I am trying to compile the CUDA samples on suse 10.2
The problem is that I have GCC 4.2.2 installed as well as the standard gcc. Compiling with 4.2.2 gives errors, and I’d like to use 4.1.2.
But when I type
export CC=/usr/bin/gcc
make
the compilation still uses the wrong version of the compiler.
comperrors.txt (22.6 KB)
Are you sure that gcc is the 4.1.2 version? To have nvcc use the 4.1.2 take a look at some of the opensuse 10.3 threads in this forum.
Tony
OK, now I get a different error:
brodskie@linux-bw3k:~/NVIDIA_CUDA_SDK> make
make[1]: Entering directory /home/brodskie/NVIDIA_CUDA_SDK/common' a - obj/release/bank_checker.cpp_o a - obj/release/cmd_arg_reader.cpp_o a - obj/release/cutil.cpp_o a - obj/release/stopwatch.cpp_o a - obj/release/stopwatch_linux.cpp_o a - obj/release/multithreading.cpp_o make[1]: Leaving directory /home/brodskie/NVIDIA_CUDA_SDK/common’
make[1]: Entering directory /home/brodskie/NVIDIA_CUDA_SDK/common' a - obj/release/paramgl.cpp_o a - obj/release/param.cpp_o make[1]: Leaving directory /home/brodskie/NVIDIA_CUDA_SDK/common’
make -C projects/template/
make[1]: Entering directory /home/brodskie/NVIDIA_CUDA_SDK/projects/template' ../../lib/libcutil.a(cutil.cpp_o): In function CUTBoolean (anonymous namespace)::cutWriteFile(char const*, unsigned char const*, unsigned int, unsigned char, bool)‘:
cutil.cpp:(.text+0x868): undefined reference to `std::basic_ostream<char, std::char_traits >& std::__ostream_insert<char, std::char_traits >(std::basic_ostream<char, std::char_traits >&, char const*, int)’
and the errors continue like that ( I attached the full message)
I can compile other applications perfectly well, I don’t understand why make cannot find the STL (that is what I at least think is wrong).
errors.txt (30 KB)
I did solve the problem , but only by deleting all traces of the second installation of gcc 4.2.1 Since the developers of gcc didn’t bother to include an uninstaller, this involved a fair amount of hunting.
I still find it very interesting that the building process of the nvidia samples used the wrong versions of the libraries included with the 4.2.1 compiler; my build environment was sane enough to allow other programs to compile just fine.
CUDA is tested & qualified against specific Linux distributions. Using a compiler toolchain which didn’t ship with a qualified Linux distribution is effectively equivalent to using an unqualified Linux distribution.
Please note that if you wish to use an alternative gcc, you should be specifying the --compiler-bindir option:
–compiler-bindir (-ccbin)
Specify the directory in which the compiler executable (Microsoft Visual
Studion cl, or a gcc derivative) resides. By default, this executable is
expected in the current executable search path.