Cython

Cython is an "optimizing static compiler " that combines Python with C to generate optimized code. Since Cython is a superset of Python, all valid Python programs are also valid Cython programs. However, by providing hints and static typing, we can get much faster programs. Note that while numba often provides similar speedups with less work,, an advantage of Cython is that it is easy to distribute optimized Cython modules since they can be built with the standard Python setup.py script.

We have already seen how to use Cython to wrap C and C++ functions from existing libraries. Here we will see how to use Cython to speed up Python functions.


In [1]:
%matplotlib inline
import matplotlib.pyplot as plt

Utility function for timing functions


In [2]:
import time

In [3]:
def timer(f, *args, **kwargs):
    start = time.clock()
    ans = f(*args, **kwargs)
    return ans, time.clock() - start

In [4]:
def report(fs, *args, **kwargs):
    ans, t = timer(fs[0], *args, **kwargs)
    for f in fs[1:]:
        ans_, t_ = timer(f, *args, **kwargs)
        print('%s: %.1f' % (f.__name__, t/t_))

Incremental improvements

Generally, we start with a pure Python function, run it through Cython with the annotate -a flag, and incrementally modify the code until the yellow parts are minimized.

How to build Cython modules

From official docs

Using Cython consists of these steps:

  • Write a .pyx source file
  • Run the Cython compiler to generate a C file
  • Run a C compiler to generate a compiled library
  • Run the Python interpreter and ask it to import the module

In the Jupyter notebook, we can use the %%cython cell magic to automate these steps.


In [5]:
%load_ext cython

Matrix multiplication example


In [6]:
def matrix_multiply(u, v, res):
    m, n = u.shape
    n, p = v.shape
    for i in range(m):
        for j in range(p):
            res[i,j] = 0
            for k in range(n):
                res[i,j] += u[i,k] * v[k,j]
    return res

In [7]:
import numpy as np
u = np.random.random((10,20))
v = np.random.random((20,5))

In [8]:
res = np.zeros((u.shape[0], v.shape[1]))
matrix_multiply(u, v, res)


Out[8]:
array([[ 3.11449823,  5.35977303,  5.48778181,  5.14085938,  4.54520686],
       [ 3.68374999,  4.67333347,  4.37704824,  4.61282881,  3.85293391],
       [ 4.01582061,  5.53693287,  5.23375753,  4.67309499,  4.59460048],
       [ 4.07493259,  4.63093078,  5.00670814,  4.92908673,  4.23099528],
       [ 3.56821364,  4.64831502,  3.89552072,  4.74094422,  3.83194289],
       [ 3.59114576,  4.14569789,  4.77227321,  4.45942977,  3.77867553],
       [ 5.17616931,  5.73851388,  5.6071998 ,  5.99414005,  5.01139514],
       [ 2.89575575,  4.00384915,  4.71317099,  4.75876412,  4.30115292],
       [ 4.08424318,  5.54013852,  5.29577668,  5.74588951,  4.45967311],
       [ 3.95142186,  5.68656356,  5.02933096,  5.7503449 ,  5.06054258]])

In [9]:
res = np.zeros((u.shape[0], v.shape[1]))
%timeit -r3 -n3 matrix_multiply(u, v, res)


3 loops, best of 3: 653 µs per loop

Using Cython annnotations to identify bottlenecks


In [10]:
%%cython -a

import numpy as np

def matrix_multiply1(u, v, res):
    m, n = u.shape
    n, p = v.shape
    for i in range(m):
        for j in range(p):
            res[i,j] = 0
            for k in range(n):
                res[i,j] += u[i,k] * v[k,j]
    return res


Out[10]:
Cython: _cython_magic_65555dffd12a135eb525e1912be8f576.pyx

Generated by Cython 0.23.5

Yellow lines hint at Python interaction.
Click on a line that starts with a "+" to see the C code that Cython generated for it.

 01: 
+02: import numpy as np
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_numpy, 0, -1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_1);
  if (PyDict_SetItem(__pyx_d, __pyx_n_s_np, __pyx_t_1) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 03: 
+04: def matrix_multiply1(u, v, res):
/* Python wrapper */
static PyObject *__pyx_pw_46_cython_magic_65555dffd12a135eb525e1912be8f576_1matrix_multiply1(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
static PyMethodDef __pyx_mdef_46_cython_magic_65555dffd12a135eb525e1912be8f576_1matrix_multiply1 = {"matrix_multiply1", (PyCFunction)__pyx_pw_46_cython_magic_65555dffd12a135eb525e1912be8f576_1matrix_multiply1, METH_VARARGS|METH_KEYWORDS, 0};
static PyObject *__pyx_pw_46_cython_magic_65555dffd12a135eb525e1912be8f576_1matrix_multiply1(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
  PyObject *__pyx_v_u = 0;
  PyObject *__pyx_v_v = 0;
  PyObject *__pyx_v_res = 0;
  PyObject *__pyx_r = 0;
  __Pyx_RefNannyDeclarations
  __Pyx_RefNannySetupContext("matrix_multiply1 (wrapper)", 0);
  {
    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_u,&__pyx_n_s_v,&__pyx_n_s_res,0};
    PyObject* values[3] = {0,0,0};
    if (unlikely(__pyx_kwds)) {
      Py_ssize_t kw_args;
      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
      switch (pos_args) {
        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
        case  0: break;
        default: goto __pyx_L5_argtuple_error;
      }
      kw_args = PyDict_Size(__pyx_kwds);
      switch (pos_args) {
        case  0:
        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_u)) != 0)) kw_args--;
        else goto __pyx_L5_argtuple_error;
        case  1:
        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_v)) != 0)) kw_args--;
        else {
          __Pyx_RaiseArgtupleInvalid("matrix_multiply1", 1, 3, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
        }
        case  2:
        if (likely((values[2] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_res)) != 0)) kw_args--;
        else {
          __Pyx_RaiseArgtupleInvalid("matrix_multiply1", 1, 3, 3, 2); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
        }
      }
      if (unlikely(kw_args > 0)) {
        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "matrix_multiply1") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
      }
    } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
      goto __pyx_L5_argtuple_error;
    } else {
      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
    }
    __pyx_v_u = values[0];
    __pyx_v_v = values[1];
    __pyx_v_res = values[2];
  }
  goto __pyx_L4_argument_unpacking_done;
  __pyx_L5_argtuple_error:;
  __Pyx_RaiseArgtupleInvalid("matrix_multiply1", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
  __pyx_L3_error:;
  __Pyx_AddTraceback("_cython_magic_65555dffd12a135eb525e1912be8f576.matrix_multiply1", __pyx_clineno, __pyx_lineno, __pyx_filename);
  __Pyx_RefNannyFinishContext();
  return NULL;
  __pyx_L4_argument_unpacking_done:;
  __pyx_r = __pyx_pf_46_cython_magic_65555dffd12a135eb525e1912be8f576_matrix_multiply1(__pyx_self, __pyx_v_u, __pyx_v_v, __pyx_v_res);
  int __pyx_lineno = 0;
  const char *__pyx_filename = NULL;
  int __pyx_clineno = 0;

