admin管理员组文章数量:1310254
I am experimenting with the use of the ctypes package in Python. Currently attempting to pass a 3d NumPy array to a C function that takes a triple pointer as an argument and updates the values of the elements of the array. The C function is:
void effect_array(int ***ptr, int rows, int cols, int depth)
{
for(int i = 0; i < rows; i++)
{
for(int j = 0; j < cols; j++)
{
for(int k = 0; k < depth; k++)
{
ptr[i][j][k] *= 2;
}
}
}
}
Currently, I have tried:
import ctypes as ct
import numpy as np
arr = np.array([
[
[1, 2, 3, 4],
[5, 6, 7, 8],
[9, 10, 11, 12]
],
[
[13, 14, 15, 16],
[17, 18, 19, 20],
[21, 22, 23, 24]
]
])
arr_c = np.ascontiguousarray(arr)
lib = ct.CDLL("path/to/share_object.so")
_effect_array = lib.effect_array
_effect_array.argtypes = [ct.POINTER(ct.POINTER(ct.POINTER(ct.c_int))), ct.c_int, ct.c_int, ct.c_int]
_effect_array.restype = None
rows, cols, depth = arr.shape
t_ptr = (ct.POINTER(ct.POINTER(ct.c_int)) * rows)()
for i in range(rows):
t_ptr[i] = (ct.POINTER(ct.c_int) * cols)()
for j in range(cols):
t_ptr[i][j] = arr_c[i][j].ctypes.data_as(ct.POINTER(ct.c_int))
print("Original array =")
print(arr_c)
print()
_effect_array(t_ptr, rows, cols, depth)
print("Array after pass to function =")
print(arr_c)
This has resulted in an output of:
Original array =
[[[ 1 2 3 4]
[ 5 6 7 8]
[ 9 10 11 12]]
[[13 14 15 16]
[17 18 19 20]
[21 22 23 24]]]
Array after pass to function =
[[[ 2 4 3 4]
[10 12 7 8]
[18 20 11 12]]
[[26 28 15 16]
[34 36 19 20]
[42 44 23 24]]]
What I would like to happen would be:
Original array =
[[[ 1 2 3 4]
[ 5 6 7 8]
[ 9 10 11 12]]
[[13 14 15 16]
[17 18 19 20]
[21 22 23 24]]]
Array after pass to function =
[[[ 2 4 6 8]
[10 12 14 16]
[18 20 22 24]]
[[26 28 30 32]
[34 36 38 40]
[42 44 46 48]]]
I am not sure why the C function is not able to access the elements beyond the first two in a row. My understanding is that my current attempt is provided access to the start of the array and that it should be able to iterate through the rest of the array as can be done in this C example:
#include <stdio.h>
#include <stdlib.h>
void effect_array(int ***ptr, int rows, int cols, int depth)
{
for(int i = 0; i < rows; i++)
{
for(int j = 0; j < cols; j++)
{
for(int k = 0; k < depth; k++)
{
ptr[i][j][k] *= 2;
}
}
}
}
int main()
{
int arr[2][2][2] = {
{
{1,2},
{3,4}
},
{
{5,6},
{7,8}
}
};
int arr2[5];
int *p = arr2;
int ***ptr = (int ***)malloc(2 * sizeof(int **));
for(int i = 0; i < 2; i++)
{
ptr[i] = (int **)malloc(2 * sizeof(int *));
for(int j = 0; j < 2; j++)
{
ptr[i][j] = &arr[i][j][0];
}
}
printf("Print array before:\n");
for(int i = 0; i < 2; i++)
{
for(int j = 0; j < 2; j++)
{
for(int k = 0; k < 2; k++)
{
printf("%d ", arr[i][j][k]);
}
printf("\n");
}
printf("\n");
}
effect_array(ptr, 2, 2, 2);
printf("Print array after:\n");
for(int i = 0; i < 2; i++)
{
for(int j = 0; j < 2; j++)
{
for(int k = 0; k < 2; k++)
{
printf("%d ", arr[i][j][k]);
}
printf("\n");
}
printf("\n");
}
for(int i = 0; i < 2; i++) free(ptr[i]);
free(ptr);
return 0;
}
This is also based off the discussion on a previous question I asked, credit to @SR143 for the working C example.
Any and all help is greatly appreciated.
