Loop over a Numpy array with Cython

Question

Let a and b be two numpy.float arrays of length 1024, defined with

cdef numpy.ndarray a
cdef numpy.ndarray b

I notice that:

cdef int i
for i in range(1024):
    b[i] += a[i]

is considerably slower than:

b += a

Why?

I really need to be able to loop manually over arrays.

You're read and followed docs.cython.org/src/tutorial/numpy.html? I mean I doubt you are going to beat b+=a in terms of speed, but what kind of speeds are you actually getting? — derricw
– derricw, Commented May 14, 2015 at 15:39
The latter is still using Numpy's own ufunc loops which can include a number of enhancements depending on how your Numpy was compiled. The former does not, and is further harmed by slow indexing of the ndarray objects; the link @ballsdotballs provided discusses this and how to fix it. — Iguananaut
– Iguananaut, Commented May 14, 2015 at 15:43
Just a side comment : is cdef float x a 32-bit or 64-bit float? Which cdef type will work with numpy.float32, numpy.float64? I looked at the doc but didn't find precise correspondance. — Basj
– Basj, Commented May 14, 2015 at 15:46
For future reference, I checked : float <=> numpy.float32, double <=> numpy.float64 = numpy.float — Basj
– Basj, Commented May 14, 2015 at 16:22

Saullo G. P. Castro · Accepted Answer · 2015-05-14 23:13:51Z

3

The difference will be smaller if you tell Cython the data type and the number of dimensions for a and b:

cdef numpy.ndarray[np.float64_t, ndim=1] a, b

Although the difference will be smaller, you won't beat b += a because this is using NumPy's SIMD-boosted functions (which will perform depending if your CPU supports SIMD).

answered May 14, 2015 at 23:13

Saullo G. P. Castro

59.4k28 gold badges191 silver badges244 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Loop over a Numpy array with Cython

1 Answer 1

Comments

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Related