  /* function exit code */
  __Pyx_RefNannyFinishContext();
  return __pyx_r;
}

static PyObject *__pyx_pf_46_cython_magic_65555dffd12a135eb525e1912be8f576_matrix_multiply1(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_u, PyObject *__pyx_v_v, PyObject *__pyx_v_res) {
  PyObject *__pyx_v_m = NULL;
  PyObject *__pyx_v_n = NULL;
  PyObject *__pyx_v_p = NULL;
  PyObject *__pyx_v_i = NULL;
  PyObject *__pyx_v_j = NULL;
  PyObject *__pyx_v_k = NULL;
  PyObject *__pyx_r = NULL;
  __Pyx_RefNannyDeclarations
  __Pyx_RefNannySetupContext("matrix_multiply1", 0);
/* … */
  /* function exit code */
  __pyx_L1_error:;
  __Pyx_XDECREF(__pyx_t_1);
  __Pyx_XDECREF(__pyx_t_2);
  __Pyx_XDECREF(__pyx_t_3);
  __Pyx_XDECREF(__pyx_t_4);
  __Pyx_XDECREF(__pyx_t_12);
  __Pyx_XDECREF(__pyx_t_13);
  __Pyx_XDECREF(__pyx_t_14);
  __Pyx_XDECREF(__pyx_t_15);
  __Pyx_AddTraceback("_cython_magic_65555dffd12a135eb525e1912be8f576.matrix_multiply1", __pyx_clineno, __pyx_lineno, __pyx_filename);
  __pyx_r = NULL;
  __pyx_L0:;
  __Pyx_XDECREF(__pyx_v_m);
  __Pyx_XDECREF(__pyx_v_n);
  __Pyx_XDECREF(__pyx_v_p);
  __Pyx_XDECREF(__pyx_v_i);
  __Pyx_XDECREF(__pyx_v_j);
  __Pyx_XDECREF(__pyx_v_k);
  __Pyx_XGIVEREF(__pyx_r);
  __Pyx_RefNannyFinishContext();
  return __pyx_r;
}
/* … */
  __pyx_tuple_ = PyTuple_Pack(9, __pyx_n_s_u, __pyx_n_s_v, __pyx_n_s_res, __pyx_n_s_m, __pyx_n_s_n, __pyx_n_s_p, __pyx_n_s_i, __pyx_n_s_j, __pyx_n_s_k); if (unlikely(!__pyx_tuple_)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_tuple_);
  __Pyx_GIVEREF(__pyx_tuple_);
/* … */
  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_46_cython_magic_65555dffd12a135eb525e1912be8f576_1matrix_multiply1, NULL, __pyx_n_s_cython_magic_65555dffd12a135eb5); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_1);
  if (PyDict_SetItem(__pyx_d, __pyx_n_s_matrix_multiply1, __pyx_t_1) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+05:     m, n = u.shape
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_u, __pyx_n_s_shape); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 5; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_1);
  if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
    PyObject* sequence = __pyx_t_1;
    #if CYTHON_COMPILING_IN_CPYTHON
    Py_ssize_t size = Py_SIZE(sequence);
    #else
    Py_ssize_t size = PySequence_Size(sequence);
    #endif
    if (unlikely(size != 2)) {
      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
      {__pyx_filename = __pyx_f[0]; __pyx_lineno = 5; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    }
    #if CYTHON_COMPILING_IN_CPYTHON
    if (likely(PyTuple_CheckExact(sequence))) {
      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 0); 
      __pyx_t_3 = PyTuple_GET_ITEM(sequence, 1); 
    } else {
      __pyx_t_2 = PyList_GET_ITEM(sequence, 0); 
      __pyx_t_3 = PyList_GET_ITEM(sequence, 1); 
    }
    __Pyx_INCREF(__pyx_t_2);
    __Pyx_INCREF(__pyx_t_3);
    #else
    __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 5; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __Pyx_GOTREF(__pyx_t_2);
    __pyx_t_3 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 5; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __Pyx_GOTREF(__pyx_t_3);
    #endif
    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
  } else {
    Py_ssize_t index = -1;
    __pyx_t_4 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 5; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __Pyx_GOTREF(__pyx_t_4);
    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
    __pyx_t_5 = Py_TYPE(__pyx_t_4)->tp_iternext;
    index = 0; __pyx_t_2 = __pyx_t_5(__pyx_t_4); if (unlikely(!__pyx_t_2)) goto __pyx_L3_unpacking_failed;
    __Pyx_GOTREF(__pyx_t_2);
    index = 1; __pyx_t_3 = __pyx_t_5(__pyx_t_4); if (unlikely(!__pyx_t_3)) goto __pyx_L3_unpacking_failed;
    __Pyx_GOTREF(__pyx_t_3);
    if (__Pyx_IternextUnpackEndCheck(__pyx_t_5(__pyx_t_4), 2) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 5; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __pyx_t_5 = NULL;
    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
    goto __pyx_L4_unpacking_done;
    __pyx_L3_unpacking_failed:;
    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
    __pyx_t_5 = NULL;
    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
    {__pyx_filename = __pyx_f[0]; __pyx_lineno = 5; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __pyx_L4_unpacking_done:;
  }
  __pyx_v_m = __pyx_t_2;
  __pyx_t_2 = 0;
  __pyx_v_n = __pyx_t_3;
  __pyx_t_3 = 0;
+06:     n, p = v.shape
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_v, __pyx_n_s_shape); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_1);
  if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
    PyObject* sequence = __pyx_t_1;
    #if CYTHON_COMPILING_IN_CPYTHON
    Py_ssize_t size = Py_SIZE(sequence);
    #else
    Py_ssize_t size = PySequence_Size(sequence);
    #endif
    if (unlikely(size != 2)) {
      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
      {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    }
    #if CYTHON_COMPILING_IN_CPYTHON
    if (likely(PyTuple_CheckExact(sequence))) {
      __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 1); 
    } else {
      __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
      __pyx_t_2 = PyList_GET_ITEM(sequence, 1); 
    }
    __Pyx_INCREF(__pyx_t_3);
    __Pyx_INCREF(__pyx_t_2);
    #else
    __pyx_t_3 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __Pyx_GOTREF(__pyx_t_3);
    __pyx_t_2 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __Pyx_GOTREF(__pyx_t_2);
    #endif
    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
  } else {
    Py_ssize_t index = -1;
    __pyx_t_4 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __Pyx_GOTREF(__pyx_t_4);
    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
    __pyx_t_5 = Py_TYPE(__pyx_t_4)->tp_iternext;
    index = 0; __pyx_t_3 = __pyx_t_5(__pyx_t_4); if (unlikely(!__pyx_t_3)) goto __pyx_L5_unpacking_failed;
    __Pyx_GOTREF(__pyx_t_3);
    index = 1; __pyx_t_2 = __pyx_t_5(__pyx_t_4); if (unlikely(!__pyx_t_2)) goto __pyx_L5_unpacking_failed;
    __Pyx_GOTREF(__pyx_t_2);
    if (__Pyx_IternextUnpackEndCheck(__pyx_t_5(__pyx_t_4), 2) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __pyx_t_5 = NULL;
    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
    goto __pyx_L6_unpacking_done;
    __pyx_L5_unpacking_failed:;
