I have the exact same error on CentOS. I just installed g++4.4. How do I
change the compiler setting ?
PhD Student in Physics
On Fri, Dec 3, 2010 at 2:28 AM, avidday <avidday(a)gmail.com> wrote:
I think that
your boost libraries are not found, as said here :
> *** Error occurred in plausibility checking for path of Boost Python
> *** Error occurred in plausibility checking
for path of Boost Thread
try to set the names according to what you have in /usr/local/lib
Furthermore, you installed boost via yum I suppose and that gave you a
old boost distribution (1.33), you should get at least a boost 1.39
if I remember well and thus install it "by hand" as explained in the
installing instructions. Before doing that, remove the current boost 1.33
from you system.
No that will make no difference. The modern pyCUDA distribution
includes what boost stuff it needs internally and upgading boost won't
change anything. I have functional 0.94.2 installations on Centos 5.2
systems with the stock boost 1.33. The other warnings (boost and cuda
libraries) are just that, warnings, and the build will finish with or
The source of these errors is a compiler bug in the Redhat 5 g++
compiler for versions 4.1.2-48 and lower which won't compile the boost
templating pyCUDA contains. There is a patch in the -49 release which
their bug database says fixes it (I am not sure whether it is
available via updates from Redhat or not and I can't verify the fix
works), otherwise use the g++ 4.4 "preview" to build pyCUDA.
To the original poster: there is an incompatibility in the 0.94.1
release code with Python 2.4. Andreas kindly checked in a small patch
for _pycuda_struct.c for me a few weeks ago which fixes it. I hope
these adventures with pyCUDA have nothing to do with a posting on the
nvidia CUDA forum a few days ago.....
PyCUDA mailing list