Optimizing numpy.dot with Cython

Question

I have the following piece of code which I'd like to optimize using Cython:

sim = numpy.dot(v1, v2) / (sqrt(numpy.dot(v1, v1)) * sqrt(numpy.dot(v2, v2))) 
dist = 1-sim
return dist

I have written and compiled the .pyx file and when I ran the code I do not see any significant improvement in performance. According to the Cython documentation I have to add c_types. The HTML file generated by Cython indicates that the bottleneck is the dot products (which is expected of course). Does this mean that I have to define a C function for the dot products? If yes how do I do that?

EDIT:

After some research I have come up with the following code. The improvement is only marginal. I am not sure if there is something I can do to improve it :

from __future__ import division
import numpy as np
import math as m
cimport numpy as np
cimport cython

cdef extern from "math.h":
    double c_sqrt "sqrt"(double)

ctypedef np.float reals #typedef_for easier readding

cdef inline double dot(np.ndarray[reals,ndim = 1] v1, np.ndarray[reals,ndim = 1] v2):
  cdef double result = 0
  cdef int i = 0
  cdef int length = v1.size
  cdef double el1 = 0
  cdef double el2 = 0
  for i in range(length):
    el1 = v1[i]
    el2 = v2[i]
    result += el1*el2
  return result

@cython.cdivision(True)
def distance(np.ndarray[reals,ndim = 1] ex1, np.ndarray[reals,ndim = 1] ex2):
  cdef double dot12 = dot(ex1, ex2)
  cdef double dot11 = dot(ex1, ex1)
  cdef double dot22 = dot(ex2, ex2)
  cdef double sim = dot12 / (c_sqrt(dot11 * dot22))
  cdef double dist = 1-sim    
  return dist

Take a look at this recent very similar question: http://stackoverflow.com/questions/10772513/power-spectrum-with-cython/10773360 — Nick Craig-Wood, May 28 '12 at 18:17

JoshAdel · Accepted Answer · 2012-05-29T02:47:34.537

13

As a general note, if you are calling numpy functions from within cython and doing little else, you generally will see only marginal gains if any at all. You generally only get massive speed-ups if you are statically typing code that makes use of an explicit for loop at the python level (not in something that is calling the Numpy C-API already).

You could try writing out the code for a dot product with all of the static typing of the counter, input numpy arrays, etc, with wraparound and boundscheck set to False, import the clib version of the sqrt function and then try to leverage the parallel for loop (prange) to make use of openmp.

edited May 29 '12 at 02:47

answered May 29 '12 at 00:20

JoshAdel

66,734
27
141
140

That's what I discovered so far. I'll try your suggestions. Thanks – George Eracleous May 29 '12 at 01:32
Maybe I misunderstood your comment. Do you suggest that my code as it stands cannot be made any faster with Cython or do you mean that if I just use the compiled Cython code without any static typing then I won't have any gain? Please, have a look at my revised code in the question! I am not sure if that's how I'm supposed to do it. – George Eracleous May 29 '12 at 10:19
@GeorgeEracleous what I was trying to say was that your original code that called `np.dot` from within cython wasn't expected to get any sort of large speed-up. The new code that you posted was more like what I was imagining trying. Just a couple of note (1) I'm not sure if your `dot` implementation is a good candidate for inlining. (2) Since you are looping over a numpy array, I recommend using the `boundscheck` and `wraparound` decorators and setting them to `False`. (3) In your setup.py file, make sure you are using optimization flags (e.g. `-O3`). – JoshAdel May 29 '12 at 11:18
2

In general though, numpy's `dot` is going to be highly optimized if compiled against BLAS or MKL – JoshAdel May 29 '12 at 11:20

score 0 · Answer 2 · answered May 28 '12 at 17:18

0

You can change the expression

sim = numpy.dot(v1, v2) / (sqrt(numpy.dot(v1, v1)) * sqrt(numpy.dot(v2, v2)))

to

sim = numpy.dot(v1, v2) / sqrt(numpy.dot(v1, v1) * numpy.dot(v2, v2))

answered May 28 '12 at 17:18

YXD

31,741
15
75
115

Hey, you're right. I should have noticed that. However, I've just tried it and still no significant improvement! But the point is to manage to run this in cython anw – George Eracleous May 28 '12 at 17:23

Optimizing numpy.dot with Cython

2 Answers2