    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
    __pyx_t_5 = NULL;
    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
    {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __pyx_L6_unpacking_done:;
  }
  __Pyx_DECREF_SET(__pyx_v_n, __pyx_t_3);
  __pyx_t_3 = 0;
  __pyx_v_p = __pyx_t_2;
  __pyx_t_2 = 0;
+07:     for i in range(m):
  __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 7; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_1);
  __Pyx_INCREF(__pyx_v_m);
  __Pyx_GIVEREF(__pyx_v_m);
  PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_v_m);
  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_1, NULL); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 7; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_2);
  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_6 = 0;
    __pyx_t_7 = NULL;
  } else {
    __pyx_t_6 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 7; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __Pyx_GOTREF(__pyx_t_1);
    __pyx_t_7 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_7)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 7; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  }
  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
  for (;;) {
    if (likely(!__pyx_t_7)) {
      if (likely(PyList_CheckExact(__pyx_t_1))) {
        if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_1)) break;
        #if CYTHON_COMPILING_IN_CPYTHON
        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 7; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        #else
        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 7; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        __Pyx_GOTREF(__pyx_t_2);
        #endif
      } else {
        if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
        #if CYTHON_COMPILING_IN_CPYTHON
        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 7; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        #else
        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 7; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        __Pyx_GOTREF(__pyx_t_2);
        #endif
      }
    } else {
      __pyx_t_2 = __pyx_t_7(__pyx_t_1);
      if (unlikely(!__pyx_t_2)) {
        PyObject* exc_type = PyErr_Occurred();
        if (exc_type) {
          if (likely(exc_type == PyExc_StopIteration || PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
          else {__pyx_filename = __pyx_f[0]; __pyx_lineno = 7; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        }
        break;
      }
      __Pyx_GOTREF(__pyx_t_2);
    }
    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_2);
    __pyx_t_2 = 0;
/* … */
  }
  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+08:         for j in range(p):
    __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 8; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __Pyx_GOTREF(__pyx_t_2);
    __Pyx_INCREF(__pyx_v_p);
    __Pyx_GIVEREF(__pyx_v_p);
    PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_p);
    __pyx_t_3 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_2, NULL); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 8; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    __Pyx_GOTREF(__pyx_t_3);
    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
    if (likely(PyList_CheckExact(__pyx_t_3)) || PyTuple_CheckExact(__pyx_t_3)) {
      __pyx_t_2 = __pyx_t_3; __Pyx_INCREF(__pyx_t_2); __pyx_t_8 = 0;
      __pyx_t_9 = NULL;
    } else {
      __pyx_t_8 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 8; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
      __Pyx_GOTREF(__pyx_t_2);
      __pyx_t_9 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_9)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 8; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
    }
    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
    for (;;) {
      if (likely(!__pyx_t_9)) {
        if (likely(PyList_CheckExact(__pyx_t_2))) {
          if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_2)) break;
          #if CYTHON_COMPILING_IN_CPYTHON
          __pyx_t_3 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_8); __Pyx_INCREF(__pyx_t_3); __pyx_t_8++; if (unlikely(0 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 8; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
          #else
          __pyx_t_3 = PySequence_ITEM(__pyx_t_2, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 8; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
          __Pyx_GOTREF(__pyx_t_3);
          #endif
        } else {
          if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
          #if CYTHON_COMPILING_IN_CPYTHON
          __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_8); __Pyx_INCREF(__pyx_t_3); __pyx_t_8++; if (unlikely(0 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 8; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
          #else
          __pyx_t_3 = PySequence_ITEM(__pyx_t_2, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 8; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
          __Pyx_GOTREF(__pyx_t_3);
          #endif
        }
      } else {
        __pyx_t_3 = __pyx_t_9(__pyx_t_2);
        if (unlikely(!__pyx_t_3)) {
          PyObject* exc_type = PyErr_Occurred();
          if (exc_type) {
            if (likely(exc_type == PyExc_StopIteration || PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
            else {__pyx_filename = __pyx_f[0]; __pyx_lineno = 8; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
          }
          break;
        }
        __Pyx_GOTREF(__pyx_t_3);
      }
      __Pyx_XDECREF_SET(__pyx_v_j, __pyx_t_3);
      __pyx_t_3 = 0;
/* … */
    }
    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+09:             res[i,j] = 0
      __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 9; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
      __Pyx_GOTREF(__pyx_t_3);
      __Pyx_INCREF(__pyx_v_i);
      __Pyx_GIVEREF(__pyx_v_i);
      PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_i);
      __Pyx_INCREF(__pyx_v_j);
      __Pyx_GIVEREF(__pyx_v_j);
      PyTuple_SET_ITEM(__pyx_t_3, 1, __pyx_v_j);
      if (unlikely(PyObject_SetItem(__pyx_v_res, __pyx_t_3, __pyx_int_0) < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 9; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+10:             for k in range(n):
