Paweł Bylina <pawel.bylina(a)renegatt.com> writes:
> I would like to download examples pycuda from wiki but seems that wiki has
> a problem ;-)
What exactly were you trying to do? The above URL won't work in a
browser, that's expected. You can use
examples/download-examples-from-wiki.py to grab all examples in an
automated way. I just tried that, it seems to work for me. You can also
copy and paste them off the wiki one by one:
Either way should work.
I installed PyCuda per Christoph Bussler's instructions.
However, I am using _later_ libraries than the ones the instruction suggest.
Of course, I am getting "DLL Load Failed, the specified module could not be
When I do a dependency walk on both the 32 bit and 64 bit NVCUDA.DLL, I get
errors with both, but it seems the 32-bit is the one that's being used (is
this correct?). The 32-bit dll is missing
IESHIMS.DLL (the path to it in "Internet Explorer" is set to include both
the 32 bit and 64 bit versions);
MF.DLL - still looking for this ;
MFPLAT.DLL - still looking for this;
MFREADWRITE.DLL - still looking for this.
Any help will be greatly appreciated!
My setup is as below (all x64). Original instructions as per Christoph
Windows 7 64-bit Ultimate N
Python - I am using 2.7
get Windows X86-64 MSI Installer (2.7)
PyCUDA - I am using pycuda-2012.1.win-amd64-py2.7.exe
Boost.Python - I am using boost_python-1.52.win-amd64-py2.7.exe
NumPy - I am using numpy-unoptimized-1.7.1rc1.win-amd64-py2.7.exe
CUDA - I am using 5.0, although 3.2 is installed as well.
Environment - this has been set
set HOME environment variable, e.g. to %HOMEDRIVE%%HOMEPATH%
PyTools - I am using pytools-2012.1.tar.gz (md5)
install with "python setup.py install"
Visual C++ - I have 2008 as well as 2010 installed. VC++ path points to 2008
(Version 8 and 9 are supported by PyCUDA, not 10!. Trying to install 10 will
not work as PyCUDA checks the version on compilation)
install ensure that the path to the cl.exe (C++ compiler) is in the path
C:\Program Files (x86)\Microsoft Visual Studio 9.0\VC\bin;C:\Program Files
(x86)\Microsoft Visual Studio9.0\VC\bin\amd64;C:\Program Files
(x86)\Microsoft Visual Studio 9.0\Common7\IDE;
nvcc.profile - I have added this to both 3.2 and 5.0
Edit nvcc.profile (in CUDA\3.2\bin) and set INCLUDES flag to this:
INCLUDES += "-I$(TOP)/include" "-I$(TOP)/include/cudart" "-IC:/Program Files
(x86)/Microsoft Visual Studio 9.0/VC/include" $(_SPACE_)
It is perfectly possible to use several GPU in parallel. I believe it was discussed in the past on the list, but if you want a relatively simple example you can look at the PyNX library : it was written for X-ray scattering calculations on large number of atoms, and can either use pyOpenCL or pyCUDA. It uses python threads to distribute jobs on several GPU.
Most relevant part is the gnu.py file (http://pynx.svn.sourceforge.net/viewvc/pynx/trunk/pynx/gpu.py?revision=24...)
Vincent Favre-Nicolin http://inac.cea.fr
CEA/Grenoble Institut Nanosciences & Cryogénie
Laboratoire SP2M/Nano-structures et Rayonnement Synchrotron
17, rue des Martyrs
38054 Grenoble Cedex 9 - France
Université Joseph Fourier http://www.ujf-grenoble.fr
tél: (+33) 4 38 78 95 40 fax: (+33) 4 38 78 51 38