce Southey wrote:
Hi, I installed the 'official binaries' on a Intel Celeron M 530 that supports SSE2 and SSE3 running MS Vista. All tests passed and with regards to the ticket: numpy.inner(F,F) resulted in 'array([[ Inf]])'
I thought that problem had be gone for good, but you're right, I could reproduce on one machine. I have added a regression test on 1.1.x and trunk to avoid letting this slip one more time, and I am now investigating the problem itself, thanks for the bug report, David

Bruce Southey wrote:
Hi, I installed the 'official binaries' on a Intel Celeron M 530 that supports SSE2 and SSE3 running MS Vista. All tests passed and with regards to the ticket: numpy.inner(F,F) resulted in 'array([[ Inf]])'
Ok, it looks like this is entirely due to my own stupidity. I screwed up, and linked the binaries to an earlier version of ATLAS, the one with the problem... I will try to generate new binaries within today, cheers, David

On Mon, Aug 4, 2008 at 7:07 AM, David Cournapeau <david@ar.media.kyoto-u.ac.jp> wrote:
Bruce Southey wrote:
Hi, I installed the 'official binaries' on a Intel Celeron M 530 that supports SSE2 and SSE3 running MS Vista. All tests passed and with regards to the ticket: numpy.inner(F,F) resulted in 'array([[ Inf]])'
Ok, it looks like this is entirely due to my own stupidity. I screwed up, and linked the binaries to an earlier version of ATLAS, the one with the problem...
I will try to generate new binaries within today,
The binaries on sourceforge are now the new ones, with the correct atlas linked in. If there are still problems, don't hesitate to bring it on the ML. cheers, David
cheers,
David _______________________________________________ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion

On Sun, Jul 27, 2008 at 11:26 PM, David Cournapeau < david@ar.media.kyoto-u.ac.jp> wrote:
Hi,
After some delay, here are the win32 binaries for numpy 1.1.1rc2:
http://www.ar.media.kyoto-u.ac.jp/members/david/numpy-1.1.1.dev5559-win32-su...
Notes on those binaries: - Based on Atlas 3.8.2 (the 1.1.0 was built against 3.8.0, which had a serious bug wrt dgemm, which is used for numpy.dot. It should solve #844 (problem with numpy.inner) - It is not stricly based on 1.1.rc2, but on 1.1.x trunk. The only difference, though, is a small fix in MANIFEST.in which was broken wrt sdist target and the version of course.
If those work out, I will also prepare 2.4 binaries. I am sorry for the delay,
Great. Now I just need to write up release notes ;) ...Chuck

On Mon, 28 Jul 2008 14:26:59 +0900 David Cournapeau <david@ar.media.kyoto-u.ac.jp> wrote:
Hi,
After some delay, here are the win32 binaries for numpy 1.1.1rc2:
http://www.ar.media.kyoto-u.ac.jp/members/david/numpy-1.1.1.dev5559-win32-su...
Notes on those binaries: - Based on Atlas 3.8.2 (the 1.1.0 was built against 3.8.0, which had a serious bug wrt dgemm, which is used for numpy.dot. It should solve #844 (problem with numpy.inner) - It is not stricly based on 1.1.rc2, but on 1.1.x trunk. The only difference, though, is a small fix in MANIFEST.in which was broken wrt sdist target and the version of course.
If those work out, I will also prepare 2.4 binaries. I am sorry for the delay,
cheers,
David _______________________________________________ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
David, Did you also try ATLAS3.9.1 ? Is it recommended to use the stable version (3.8.2) Nils

On Mon, Jul 28, 2008 at 11:08 AM, Nils Wagner <nwagner@iam.uni-stuttgart.de>wrote:
On Mon, 28 Jul 2008 14:26:59 +0900 David Cournapeau <david@ar.media.kyoto-u.ac.jp> wrote:
Hi,
After some delay, here are the win32 binaries for numpy 1.1.1rc2:
http://www.ar.media.kyoto-u.ac.jp/members/david/numpy-1.1.1.dev5559-win32-su...
Notes on those binaries: - Based on Atlas 3.8.2 (the 1.1.0 was built against 3.8.0, which had a serious bug wrt dgemm, which is used for numpy.dot. It should solve #844 (problem with numpy.inner) - It is not stricly based on 1.1.rc2, but on 1.1.x trunk. The only difference, though, is a small fix in MANIFEST.in which was broken wrt sdist target and the version of course.
If those work out, I will also prepare 2.4 binaries. I am sorry for the delay,
cheers,
David _______________________________________________ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
David,
Did you also try ATLAS3.9.1 ? Is it recommended to use the stable version (3.8.2)
I compiled ATLAS3.9.1 and the make time tests didn't run any faster. It's also very new, so it might be best to let it stew a bit to uncover any little oopsies. Chuck

Nils Wagner wrote:
David,
Did you also try ATLAS3.9.1 ? Is it recommended to use the stable version (3.8.2)
I did not try atlas 3.9.1, but anyone can try. I personally do not want to package unstable versions, but the vendor directory in numpy repository should make it relatively easy to try for yourself (once you add atlas 3.9.1 sources). cheers, David
participants (7)
-
Alan G Isaac
-
Bruce Southey
-
Charles R Harris
-
David Cournapeau
-
David Cournapeau
-
Jarrod Millman
-
Nils Wagner