      __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 10; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
      __Pyx_GOTREF(__pyx_t_3);
      __Pyx_INCREF(__pyx_v_n);
      __Pyx_GIVEREF(__pyx_v_n);
      PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_n);
      __pyx_t_4 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_3, NULL); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 10; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
      __Pyx_GOTREF(__pyx_t_4);
      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
      if (likely(PyList_CheckExact(__pyx_t_4)) || PyTuple_CheckExact(__pyx_t_4)) {
        __pyx_t_3 = __pyx_t_4; __Pyx_INCREF(__pyx_t_3); __pyx_t_10 = 0;
        __pyx_t_11 = NULL;
      } else {
        __pyx_t_10 = -1; __pyx_t_3 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 10; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        __Pyx_GOTREF(__pyx_t_3);
        __pyx_t_11 = Py_TYPE(__pyx_t_3)->tp_iternext; if (unlikely(!__pyx_t_11)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 10; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
      }
      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
      for (;;) {
        if (likely(!__pyx_t_11)) {
          if (likely(PyList_CheckExact(__pyx_t_3))) {
            if (__pyx_t_10 >= PyList_GET_SIZE(__pyx_t_3)) break;
            #if CYTHON_COMPILING_IN_CPYTHON
            __pyx_t_4 = PyList_GET_ITEM(__pyx_t_3, __pyx_t_10); __Pyx_INCREF(__pyx_t_4); __pyx_t_10++; if (unlikely(0 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 10; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
            #else
            __pyx_t_4 = PySequence_ITEM(__pyx_t_3, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 10; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
            __Pyx_GOTREF(__pyx_t_4);
            #endif
          } else {
            if (__pyx_t_10 >= PyTuple_GET_SIZE(__pyx_t_3)) break;
            #if CYTHON_COMPILING_IN_CPYTHON
            __pyx_t_4 = PyTuple_GET_ITEM(__pyx_t_3, __pyx_t_10); __Pyx_INCREF(__pyx_t_4); __pyx_t_10++; if (unlikely(0 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 10; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
            #else
            __pyx_t_4 = PySequence_ITEM(__pyx_t_3, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 10; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
            __Pyx_GOTREF(__pyx_t_4);
            #endif
          }
        } else {
          __pyx_t_4 = __pyx_t_11(__pyx_t_3);
          if (unlikely(!__pyx_t_4)) {
            PyObject* exc_type = PyErr_Occurred();
            if (exc_type) {
              if (likely(exc_type == PyExc_StopIteration || PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
              else {__pyx_filename = __pyx_f[0]; __pyx_lineno = 10; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
            }
            break;
          }
          __Pyx_GOTREF(__pyx_t_4);
        }
        __Pyx_XDECREF_SET(__pyx_v_k, __pyx_t_4);
        __pyx_t_4 = 0;
/* … */
      }
      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+11:                 res[i,j] += u[i,k] * v[k,j]
        __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 11; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        __Pyx_GOTREF(__pyx_t_4);
        __Pyx_INCREF(__pyx_v_i);
        __Pyx_GIVEREF(__pyx_v_i);
        PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_i);
        __Pyx_INCREF(__pyx_v_j);
        __Pyx_GIVEREF(__pyx_v_j);
        PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_v_j);
        __pyx_t_12 = PyObject_GetItem(__pyx_v_res, __pyx_t_4); if (unlikely(__pyx_t_12 == NULL)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 11; __pyx_clineno = __LINE__; goto __pyx_L1_error;};
        __Pyx_GOTREF(__pyx_t_12);
        __pyx_t_13 = PyTuple_New(2); if (unlikely(!__pyx_t_13)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 11; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        __Pyx_GOTREF(__pyx_t_13);
        __Pyx_INCREF(__pyx_v_i);
        __Pyx_GIVEREF(__pyx_v_i);
        PyTuple_SET_ITEM(__pyx_t_13, 0, __pyx_v_i);
        __Pyx_INCREF(__pyx_v_k);
        __Pyx_GIVEREF(__pyx_v_k);
        PyTuple_SET_ITEM(__pyx_t_13, 1, __pyx_v_k);
        __pyx_t_14 = PyObject_GetItem(__pyx_v_u, __pyx_t_13); if (unlikely(__pyx_t_14 == NULL)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 11; __pyx_clineno = __LINE__; goto __pyx_L1_error;};
        __Pyx_GOTREF(__pyx_t_14);
        __Pyx_DECREF(__pyx_t_13); __pyx_t_13 = 0;
        __pyx_t_13 = PyTuple_New(2); if (unlikely(!__pyx_t_13)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 11; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        __Pyx_GOTREF(__pyx_t_13);
        __Pyx_INCREF(__pyx_v_k);
        __Pyx_GIVEREF(__pyx_v_k);
        PyTuple_SET_ITEM(__pyx_t_13, 0, __pyx_v_k);
        __Pyx_INCREF(__pyx_v_j);
        __Pyx_GIVEREF(__pyx_v_j);
        PyTuple_SET_ITEM(__pyx_t_13, 1, __pyx_v_j);
        __pyx_t_15 = PyObject_GetItem(__pyx_v_v, __pyx_t_13); if (unlikely(__pyx_t_15 == NULL)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 11; __pyx_clineno = __LINE__; goto __pyx_L1_error;};
        __Pyx_GOTREF(__pyx_t_15);
        __Pyx_DECREF(__pyx_t_13); __pyx_t_13 = 0;
        __pyx_t_13 = PyNumber_Multiply(__pyx_t_14, __pyx_t_15); if (unlikely(!__pyx_t_13)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 11; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        __Pyx_GOTREF(__pyx_t_13);
        __Pyx_DECREF(__pyx_t_14); __pyx_t_14 = 0;
        __Pyx_DECREF(__pyx_t_15); __pyx_t_15 = 0;
        __pyx_t_15 = PyNumber_InPlaceAdd(__pyx_t_12, __pyx_t_13); if (unlikely(!__pyx_t_15)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 11; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        __Pyx_GOTREF(__pyx_t_15);
        __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
        __Pyx_DECREF(__pyx_t_13); __pyx_t_13 = 0;
        if (unlikely(PyObject_SetItem(__pyx_v_res, __pyx_t_4, __pyx_t_15) < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 11; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
        __Pyx_DECREF(__pyx_t_15); __pyx_t_15 = 0;
        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+12:     return res
  __Pyx_XDECREF(__pyx_r);
  __Pyx_INCREF(__pyx_v_res);
  __pyx_r = __pyx_v_res;
  goto __pyx_L0;

