Re: [SciPy-user] scipy on old CPU crashes
sorry, but scipy-0.7.0b1-win32-superpack-python2.4.exe and numpy-1.2.1-win32-superpack-python2.4.exe crash on my old pc too, which uses duron 750MHz. So now I think it is not the problem with non-sse/sse/sse2 instruction is there any method to find out the real reason except to compile from the source thanx (I know Duron is too old, but currently I don't have the money to buy a new PC or even update it :( so bad) [code] from scipy.integrate import quad print quad(lambda x:x, 1, 2) #crash soon [/code]
Date: Thu, 13 Nov 2008 12:41:33 +0900 From: "David Cournapeau" <cournape@gmail.com> Subject: Re: [SciPy-user] scipy on old CPU crashes To: "SciPy Users List" <scipy-user@scipy.org> Message-ID: <5b8d13220811121941va8442f2gcb3a997874878b4b@mail.gmail.com> Content-Type: text/plain; charset=UTF-8
On Thu, Nov 13, 2008 at 12:16 PM, oyster <lepto.python@gmail.com> wrote:
hi, all I am using an old AMD Duron CPU with Win2k, which seems does not support SSE/SSE2
indeed, old Duron does not support SSE IIRC.
I found that there are 3 verisons in numpy-1.2.1-win32-superpack-python2.5.exe(numpy-1.2.1-sse3.exe, numpy-1.2.1-sse2.exe and numpy-1.2.1-nosse.exe)
Yep, the superpack is just a simple wrapper around the correct installer, nothing fancy.
Is there a precompiled scipy that judges nosse/sse/sse2 automatically?
No, but there will be for 0.7, which hopfully is only days away now.
or is there a way to change ATLAS only according to my CPU?
Unfortunately not without rebuilding scipy yourself. Win32 binaries are built by linking atlas statically.
David
On Tue, Dec 2, 2008 at 4:36 PM, oyster <lepto.python@gmail.com> wrote:
sorry, but scipy-0.7.0b1-win32-superpack-python2.4.exe and numpy-1.2.1-win32-superpack-python2.4.exe crash on my old pc too, which uses duron 750MHz. So now I think it is not the problem with non-sse/sse/sse2 instruction
Ok. Just to be sure it is not a bug in the installer, can you tell me what the following commands give you ? python -c "import numpy; print numpy.show_config()" python -c "import scipy; print scipy.show_config()"
is there any method to find out the real reason except to compile from the source
It may not be easy, and you will need some tools, like a debugger to get a backtrace. What it the error you get exactly ? Illegal instruction or something else ? David
On Tue, Dec 2, 2008 at 11:52 PM, David Cournapeau <cournape@gmail.com> wrote:
On Tue, Dec 2, 2008 at 4:36 PM, oyster <lepto.python@gmail.com> wrote:
sorry, but scipy-0.7.0b1-win32-superpack-python2.4.exe and numpy-1.2.1-win32-superpack-python2.4.exe crash on my old pc too, which uses duron 750MHz. So now I think it is not the problem with non-sse/sse/sse2 instruction
Ok, I checked the machine code in scipy and it seems that the quadpack module (used by scipy.integrate) has a couple of SSE instructions. Code-wise, it is trivial to solve, but we may need a new numpy version for that. thanks for the report, and sorry for the trouble, David
On Wed, Dec 3, 2008 at 4:06 PM, David Cournapeau <cournape@gmail.com> wrote:
On Tue, Dec 2, 2008 at 11:52 PM, David Cournapeau <cournape@gmail.com> wrote:
On Tue, Dec 2, 2008 at 4:36 PM, oyster <lepto.python@gmail.com> wrote:
sorry, but scipy-0.7.0b1-win32-superpack-python2.4.exe and numpy-1.2.1-win32-superpack-python2.4.exe crash on my old pc too, which uses duron 750MHz. So now I think it is not the problem with non-sse/sse/sse2 instruction
Ok, I checked the machine code in scipy and it seems that the quadpack module (used by scipy.integrate) has a couple of SSE instructions. Code-wise, it is trivial to solve, but we may need a new numpy version for that.
Could you try this installer ? http://www.ar.media.kyoto-u.ac.jp/members/david/archives/scipy/scipy-0.7.0.d... This one should hopefully contains no SSE instructions at all for old CPU (the _quadpack.pyd contains no SSE2 instructions anymore), and should work. David
participants (2)
-
David Cournapeau
-
oyster