Set up compiler preferences correclty for nvcc usage about brian2cuda HOT 3 CLOSED

Just for reference: it's of course the same issue for brian2genn, which now has a devices.genn.extra_compile_args_nvcc preference which is passed to the nvcc call. It also uses the flags returned by get_compiler_and_args (i.e. the gcc or msvc arguments, depending on the system) for the standard C++ compiler calls.

from brian2cuda.

Oh yeah, right. Thanks for pointing it out.
So looking at the brian2genn makefile template and the GeNN makefile (and the corresponding WIN versions) should give a good idea on how to do this.
And brian2genn also uses its own get_compile_args(), defined in brian2genn/device.py.

from brian2cuda.

Just one more comment on that: you should be fine with just using Brian's standard get_compiler_and_args (which gives you the msvc preferences on a Windows system, and the gcc preferences otherwise). The reason why brian2genn has its own get_compile_args is that GeNN itself does code generation and has two separate preferences for Windows and Linux/OS-X, so we simply give it both.

from brian2cuda.