Using Cython cdefs and directives


In [11]:
%%cython -a

import cython

@cython.boundscheck(False)
@cython.wraparound(False)
def matrix_multiply1(double[:,:] u, double[:, :] v, double[:, :] res):
    cdef int i, j, k
    cdef int m, n, p

    m = u.shape[0]
    n = u.shape[1]
    p = v.shape[1]

    with cython.nogil:
        for i in range(m):
            for j in range(p):
                res[i,j] = 0
                for k in range(n):
                    res[i,j] += u[i,k] * v[k,j]


Out[11]:
Cython: _cython_magic_1d20d9c3ed3cbd86dd4904fc3a8795cf.pyx

Generated by Cython 0.23.5

Yellow lines hint at Python interaction.
Click on a line that starts with a "+" to see the C code that Cython generated for it.

 01: 
+02: import cython
  __pyx_t_1 = PyDict_New(); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_1);
  if (PyDict_SetItem(__pyx_d, __pyx_n_s_test, __pyx_t_1) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 03: 
 04: @cython.boundscheck(False)
 05: @cython.wraparound(False)
+06: def matrix_multiply1(double[:,:] u, double[:, :] v, double[:, :] res):
/* Python wrapper */
static PyObject *__pyx_pw_46_cython_magic_1d20d9c3ed3cbd86dd4904fc3a8795cf_1matrix_multiply1(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
static PyMethodDef __pyx_mdef_46_cython_magic_1d20d9c3ed3cbd86dd4904fc3a8795cf_1matrix_multiply1 = {"matrix_multiply1", (PyCFunction)__pyx_pw_46_cython_magic_1d20d9c3ed3cbd86dd4904fc3a8795cf_1matrix_multiply1, METH_VARARGS|METH_KEYWORDS, 0};
static PyObject *__pyx_pw_46_cython_magic_1d20d9c3ed3cbd86dd4904fc3a8795cf_1matrix_multiply1(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
  __Pyx_memviewslice __pyx_v_u = { 0, 0, { 0 }, { 0 }, { 0 } };
  __Pyx_memviewslice __pyx_v_v = { 0, 0, { 0 }, { 0 }, { 0 } };
  __Pyx_memviewslice __pyx_v_res = { 0, 0, { 0 }, { 0 }, { 0 } };
  PyObject *__pyx_r = 0;
  __Pyx_RefNannyDeclarations
  __Pyx_RefNannySetupContext("matrix_multiply1 (wrapper)", 0);
  {
    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_u,&__pyx_n_s_v,&__pyx_n_s_res,0};
    PyObject* values[3] = {0,0,0};
    if (unlikely(__pyx_kwds)) {
      Py_ssize_t kw_args;
      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
      switch (pos_args) {
        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
        case  0: break;
        default: goto __pyx_L5_argtuple_error;
      }
      kw_args = PyDict_Size(__pyx_kwds);
      switch (pos_args) {
        case  0:
        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_u)) != 0)) kw_args--;
        else goto __pyx_L5_argtuple_error;
        case  1:
        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_v)) != 0)) kw_args--;
        else {
          __Pyx_RaiseArgtupleInvalid("matrix_multiply1", 1, 3, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
        }
        case  2:
        if (likely((values[2] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_res)) != 0)) kw_args--;
        else {
          __Pyx_RaiseArgtupleInvalid("matrix_multiply1", 1, 3, 3, 2); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
        }
      }
      if (unlikely(kw_args > 0)) {
        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "matrix_multiply1") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
      }
    } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
      goto __pyx_L5_argtuple_error;
    } else {
      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
    }
    __pyx_v_u = __Pyx_PyObject_to_MemoryviewSlice_dsds_double(values[0]); if (unlikely(!__pyx_v_u.memview)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
    __pyx_v_v = __Pyx_PyObject_to_MemoryviewSlice_dsds_double(values[1]); if (unlikely(!__pyx_v_v.memview)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
    __pyx_v_res = __Pyx_PyObject_to_MemoryviewSlice_dsds_double(values[2]); if (unlikely(!__pyx_v_res.memview)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
  }
  goto __pyx_L4_argument_unpacking_done;
  __pyx_L5_argtuple_error:;
  __Pyx_RaiseArgtupleInvalid("matrix_multiply1", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
  __pyx_L3_error:;
  __Pyx_AddTraceback("_cython_magic_1d20d9c3ed3cbd86dd4904fc3a8795cf.matrix_multiply1", __pyx_clineno, __pyx_lineno, __pyx_filename);
  __Pyx_RefNannyFinishContext();
  return NULL;
  __pyx_L4_argument_unpacking_done:;
  __pyx_r = __pyx_pf_46_cython_magic_1d20d9c3ed3cbd86dd4904fc3a8795cf_matrix_multiply1(__pyx_self, __pyx_v_u, __pyx_v_v, __pyx_v_res);
  int __pyx_lineno = 0;
  const char *__pyx_filename = NULL;
  int __pyx_clineno = 0;

  /* function exit code */
  __Pyx_RefNannyFinishContext();
  return __pyx_r;
}

static PyObject *__pyx_pf_46_cython_magic_1d20d9c3ed3cbd86dd4904fc3a8795cf_matrix_multiply1(CYTHON_UNUSED PyObject *__pyx_self, __Pyx_memviewslice __pyx_v_u, __Pyx_memviewslice __pyx_v_v, __Pyx_memviewslice __pyx_v_res) {
  int __pyx_v_i;
  int __pyx_v_j;
  int __pyx_v_k;
  int __pyx_v_m;
  int __pyx_v_n;
  int __pyx_v_p;
  PyObject *__pyx_r = NULL;
  __Pyx_RefNannyDeclarations
  __Pyx_RefNannySetupContext("matrix_multiply1", 0);
/* … */
  /* function exit code */
  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
  __PYX_XDEC_MEMVIEW(&__pyx_v_u, 1);
  __PYX_XDEC_MEMVIEW(&__pyx_v_v, 1);
  __PYX_XDEC_MEMVIEW(&__pyx_v_res, 1);
  __Pyx_XGIVEREF(__pyx_r);
  __Pyx_RefNannyFinishContext();
  return __pyx_r;
}
/* … */
  __pyx_tuple__14 = PyTuple_Pack(9, __pyx_n_s_u, __pyx_n_s_v, __pyx_n_s_res, __pyx_n_s_i, __pyx_n_s_j, __pyx_n_s_k, __pyx_n_s_m, __pyx_n_s_n, __pyx_n_s_p); if (unlikely(!__pyx_tuple__14)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_tuple__14);
  __Pyx_GIVEREF(__pyx_tuple__14);
/* … */
  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_46_cython_magic_1d20d9c3ed3cbd86dd4904fc3a8795cf_1matrix_multiply1, NULL, __pyx_n_s_cython_magic_1d20d9c3ed3cbd86dd); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_1);
  if (PyDict_SetItem(__pyx_d, __pyx_n_s_matrix_multiply1, __pyx_t_1) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
  __pyx_codeobj__15 = (PyObject*)__Pyx_PyCode_New(3, 0, 9, 0, 0, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__14, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_home_jovyan_cache_ipython_cytho, __pyx_n_s_matrix_multiply1, 6, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__15)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 6; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
 07:     cdef int i, j, k
 08:     cdef int m, n, p
 09: 
+10:     m = u.shape[0]
  __pyx_v_m = (__pyx_v_u.shape[0]);
+11:     n = u.shape[1]
  __pyx_v_n = (__pyx_v_u.shape[1]);
+12:     p = v.shape[1]
  __pyx_v_p = (__pyx_v_v.shape[1]);
 13: 
+14:     with cython.nogil:
  {
      #ifdef WITH_THREAD
      PyThreadState *_save;
      Py_UNBLOCK_THREADS
      #endif
      /*try:*/ {
/* … */
      /*finally:*/ {
        /*normal exit:*/{
          #ifdef WITH_THREAD
          Py_BLOCK_THREADS
          #endif
          goto __pyx_L5;
        }
        __pyx_L5:;
      }
  }
+15:         for i in range(m):
        __pyx_t_1 = __pyx_v_m;
        for (__pyx_t_2 = 0; __pyx_t_2 < __pyx_t_1; __pyx_t_2+=1) {
          __pyx_v_i = __pyx_t_2;
+16:             for j in range(p):
          __pyx_t_3 = __pyx_v_p;
          for (__pyx_t_4 = 0; __pyx_t_4 < __pyx_t_3; __pyx_t_4+=1) {
            __pyx_v_j = __pyx_t_4;
+17:                 res[i,j] = 0
            __pyx_t_5 = __pyx_v_i;
            __pyx_t_6 = __pyx_v_j;
            *((double *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_res.data + __pyx_t_5 * __pyx_v_res.strides[0]) ) + __pyx_t_6 * __pyx_v_res.strides[1]) )) = 0.0;
+18:                 for k in range(n):
            __pyx_t_7 = __pyx_v_n;
            for (__pyx_t_8 = 0; __pyx_t_8 < __pyx_t_7; __pyx_t_8+=1) {
              __pyx_v_k = __pyx_t_8;
+19:                     res[i,j] += u[i,k] * v[k,j]
              __pyx_t_9 = __pyx_v_i;
              __pyx_t_10 = __pyx_v_k;
              __pyx_t_11 = __pyx_v_k;
              __pyx_t_12 = __pyx_v_j;
              __pyx_t_13 = __pyx_v_i;
              __pyx_t_14 = __pyx_v_j;
              *((double *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_res.data + __pyx_t_13 * __pyx_v_res.strides[0]) ) + __pyx_t_14 * __pyx_v_res.strides[1]) )) += ((*((double *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_u.data + __pyx_t_9 * __pyx_v_u.strides[0]) ) + __pyx_t_10 * __pyx_v_u.strides[1]) ))) * (*((double *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_v.data + __pyx_t_11 * __pyx_v_v.strides[0]) ) + __pyx_t_12 * __pyx_v_v.strides[1]) ))));
            }
          }
        }
      }

