Is installed CUDA capable graphics card necessary (in Linux) for compiling CUDA programs with nvcc? Or one can compile programs everywhere and run only on such systems?
Is installed CUDA capable graphics card necessary (in Linux) for compiling CUDA programs with
Share
No, a graphics card is not necessary for compilation.
You don’t even need one to run the program; you can have it emulated in software.
UPDATE Ok, apparently, SW emulation hasn’t been supported since CUDA 3.0.