How can I make better use of CPU to speed up np.linspace

Numba Community Support
1 
from numba import njit, prange
# from skimage.draw import line_nd
import numpy as np
import random


def _round_safe(coords):
    if len(coords) > 1 and coords[0] % 1 == 0.5 and coords[1] - coords[0] == 1:
        _round_function = np.floor
    else:
        _round_function = np.round
    return _round_function(coords).astype(int)
def line_nd(start, stop, *, endpoint=False, integer=True):
    start = np.asarray(start)
    stop = np.asarray(stop)
    npoints = int(np.ceil(np.max(np.abs(stop - start))))
    if endpoint:
        npoints += 1
    coords = np.linspace(start, stop, num=npoints, endpoint=endpoint).T
    if integer:
        for dim in range(len(start)):
            coords[dim, :] = _round_safe(coords[dim, :])
        coords = coords.astype(int)
    return tuple(coords)




@njit(parallel=True)
def get_all(a1):
    for i1 in prange(1,999999):  # Use prange for parallelism
        a11 = random.randint(100, 10000)
        start = np.asarray((11*a11*0.01, 22*a11*0.02, 33*a11*0.03))
        stop = np.asarray((66.1, 77.1, 99.1))
        line_nd(start,stop)


get_all(1)

numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Untyped global name ‘line_nd’: Cannot determine Numba type of <class ‘function’>

File “测试1.py”, line 35:
def get_all(a1):

stop = np.asarray((66.1, 77.1, 99.1))
line_nd(start,stop)
^

During: Pass nopython_type_inference

2

As a general rule, it only makes sense for jitted functions to call other jitted functions. I’d recommend starting with the lowest-level function and making sure it works in both plain-python and jit-mode and work up from there.