In [12]:
res = np.zeros((u.shape[0], v.shape[1]))
%timeit -r3 -n3 matrix_multiply1(u, v, res)


3 loops, best of 3: 12.4 µs per loop

Parallel execution with Cython

Will not work unless OpenMP is installed.


In [13]:
%%cython --compile-args=-fopenmp --link-args=-fopenmp --force

import cython
from cython.parallel import parallel, prange

@cython.boundscheck(False)
@cython.wraparound(False)
def matrix_multiply2(double[:,:] u, double[:, :] v, double[:, :] res):
    cdef int i, j, k
    cdef int m, n, p

    m = u.shape[0]
    n = u.shape[1]
    p = v.shape[1]

    with cython.nogil, parallel():
        for i in prange(m):
            for j in prange(p):
                res[i,j] = 0
                for k in range(n):
                    res[i,j] += u[i,k] * v[k,j]

In [14]:
res = np.zeros((u.shape[0], v.shape[1]))
%timeit -r3 -n3 matrix_multiply2(u, v, res)


The slowest run took 112.92 times longer than the fastest. This could mean that an intermediate result is being cached.
3 loops, best of 3: 90.2 µs per loop

In [15]:
import time

In [16]:
# color function for point at (x, y)
def mandel(x, y, max_iters):
    c = complex(x, y)
    z = 0.0j
    for i in range(max_iters):
        z = z*z + c
        if z.real*z.real + z.imag*z.imag >= 4:
            return i
    return max_iters

def create_fractal(xmin, xmax, ymin, ymax, image, iters):
    height, width = image.shape
    
    pixel_size_x = (xmax - xmin)/width
    pixel_size_y = (ymax - ymin)/height
        
    for x in range(width):
        real = xmin + x*pixel_size_x
        for y in range(height):
            imag = ymin + y*pixel_size_y
            color = mandel(real, imag, iters)
            image[y, x]  = color

In [17]:
gimage = np.zeros((1024, 1536), dtype=np.uint8)
xmin, xmax, ymin, ymax = np.array([-2.0, 1.0, -1.0, 1.0]).astype('float32')
iters = 50

start = time.clock()
create_fractal(xmin, xmax, ymin, ymax, gimage, iters)
dt = time.clock() - start

print("Mandelbrot created on CPU in %f s" % dt)
plt.grid(False)
plt.imshow(gimage, cmap='jet')
pass


Mandelbrot created on CPU in 15.622738 s

In [18]:
%%cython -a

cimport cython

cdef extern from "complex.h":
    double cabs(double complex)

# color function for point at (x, y)
cdef unsigned char mandel_cython(double x, double y, int max_iters):
    cdef double complex c, z

    c = x + y*1j
    z = 0.0j
    for i in range(max_iters):
        z = z*z + c
        if cabs(z) >= 2:
            return i
    return max_iters

@cython.cdivision(True)
def create_fractal_cython(double xmin, double xmax, double ymin, double ymax, unsigned char[:, :] image, int iters):

    cdef int x, y
    cdef int height, width
    cdef double pixel_size_x, pixel_size_y
    cdef double real, imag
    cdef unsigned char color

    height = image.shape[0]
    width = image.shape[1]
    
    pixel_size_x = (xmax - xmin)/width
    pixel_size_y = (ymax - ymin)/height
        
    for x in range(width):
        real = xmin + x*pixel_size_x
        for y in range(height):
            imag = ymin + y*pixel_size_y
            color = mandel_cython(real, imag, iters)
            image[y, x]  = color


Out[18]:
Cython: _cython_magic_176141650497de6acdeb56d4e03ce71b.pyx

Generated by Cython 0.23.5

Yellow lines hint at Python interaction.
Click on a line that starts with a "+" to see the C code that Cython generated for it.

