While compilation in standalone mode is significantly faster than in Cython runtime (which we are already working on replacing completely, see #1769), it is still slower than it could/should be, especially in Brian2CUDA (see brian-team/brian2cuda#179). We should investigate ways of improving this, e.g. by merging code object files together (to have fewer files to compile in total), by pre-compiling header files and library functions, and potentially by using helper tools such as ccache.
While compilation in standalone mode is significantly faster than in Cython runtime (which we are already working on replacing completely, see #1769), it is still slower than it could/should be, especially in Brian2CUDA (see brian-team/brian2cuda#179). We should investigate ways of improving this, e.g. by merging code object files together (to have fewer files to compile in total), by pre-compiling header files and library functions, and potentially by using helper tools such as ccache.