Compile errors building sdk on linux

I’ve just installed the cuda toolkit and sdk on linux, and started compiling the examples. It choked at this point:

g++ -m32 -W -Wall -Wimplicit -Wswitch -Wformat -Wchar-subscripts -Wparentheses -Wmultichar -Wtrigraphs -Wpointer-arith -Wcast-align -Wreturn-type -Wno-unused-function  -I. -I/usr/local/cuda/include -I../../common/inc -DUNIX -O3 -o ../../bin/obj/release/matrixmul_drv.cpp_o -c matrixmul_drv.cpp

matrixmul_drv.cpp: In function ‘CUresult initCUDA(char*, CUfunc_st**)’:

matrixmul_drv.cpp:207: error: jump to label ‘Error’

matrixmul_drv.cpp:191: error:   from here

matrixmul_drv.cpp:193: error:   crosses initialization of ‘char* module_path’

matrixmul_drv.cpp:207: error: jump to label ‘Error’

matrixmul_drv.cpp:187: error:   from here

matrixmul_drv.cpp:193: error:   crosses initialization of ‘char* module_path’

matrixmul_drv.cpp:207: error: jump to label ‘Error’

matrixmul_drv.cpp:184: error:   from here

matrixmul_drv.cpp:193: error:   crosses initialization of ‘char* module_path’

This looks like invalid code! you aren’t allowed to jump across variables – anyone know what is up with this?

Thanks!

Which version of G++ are you running? You should try with g++ 3.4, I had a similar problem with my g++ 4.1 and it disappeared with g+±3.4.

Hi Baisa,

This is a known bug. Please see this thread:
[url=“http://forums.nvidia.com/index.php?showtopic=28505&hl=matrixmul_drv”]The Official NVIDIA Forums | NVIDIA

Thanks,
Mark