Numba CUDA: Compilation error while trying to make a dot product + comparation

Question

I'm very new to numba.cuda and the following code does not work. Was hoping you could identify the reason

from numba import cuda
import numpy as np
from scipy.spatial import ConvexHull

random_points = np.random.uniform(0, 6, (1000, 2))
result = np.full(random_points.shape[0], True)
hull = ConvexHull(np.array([(1, 2), (3, 4), (3, 6), (2, 4.5), (2.5, 5)]))

@cuda.jit
def my_kernel(points, eq, result):
    pos = cuda.grid(1)

    if pos < points.shape[0]:
        for i in range(eq.shape[0]):
            temp = 0.
            for j in range(points.shape[1]):
                temp += eq[i,j]*points[pos,j]
            temp += eq[i,-1]
            if temp > 1e-12:
                result[pos] = False
    

# Host code   
threadsperblock = 256
blockspergrid = int(np.ceil(random_points.shape[0] / threadsperblock))
my_kernel[blockspergrid, threadsperblock](random_points, hull.equations, result)
print(result)

Error:
NvvmError: Failed to compile

<unnamed> (35, 32): parse expected comma after getelementptr's type
NVVM_ERROR_COMPILATION

Answer 1

The code is fine, it's just the version of numba/cudatoolkit. @MichaelSzczesny tried it with numba 0.56.0 and cuda 11.2and it worked, so thanks to him!

Version that did NOT work: numba 0.45.1 & cudatoolkit 11.3.1, installed through conda