Cuda 8.0 32-bit cross compilation

I want to compile my 32 bit code on 64 bit machine. So I am using -m32 flag with nvcc. But while linking cuda libraries, the compiler gives the following error :

skipping incompatible /usr/local/cuda-8.0/targets/x86_64-linux/lib//libcudadevrt.a when searching for -lcudadevrt

The reason why I am compiling with m32 flag is that I am using an external 32 bit library whose source code is not available.

So is there any way I can compile my cuda based 32 bit code on 64 bit machine.

I am using cuda 8.0 on linux machine.

Thanks.