On Mon, 20 Jun 2011 16:03:39 -0700, Scott Biersdorff <sco...@cs.uoregon.edu> wrote: > On Mon, 2011-06-20 at 18:42 -0400, Andreas Kloeckner wrote: > > > > No. The evidence is in your stack trace, actually. The failing > > compilation is trying to compile a kernel for an array operation (the > > 2*a in your case). I suppose we should add a way to pass options to nvcc > > for behind-your-back compilations? A global variable maybe (shudder)? > > That's what I ended up doing myself (Like you have done for > CUDA_DEBUGGING). However you might want a more flexible solution so you > don't end up dealing with all possible nvcc options as a separate global > variables. Also be warned that nvcc will not allow some options to set > more than once on the same command.
Added. http://documen.tician.de/pycuda/driver.html#pycuda.compiler.DEFAULT_NVCC_FLAGS Andreas
pgpYLVUMJ5WVF.pgp
Description: PGP signature
_______________________________________________ PyCUDA mailing list PyCUDA@tiker.net http://lists.tiker.net/listinfo/pycuda