NVCC flag short-names don't work with --device-link (-dlink) CUDA 6.5

Accelerated Computing CUDA CUDA Setup and Installation
1

After installing CUDA 6.5 on OS X 10.11.2 (El Capitan), I tried running the example code for separable compilation:
http://devblogs.nvidia.com/parallelforall/separate-compilation-linking-cuda-device-code/

At the critical step the --device-link flag command yields an error:
$ nvcc –arch=sm_20 –dlink v3.o particle.o main.o –o gpuCode.o
nvcc fatal : Don’t know what to do with ‘–o’

If you encounter this, all you need to do is replace the short names with the long ones…
nvcc --gpu-architecture=sm_20 --device-link v3.o particle.o main.o --output-file gpuCode.o

Any advice on how to fix this, or how to report it?

2

You’re using an incorrect dash symbol.

Your command that you have shown here:

$ nvcc –arch=sm_20 –dlink v3.o particle.o main.o –o gpuCode.o

is using this dash symbol:

–

the correct dash symbol is this one:

So try this command:

$ nvcc -arch=sm_20 -dlink v3.o particle.o main.o -o gpuCode.o