I am experimenting with the use of the ctypes package in Python. Currently attempting to pass a 3d NumPy array to a C function that takes a triple pointer as an argument and updates the values of the elements of the array. The C function is:
void effect_array(int ***ptr, int rows, int cols, int depth)
{
for(int i = 0; i < rows; i++)
{
for(int j = 0; j < cols; j++)
{
for(int k = 0; k < depth; k++)
{
ptr[i][j][k] *= 2;
}
}
}
}
Currently, I have tried:
import ctypes as ct
import numpy as np
arr = np.array([
[
[1, 2, 3, 4],
[5, 6, 7, 8],
[9, 10, 11, 12]
],
[
[13, 14, 15, 16],
[17, 18, 19, 20],
[21, 22, 23, 24]
]
])
arr_c = np.ascontiguousarray(arr)
lib = ct.CDLL("path/to/share_object.so")
_effect_array = lib.effect_array
_effect_array.argtypes = [ct.POINTER(ct.POINTER(ct.POINTER(ct.c_int))), ct.c_int, ct.c_int, ct.c_int]
_effect_array.restype = None
rows, cols, depth = arr.shape
t_ptr = (ct.POINTER(ct.POINTER(ct.c_int)) * rows)()
for i in range(rows):
t_ptr[i] = (ct.POINTER(ct.c_int) * cols)()
for j in range(cols):
t_ptr[i][j] = arr_c[i][j].ctypes.data_as(ct.POINTER(ct.c_int))
print("Original array =")
print(arr_c)
print()
_effect_array(t_ptr, rows, cols, depth)
print("Array after pass to function =")
print(arr_c)
This has resulted in an output of:
Original array =
[[[ 1 2 3 4]
[ 5 6 7 8]
[ 9 10 11 12]]
[[13 14 15 16]
[17 18 19 20]
[21 22 23 24]]]
Array after pass to function =
[[[ 2 4 3 4]
[10 12 7 8]
[18 20 11 12]]
[[26 28 15 16]
[34 36 19 20]
[42 44 23 24]]]
What I would like to happen would be:
Original array =
[[[ 1 2 3 4]
[ 5 6 7 8]
[ 9 10 11 12]]
[[13 14 15 16]
[17 18 19 20]
[21 22 23 24]]]
Array after pass to function =
[[[ 2 4 6 8]
[10 12 14 16]
[18 20 22 24]]
[[26 28 30 32]
[34 36 38 40]
[42 44 46 48]]]
I am not sure why the C function is not able to access the elements beyond the first two in a row. My understanding is that my current attempt is provided access to the start of the array and that it should be able to iterate through the rest of the array as can be done in this C example:
#include <stdio.h>
#include <stdlib.h>
void effect_array(int ***ptr, int rows, int cols, int depth)
{
for(int i = 0; i < rows; i++)
{
for(int j = 0; j < cols; j++)
{
for(int k = 0; k < depth; k++)
{
ptr[i][j][k] *= 2;
}
}
}
}
int main()
{
int arr[2][2][2] = {
{
{1,2},
{3,4}
},
{
{5,6},
{7,8}
}
};
int arr2[5];
int *p = arr2;
int ***ptr = (int ***)malloc(2 * sizeof(int **));
for(int i = 0; i < 2; i++)
{
ptr[i] = (int **)malloc(2 * sizeof(int *));
for(int j = 0; j < 2; j++)
{
ptr[i][j] = &arr[i][j][0];
}
}
printf("Print array before:\n");
for(int i = 0; i < 2; i++)
{
for(int j = 0; j < 2; j++)
{
for(int k = 0; k < 2; k++)
{
printf("%d ", arr[i][j][k]);
}
printf("\n");
}
printf("\n");
}
effect_array(ptr, 2, 2, 2);
printf("Print array after:\n");
for(int i = 0; i < 2; i++)
{
for(int j = 0; j < 2; j++)
{
for(int k = 0; k < 2; k++)
{
printf("%d ", arr[i][j][k]);
}
printf("\n");
}
printf("\n");
}
for(int i = 0; i < 2; i++) free(ptr[i]);
free(ptr);
return 0;
}
This is also based off the discussion on a previous question I asked, credit to @SR143 for the working C example.
Any and all help is greatly appreciated.
Share Improve this question edited Feb 2 at 16:31 jared 9,1463 gold badges15 silver badges43 bronze badges asked Feb 2 at 15:39 frankfrank 311 silver badge6 bronze badges 9 | Show 4 more comments1 Answer
Reset to default 2In your original Python code, the dtype of arr is np.int64. Add dtype=np.int32 to your np.array declaration and the code works.
However, be aware that you can pass a multidimensional numpy array directly to C code instead by using an int* and pointer math on the contiguous C array and be more efficient since it will save you the trouble of creating the pointer array. An np.ctypeslib.ndpointer can be used with .argtypes for better type-checking as well.
Example:
test.c (Windows example):
__declspec(dllexport)
void affect_array(int *ptr, int rows, int cols, int depth) {
for(int i = 0; i < rows; i++) {
for(int j = 0; j < cols; j++) {
for(int k = 0; k < depth; k++) {
ptr[i*cols*depth + j*depth + k] *= 2;
}
}
}
}
test.py
import ctypes as ct
import numpy as np
arr = np.array([[[1, 2, 3, 4],
[5, 6, 7, 8],
[9, 10, 11, 12]],
[[13, 14, 15, 16],
[17, 18, 19, 20],
[21, 22, 23, 24]]], dtype=np.int32)
lib = ct.CDLL('./test')
affect_array = lib.affect_array
# Require 3D numpy array of int32 as first parameter. It will be type-checked.
affect_array.argtypes = np.ctypeslib.ndpointer(dtype=np.int32, ndim=3), ct.c_int, ct.c_int, ct.c_int
affect_array.restype = None
print('Original array =')
print(arr)
print()
affect_array(arr, *arr.shape) # * unpacks tuple as 3 additional parameters.
print('Array after pass to function =')
print(arr)
Output:
Original array =
[[[ 1 2 3 4]
[ 5 6 7 8]
[ 9 10 11 12]]
[[13 14 15 16]
[17 18 19 20]
[21 22 23 24]]]
Array after pass to function =
[[[ 2 4 6 8]
[10 12 14 16]
[18 20 22 24]]
[[26 28 30 32]
[34 36 38 40]
[42 44 46 48]]]
本文标签： pythonPassing a NumPy 3d array to a C function with a triple pointer as an argumentStack Overflow
版权声明：本文标题：python - Passing a NumPy 3d array to a C function with a triple pointer as an argument - Stack Overflow 内容由网友自发贡献，该文观点仅代表作者本人， 转载请联系作者并注明出处：http://www.betaflare.com/web/1741842979a2400613.html， 本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如发现本站有涉嫌抄袭侵权/违法违规的内容，一经查实，本站将立刻删除。
t_ptr[i] = (ct.POINTER(ct.c_int) * cols)()andt_ptr[i][j] = arr_c[i][j].ctypes.data_as(ct.POINTER(ct.c_int))both discard the only remaining reference to the new array. – user2357112 Commented Feb 2 at 16:07