Numba Integration with Cython Pointers

Numba Support: How do I do ...?
1

I’m trying to call a DLL function with Numba using Ctypes. However, the function receives multiple arguments, including pointers. And it seems impossible to pass the values from a jit function.

Example:
a_ = np.asarray([1.5930, 1.9597, 0.0263*1000])
a = a_.ctypes.data_as(ctypes.POINTER(ctypes.c_float))

It is not accepted.

Cannot determine Numba type of <class 'ctypes.wintypes.LP_c_float'>

Does anybody know a way around this? Thanks!

I managed to use ‘intrinsic’ decorator, but I don’t really know how to use it properly

@intrinsic
def callingDLL(typingctx, T, P): 

    a_ = np.array([28.05/1000, 42.05/1000, 1])            
    b_ = np.array([1.5930, 1.9597, 0.0263*1000])   

    a = a_.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
    b = b_.ctypes.data_as(ctypes.POINTER(ctypes.c_float))

    result = types.CPointer(types.float64)

    def functiontobecalled(T, P, a, b):
        return functionDLL(T, P, a, b)
    
    return result, functiontobecalled

@jit(nopython=True)
def Teste(T, P): 
    result = functiontobecalled(T, P)
    print("Answer is:", result )
    return
2

There are lots of examples of calling shared libraries from numba. Do you have a complete minimal example of your problem that folks on the board could use to investigate?