The compiler flag seemed to work. pycuda now compiles and I've run both the
test_driver and test_gpuarray test programs with success. Thank-you for all
your help.  Also, I don't think I mentioned this before, but I'm running
this on openSuSE 10.3 if anyone is interested. I've also tried
unsuccessfully to install on openSuSE 11.0 with a patched version of CUDA,
but I have a feeling my lack of success is related to the patch and not a
fault in pycuda at all.  Again thanks to everyone for your help.

Brad Zima

On Tue, Nov 4, 2008 at 12:50 AM, Andreas Klöckner
<[EMAIL PROTECTED]>wrote:

> On Dienstag 04 November 2008, Brad Zima wrote:
> > Well, it looks like some tweaking of my --boost-inc-dir parameters found
> > the header file for me. Apparently, I needed the full-path for my home
> > directory instead of ~/. However, when compiling I now get an internal
> > compiler error from gcc, as follows:
> >
> > [EMAIL PROTECTED]:~/pycuda> make
> > [snip]
> >
> /home/infinite/pool/include/boost-1_36/boost/type_traits/detail/cv_traits_i
> >mpl.hpp:37: internal compiler error: in make_rtl_for_nonlocal_decl, at
> > cp/decl.c:4971 Please submit a full bug report,
> > with preprocessed source if appropriate.
> > See <URL:http://bugs.opensuse.org> for instructions.
> > error: command 'gcc' failed with exit status 1
> > make: *** [all] Error 1
>
> Let me guess: gcc 4.1.something?
>
> If so, include
>
> CXXFLAGS = ['-DBOOST_PYTHON_NO_PY_SIGNATURES']
>
> in your siteconf.py.
>
> Andreas
>
_______________________________________________
PyCuda mailing list
[email protected]
http://tiker.net/mailman/listinfo/pycuda_tiker.net

Reply via email to