Link cuda device functions to target
For example.
I have some library __device__
functions which will be compiled to some cuda application.
__device__ void er_ceil(er_ptr result, er_srcptr x) { if (x->exp <= 0) { result->frac = ceil(x->frac); if (result->frac == -0) result->frac = 0; result->exp = 0; } else { double dvalue; cuda::er_get_d(&dvalue, x); dvalue = ceil(dvalue); cuda::er_set_d(result, dvalue); } }
target_link_libraries - dont want to work
How can i link this function to some cuda target?