 01: 
+02: cimport cython
  __pyx_t_1 = PyDict_New(); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_1);
  if (PyDict_SetItem(__pyx_d, __pyx_n_s_test, __pyx_t_1) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 03: 
 04: cdef extern from "complex.h":
 05:     double cabs(double complex)
 06: 
 07: # color function for point at (x, y)
+08: cdef unsigned char mandel_cython(double x, double y, int max_iters):
static unsigned char __pyx_f_46_cython_magic_176141650497de6acdeb56d4e03ce71b_mandel_cython(double __pyx_v_x, double __pyx_v_y, int __pyx_v_max_iters) {
  __pyx_t_double_complex __pyx_v_c;
  __pyx_t_double_complex __pyx_v_z;
  int __pyx_v_i;
  unsigned char __pyx_r;
  __Pyx_RefNannyDeclarations
  __Pyx_RefNannySetupContext("mandel_cython", 0);
/* … */
  /* function exit code */
  __pyx_L0:;
  __Pyx_RefNannyFinishContext();
  return __pyx_r;
}
 09:     cdef double complex c, z
 10: 
+11:     c = x + y*1j
  __pyx_v_c = __Pyx_c_sum(__pyx_t_double_complex_from_parts(__pyx_v_x, 0), __Pyx_c_prod(__pyx_t_double_complex_from_parts(__pyx_v_y, 0), __pyx_t_double_complex_from_parts(0, 1.0)));
+12:     z = 0.0j
  __pyx_v_z = __pyx_t_double_complex_from_parts(0, 0.0);
+13:     for i in range(max_iters):
  __pyx_t_1 = __pyx_v_max_iters;
  for (__pyx_t_2 = 0; __pyx_t_2 < __pyx_t_1; __pyx_t_2+=1) {
    __pyx_v_i = __pyx_t_2;
+14:         z = z*z + c
    __pyx_v_z = __Pyx_c_sum(__Pyx_c_prod(__pyx_v_z, __pyx_v_z), __pyx_v_c);
+15:         if cabs(z) >= 2:
    __pyx_t_3 = ((cabs(__pyx_v_z) >= 2.0) != 0);
    if (__pyx_t_3) {
/* … */
    }
  }
+16:             return i
      __pyx_r = __pyx_v_i;
      goto __pyx_L0;
+17:     return max_iters
  __pyx_r = __pyx_v_max_iters;
  goto __pyx_L0;
 18: 
 19: @cython.cdivision(True)
+20: def create_fractal_cython(double xmin, double xmax, double ymin, double ymax, unsigned char[:, :] image, int iters):
/* Python wrapper */
static PyObject *__pyx_pw_46_cython_magic_176141650497de6acdeb56d4e03ce71b_1create_fractal_cython(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
static PyMethodDef __pyx_mdef_46_cython_magic_176141650497de6acdeb56d4e03ce71b_1create_fractal_cython = {"create_fractal_cython", (PyCFunction)__pyx_pw_46_cython_magic_176141650497de6acdeb56d4e03ce71b_1create_fractal_cython, METH_VARARGS|METH_KEYWORDS, 0};
static PyObject *__pyx_pw_46_cython_magic_176141650497de6acdeb56d4e03ce71b_1create_fractal_cython(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
  double __pyx_v_xmin;
  double __pyx_v_xmax;
  double __pyx_v_ymin;
  double __pyx_v_ymax;
  __Pyx_memviewslice __pyx_v_image = { 0, 0, { 0 }, { 0 }, { 0 } };
  int __pyx_v_iters;
  PyObject *__pyx_r = 0;
  __Pyx_RefNannyDeclarations
  __Pyx_RefNannySetupContext("create_fractal_cython (wrapper)", 0);
  {
    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_xmin,&__pyx_n_s_xmax,&__pyx_n_s_ymin,&__pyx_n_s_ymax,&__pyx_n_s_image,&__pyx_n_s_iters,0};
    PyObject* values[6] = {0,0,0,0,0,0};
    if (unlikely(__pyx_kwds)) {
      Py_ssize_t kw_args;
      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
      switch (pos_args) {
        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
        case  0: break;
        default: goto __pyx_L5_argtuple_error;
      }
      kw_args = PyDict_Size(__pyx_kwds);
      switch (pos_args) {
        case  0:
        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_xmin)) != 0)) kw_args--;
        else goto __pyx_L5_argtuple_error;
        case  1:
        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_xmax)) != 0)) kw_args--;
        else {
          __Pyx_RaiseArgtupleInvalid("create_fractal_cython", 1, 6, 6, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
        }
        case  2:
        if (likely((values[2] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_ymin)) != 0)) kw_args--;
        else {
          __Pyx_RaiseArgtupleInvalid("create_fractal_cython", 1, 6, 6, 2); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
        }
        case  3:
        if (likely((values[3] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_ymax)) != 0)) kw_args--;
        else {
          __Pyx_RaiseArgtupleInvalid("create_fractal_cython", 1, 6, 6, 3); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
        }
        case  4:
        if (likely((values[4] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_image)) != 0)) kw_args--;
        else {
          __Pyx_RaiseArgtupleInvalid("create_fractal_cython", 1, 6, 6, 4); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
        }
        case  5:
        if (likely((values[5] = PyDict_GetItem(__pyx_kwds, __pyx_n_s_iters)) != 0)) kw_args--;
        else {
          __Pyx_RaiseArgtupleInvalid("create_fractal_cython", 1, 6, 6, 5); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
        }
      }
      if (unlikely(kw_args > 0)) {
        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "create_fractal_cython") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
      }
    } else if (PyTuple_GET_SIZE(__pyx_args) != 6) {
      goto __pyx_L5_argtuple_error;
    } else {
      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
      values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
      values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
    }
    __pyx_v_xmin = __pyx_PyFloat_AsDouble(values[0]); if (unlikely((__pyx_v_xmin == (double)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
    __pyx_v_xmax = __pyx_PyFloat_AsDouble(values[1]); if (unlikely((__pyx_v_xmax == (double)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
    __pyx_v_ymin = __pyx_PyFloat_AsDouble(values[2]); if (unlikely((__pyx_v_ymin == (double)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
    __pyx_v_ymax = __pyx_PyFloat_AsDouble(values[3]); if (unlikely((__pyx_v_ymax == (double)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
    __pyx_v_image = __Pyx_PyObject_to_MemoryviewSlice_dsds_unsigned_char(values[4]); if (unlikely(!__pyx_v_image.memview)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
    __pyx_v_iters = __Pyx_PyInt_As_int(values[5]); if (unlikely((__pyx_v_iters == (int)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
  }
  goto __pyx_L4_argument_unpacking_done;
  __pyx_L5_argtuple_error:;
  __Pyx_RaiseArgtupleInvalid("create_fractal_cython", 1, 6, 6, PyTuple_GET_SIZE(__pyx_args)); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
  __pyx_L3_error:;
  __Pyx_AddTraceback("_cython_magic_176141650497de6acdeb56d4e03ce71b.create_fractal_cython", __pyx_clineno, __pyx_lineno, __pyx_filename);
  __Pyx_RefNannyFinishContext();
  return NULL;
  __pyx_L4_argument_unpacking_done:;
  __pyx_r = __pyx_pf_46_cython_magic_176141650497de6acdeb56d4e03ce71b_create_fractal_cython(__pyx_self, __pyx_v_xmin, __pyx_v_xmax, __pyx_v_ymin, __pyx_v_ymax, __pyx_v_image, __pyx_v_iters);
  int __pyx_lineno = 0;
  const char *__pyx_filename = NULL;
  int __pyx_clineno = 0;

