Can't compile any .cu files on x86_64

I tried compiling the samples on an x86_64 machine and it throws an error I don’t understand:
nvcc -o obj/release/dxtc.cu_o -c -m64 -I. -I/usr/local/cuda_libs/cuda64/include -I…/…/common/inc -DUNIX -O3
Internal error: check_target_config: target wchar_t is too large

Any ideas or workarounds?

Note: I added -m64 in the makefile, needed because nvcc assumes 32 bits.

EDIT: After switching to the RHEL5 runtime, it now works (I was using SUSE 10.2). I’m on kubuntu feisty btw.


Are you sure you have installed the right toolkit?
Which OS version are you running?

nvcc -O3 -m64 -c -I /usr/local/NVIDIA_CUDA_SDK/common/inc/ -DUNIX
compiles just fine on my machine (RHEL4 64bit)

I just edited my post, it now works. Thank you.

For what it is worth, I have compiled the examples on a 64bit OpenSuse 10.2 machine and did not encounter any problems… I had to install a few development items beyond the standard OpenSuse install but nothing I thought was unusual.