Cannot create 32 bit debugable binary on 64 bit machine

Accelerated Computing CUDA CUDA Programming and Performance
#1

Hi everyone,

I’m using CUDA 2.3 on a 64 Bit Linux machine and trying to compile a 32 bit binary that is debugable. Everything works fine with my “release” configuration. The following example is sufficient for producing the error:

__global__ void testKernel() { }

int main() {

	return 0;

}

Compiling with:

nvcc -m32 -o ./test.cu.o -c test.cu

works fine. But as soon as using the “-G” flag, the following error occurs:

nvcc -m32 -G -o ./test.cu.o -c test.cu

ld: Relocatable linking with relocations from format elf32-i386 (/tmp/tmpxft_..._test.sm_10.cubin.dwarf.o) to format elf64-x86-64 (test.cu.o) is not supported

Did anyone run into this problem yet? It would be very helpful if someone could test this on his 64 bit machine.

Thanks in advance.