Problem with argsort function when using cuda target

Numba
1

Good morning
i’m new with numba.
i write the following code and i get the following error:

Failed in nopython mode pipeline (step: nopython frontend)
Use of unsupported NumPy function ‘numpy.argsort’ or unsupported use of the function.

File “”, line 9:
def agsortnum(dis,disout):
disout[:]=np.argsort(dis[:])
^

from numba import guvectorize, float32
from numba import cuda
import numpy as np

@guvectorize(['void(float32[:], float32[:])'], '(n)->(n)',
             nopython=True, target='cuda')
def agsortnum(dis,disout):
   disout[:]=np.argsort(dis[:])
 
dis1=[5,2,7,1,4,2,8,9]
dis1=float32(dis1) 
dist = np.zeros(dis1.shape[0]).astype('float32')
dis11 = cuda.to_device(dis1)
dist1 = cuda.to_device(dist)
print(dis1)
agsortnum(dis11,dist1)
print(dist1)

thanks an advance

2

Many NumPy functions, including argsort are not supported on the CUDA target at present. In order to implement this gufunc, you would need to “manually” implement the function inside your gufunc.