CODEWITHSUNDEEP
pythonccython
user3646557
user3646557

Reputation: 309

Does typing increase speed of execution when wrapping C functions with Cython?

I am using Cython to wrap some C code. This is the declaration of a typical routine (the actual ones involve about 13 to 17 arguments).

void my_func(double *__restrict__ lhs, const int xlhs, const int oxlhs);

Right now, in the .pyx file this is how I define the wrapper

def preCond1D(np.ndarray[double, ndim=2] lhs not None, xlhs not None, oxlhs not None): 
   preCond1D_func( <double*> np.PyArray_DATA(lhs), xlhs, oxlhs)

Would there be any benefit (performance or otherwise) if instead I explicitly declare (i.e. typing) the int variables, as

def preCond1D(np.ndarray[double, ndim=2] lhs, xlhs: cython.int32, oxlhs: cython.int32): 
   preCond1D_func( <double*> np.PyArray_DATA(lhs), xlhs, oxlhs)

Upvotes: 1

Views: 135

Answers (1)

DavidW
DavidW

Reputation: 30929

Cython does use type hints, so generally they are useful. You can control this by the annotation_typing directive.

In this case though they'll make absolutely no difference. You will get exactly one set of data conversions from Python values to C types. With the type hints they'll happen right at the start of preCond1D. Without the type hints they'll happen at the call to preCond1D_func.

The only benefit would be the documentation that the type hints provide.

Upvotes: 1

Related Questions