//출처 : https://ko.wikipedia.org/wiki/OpenMP
//openMP를 사용하기 위해서 omp.h 헤더파일을 import
 #include <omp.h>
 #include <stdio.h>
 #include <stdlib.h>

 int main (int argc, char *argv[]) {
   int th_id, nthreads;
   //progma omp라는 문장을 추가하여 해당 이하 구문을 다수의 Thread를 만들어 실행시킴
   #pragma omp parallel private(th_id)
   {
     th_id = omp_get_thread_num();
     printf("Hello World : 스레드 %d\n", th_id);
     #pragma omp barrier
     if ( th_id == 0 ) {
       nthreads = omp_get_num_threads();
       printf("모두 %d 개의 스레드가 있습니다\n",nthreads);
     }
   }
   return EXIT_SUCCESS;
 }

//출처 : https://learn.microsoft.com/ko-kr/cpp/parallel/openmp/a-examples?view=msvc-170
#pragma omp parallel for
    for (i=1; i<n; i++)
        b[i] = (a[i] + a[i-1]) / 2.0;

#cython의 prange를 사용한 openmp 활용
from cython.parallel cimport prange
#cython코드에서 cython wrapper를 사용하기 위한 임포트
cimport cython
from libc.math cimport sin

import numpy as np

@cython.boundscheck(False)
@cython.wraparound(False)
def do_sine(double[:,:] input):
    cdef double[:,:] output = np.empty_like(input)
    cdef Py_ssize_t i, j

    for i in prange(input.shape[0], nogil=True):
        for j in range(input.shape[1]):
            output[i, j] = sin(input[i, j])
    return np.asarray(output)
    
"""
compile후 나오는 C언어 코드에서 병렬처리 된 부분
#pragma omp parallel
{
    #pragma omp for firstprivate(i) lastprivate(i) lastprivate(j)
    for (__pyx_t_8 = 0; __pyx_t_8 < __pyx_t_9; __pyx_t_8++){
        i = __pyx_t_8;
        /* body goes here */
    }
"""

%%cython --annotate
import numpy as np
cimport numpy as np
np.import_array()
DTYPE = np.int
ctypedef np.int_t DTYPE_t


from cython.parallel cimport prange
cimport cython
@cython.boundscheck(False)
@cython.wraparound(False)
def my_conv2d(np.ndarray[DTYPE_t, ndim = 2] mat, np.ndarray[DTYPE_t, ndim = 2] f):
    cdef int mat_h = mat.shape[1]
    cdef int mat_v = mat.shape[0]
    cdef int f_h = f.shape[1]
    cdef int f_v = f.shape[0]
    cdef DTYPE_t temp_val
    
    cdef int ret_mat_h = mat_h-f_h+1
    cdef int ret_mat_v = mat_v-f_v+1
    
    cdef int row, col, i_row, i_col, temp_row, temp_col;
    
    cdef np.ndarray[DTYPE_t, ndim = 2] ret = np.zeros([ret_mat_v,ret_mat_h], dtype = DTYPE)    
    for row in prange(ret_mat_v, nogil = True):
        for col in range(ret_mat_h):
            temp_val = 0
            for i_row in range(f_v):
                for i_col in range(f_h):                    
                    temp_row = row+i_row
                    temp_col = col+i_col                    
                    temp_val += (mat[temp_row, temp_col] * f[i_row, i_col])
            ret[row,col] = temp_val
            
    return ret
import time
import numpy as np
import copy
N = 1000
input = np.array([copy.deepcopy([np.random.randint(0, 10) for i in range(1000)]) for k in range(1000)])
filter = np.array([copy.deepcopy([np.random.randint(0, 10) for i in range(3)]) for k in range(3)])
st = time.perf_counter()
normal = my_conv2d(input, filter)
print(f"time cfunc : {time.perf_counter() - st}")

Error compiling Cython file:
------------------------------------------------------------
...
                for i_col in range(f_h):                    
                    temp_row = row+i_row
                    temp_col = col+i_col                    
                    #temp_val = temp_val + (mat[temp_row, temp_col] * f[i_row, i_col])
                    temp_val += (mat[temp_row, temp_col] * f[i_row, i_col])
            ret[row,col] = temp_val
                          ^
------------------------------------------------------------
ipython\cython\_cython_magic_066a197a2f983a6a606f576c0c7491ec.pyx:34:27: Cannot read reduction variable in loop body