  /* function exit code */
  __Pyx_RefNannyFinishContext();
  return __pyx_r;
}

static PyObject *__pyx_pf_46_cython_magic_176141650497de6acdeb56d4e03ce71b_create_fractal_cython(CYTHON_UNUSED PyObject *__pyx_self, double __pyx_v_xmin, double __pyx_v_xmax, double __pyx_v_ymin, double __pyx_v_ymax, __Pyx_memviewslice __pyx_v_image, int __pyx_v_iters) {
  int __pyx_v_x;
  int __pyx_v_y;
  int __pyx_v_height;
  int __pyx_v_width;
  double __pyx_v_pixel_size_x;
  double __pyx_v_pixel_size_y;
  double __pyx_v_real;
  double __pyx_v_imag;
  unsigned char __pyx_v_color;
  PyObject *__pyx_r = NULL;
  __Pyx_RefNannyDeclarations
  __Pyx_RefNannySetupContext("create_fractal_cython", 0);
/* … */
  /* function exit code */
  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
  goto __pyx_L0;
  __pyx_L1_error:;
  __Pyx_AddTraceback("_cython_magic_176141650497de6acdeb56d4e03ce71b.create_fractal_cython", __pyx_clineno, __pyx_lineno, __pyx_filename);
  __pyx_r = NULL;
  __pyx_L0:;
  __PYX_XDEC_MEMVIEW(&__pyx_v_image, 1);
  __Pyx_XGIVEREF(__pyx_r);
  __Pyx_RefNannyFinishContext();
  return __pyx_r;
}
/* … */
  __pyx_tuple__14 = PyTuple_Pack(15, __pyx_n_s_xmin, __pyx_n_s_xmax, __pyx_n_s_ymin, __pyx_n_s_ymax, __pyx_n_s_image, __pyx_n_s_iters, __pyx_n_s_x, __pyx_n_s_y, __pyx_n_s_height, __pyx_n_s_width, __pyx_n_s_pixel_size_x, __pyx_n_s_pixel_size_y, __pyx_n_s_real, __pyx_n_s_imag, __pyx_n_s_color); if (unlikely(!__pyx_tuple__14)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_tuple__14);
  __Pyx_GIVEREF(__pyx_tuple__14);
/* … */
  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_46_cython_magic_176141650497de6acdeb56d4e03ce71b_1create_fractal_cython, NULL, __pyx_n_s_cython_magic_176141650497de6acd); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_GOTREF(__pyx_t_1);
  if (PyDict_SetItem(__pyx_d, __pyx_n_s_create_fractal_cython, __pyx_t_1) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
  __pyx_codeobj__15 = (PyObject*)__Pyx_PyCode_New(6, 0, 15, 0, 0, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__14, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_home_jovyan_cache_ipython_cytho, __pyx_n_s_create_fractal_cython, 20, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__15)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 20; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
 21: 
 22:     cdef int x, y
 23:     cdef int height, width
 24:     cdef double pixel_size_x, pixel_size_y
 25:     cdef double real, imag
 26:     cdef unsigned char color
 27: 
+28:     height = image.shape[0]
  __pyx_v_height = (__pyx_v_image.shape[0]);
+29:     width = image.shape[1]
  __pyx_v_width = (__pyx_v_image.shape[1]);
 30: 
+31:     pixel_size_x = (xmax - xmin)/width
  __pyx_v_pixel_size_x = ((__pyx_v_xmax - __pyx_v_xmin) / __pyx_v_width);
+32:     pixel_size_y = (ymax - ymin)/height
  __pyx_v_pixel_size_y = ((__pyx_v_ymax - __pyx_v_ymin) / __pyx_v_height);
 33: 
+34:     for x in range(width):
  __pyx_t_1 = __pyx_v_width;
  for (__pyx_t_2 = 0; __pyx_t_2 < __pyx_t_1; __pyx_t_2+=1) {
    __pyx_v_x = __pyx_t_2;
+35:         real = xmin + x*pixel_size_x
    __pyx_v_real = (__pyx_v_xmin + (__pyx_v_x * __pyx_v_pixel_size_x));
+36:         for y in range(height):
    __pyx_t_3 = __pyx_v_height;
    for (__pyx_t_4 = 0; __pyx_t_4 < __pyx_t_3; __pyx_t_4+=1) {
      __pyx_v_y = __pyx_t_4;
+37:             imag = ymin + y*pixel_size_y
      __pyx_v_imag = (__pyx_v_ymin + (__pyx_v_y * __pyx_v_pixel_size_y));
+38:             color = mandel_cython(real, imag, iters)
      __pyx_v_color = __pyx_f_46_cython_magic_176141650497de6acdeb56d4e03ce71b_mandel_cython(__pyx_v_real, __pyx_v_imag, __pyx_v_iters);
+39:             image[y, x]  = color
      __pyx_t_5 = __pyx_v_y;
      __pyx_t_6 = __pyx_v_x;
      __pyx_t_7 = -1;
      if (__pyx_t_5 < 0) {
        __pyx_t_5 += __pyx_v_image.shape[0];
        if (unlikely(__pyx_t_5 < 0)) __pyx_t_7 = 0;
      } else if (unlikely(__pyx_t_5 >= __pyx_v_image.shape[0])) __pyx_t_7 = 0;
      if (__pyx_t_6 < 0) {
        __pyx_t_6 += __pyx_v_image.shape[1];
        if (unlikely(__pyx_t_6 < 0)) __pyx_t_7 = 1;
      } else if (unlikely(__pyx_t_6 >= __pyx_v_image.shape[1])) __pyx_t_7 = 1;
      if (unlikely(__pyx_t_7 != -1)) {
        __Pyx_RaiseBufferIndexError(__pyx_t_7);
        {__pyx_filename = __pyx_f[0]; __pyx_lineno = 39; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
      }
      *((unsigned char *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_image.data + __pyx_t_5 * __pyx_v_image.strides[0]) ) + __pyx_t_6 * __pyx_v_image.strides[1]) )) = __pyx_v_color;
    }
  }

In [19]:
gimage = np.zeros((1024, 1536), dtype=np.uint8)
xmin, xmax, ymin, ymax = np.array([-2.0, 1.0, -1.0, 1.0]).astype('float32')
iters = 50

start = time.clock()
create_fractal_cython(xmin, xmax, ymin, ymax, gimage, iters)
dt = time.clock() - start

print("Mandelbrot created on CPU in %f s" % dt)
plt.grid(False)
plt.imshow(gimage, cmap='jet')
pass


Mandelbrot created on CPU in 0.462727 s