#temp_val += (mat[temp_row, temp_col] * f[i_row, i_col]) -->
temp_val = temp_val + (mat[temp_row, temp_col] * f[i_row, i_col]) 

// for row in prange(ret_mat_v, nogile = True)의 CPython 변환 코드
{
      #ifdef WITH_THREAD
      PyThreadState *_save;
      Py_UNBLOCK_THREADS
      __Pyx_FastGIL_Remember();
      #endif
      /*try:*/ {
        __pyx_t_6 = __pyx_v_ret_mat_v;
        if ((1 == 0)) abort();
        {
            #if ((defined(__APPLE__) || defined(__OSX__)) && (defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && (__GNUC_MINOR__ > 95)))))
                #undef likely
                #undef unlikely
                #define likely(x)   (x)
                #define unlikely(x) (x)
            #endif
            __pyx_t_8 = (__pyx_t_6 - 0 + 1 - 1/abs(1)) / 1;
            if (__pyx_t_8 > 0)
            {
                #ifdef _OPENMP
                #pragma omp parallel
                #endif /* _OPENMP */
                {
                    #ifdef _OPENMP
                    #pragma omp for lastprivate(__pyx_v_col) lastprivate(__pyx_v_i_col) lastprivate(__pyx_v_i_row) firstprivate(__pyx_v_row) lastprivate(__pyx_v_row) lastprivate(__pyx_v_temp_col) lastprivate(__pyx_v_temp_row) lastprivate(__pyx_v_temp_val)
                    #endif /* _OPENMP */
                    for (__pyx_t_7 = 0; __pyx_t_7 < __pyx_t_8; __pyx_t_7++){
                        {
                            __pyx_v_row = (int)(0 + 1 * __pyx_t_7);
                            /* Initialize private variables to invalid values */
                            __pyx_v_col = ((int)0xbad0bad0);
                            __pyx_v_i_col = ((int)0xbad0bad0);
                            __pyx_v_i_row = ((int)0xbad0bad0);
                            __pyx_v_temp_col = ((int)0xbad0bad0);
                            __pyx_v_temp_row = ((int)0xbad0bad0);
                            __pyx_v_temp_val = ((__pyx_t_46_cython_magic_13dc7bc2b38a451e5c2b88564a61cb58_DTYPE_t)0xbad0bad0);
/* … */
      /*finally:*/ {
        /*normal exit:*/{
          #ifdef WITH_THREAD
          __Pyx_FastGIL_Forget();
          Py_BLOCK_THREADS
          #endif
          goto __pyx_L5;
        }
        __pyx_L5:;
      }
  }

#setup.py
from distutils.core import setup
from distutils.extension import Extension
import numpy as np
import sys
#compile에 사용된 linker, compile 매개변수 정의
#os 종류에 따라 명령어가 살짝 차이남
if sys.platform.startswith("win"):
    openmp_arg = '/openmp'
else:
    openmp_arg = '-fopenmp'

ext_modules = [Extension(
      "cy_conv2d_parallel",
      ["cy_conv2d_parallel.pyx"], 
      #compile간 사용할 매개변수 추가
        extra_compile_args=[openmp_arg],
        extra_link_args=[openmp_arg],
      )]

from Cython.Build import cythonize
setup(
    ext_modules=cythonize(
        ext_modules,
        compiler_directives={"language_level": "3"},
    ),
    include_dirs=[np.get_include()]
) 

import cy_conv2d
import cy_conv2d_parallel
import numpy as np
import copy
input = np.array([copy.deepcopy([np.random.randint(0, 10) for i in range(1000)]) for k in range(1000)])
filter = np.array([copy.deepcopy([np.random.randint(0, 10) for i in range(3)]) for k in range(3)])
st = time.perf_counter()
normal = cy_conv2d_parallel.my_conv2d(input, filter)
print(f"time cfunc with parallel : {time.perf_counter() - st}")

st = time.perf_counter()
normal = cy_conv2d.my_conv2d(input, filter)
print(f"time cfunc : {time.perf_counter() - st}")
'''
time cfunc with parallel : 0.01012540000010631
time cfunc : 0.17689960000279825
'''