Got an AssertionError: The 0-th index of a Matrix/Vector must be a compile-time constant integer

综合讨论区
#1

I got the following error trying to assign values to entries of a Matrix using data from other Vectors. My code is the following. The problematic line(s) include line 39:

k[2 * i, 2 * j] = b / a * (1 + 1 / 3 * l_y[i] * l_y[j]) * l_x[i] * l_x[j] + a / b * (1 - nu) / 2 * (1 + l_x[i] * l_x[j] / 3) * l_y[i] * l_y[j]

How do I fix it? Any suggestion is appreciated. Thanks a lot!

error
error1343×545 27.2 KB 

import taichi as ti
import numpy as np

ti.init(arch = ti.gpu)

# material parameters
nu = 0.3 #Poisson's ratio
l = 8.0 #length

# simulation parameters
n_no_x = 3 #no. of nodes in x direction
n_no_y = 2 #no. of nodes in y direction
n_no = n_no_x * n_no_y
n_no_active = (n_no_x - 1) * n_no_y #excluding Dirichlet nodes

ele = ti.Vector(4, dt = ti.i32, shape = (n_no_y - 1, n_no_x - 1)) #connectivity matrix
init_pos = ti.var(dt = ti.f32, shape = n_no * 2) # store initial node position (not updated for infinitesimal deformation)


@ti.func
def k_ele(x0, x1, x2, x3):
    l_x = ti.Vector([-1, 1, 1, -1])
    l_y = ti.Vector([-1, -1, 1, 1])
    k = ti.Matrix([[0, 0, 0, 0, 0, 0, 0, 0],
                   [0, 0, 0, 0, 0, 0, 0, 0],
                   [0, 0, 0, 0, 0, 0, 0, 0],
                   [0, 0, 0, 0, 0, 0, 0, 0],
                   [0, 0, 0, 0, 0, 0, 0, 0],
                   [0, 0, 0, 0, 0, 0, 0, 0],
                   [0, 0, 0, 0, 0, 0, 0, 0],
                   [0, 0, 0, 0, 0, 0, 0, 0]])

    a = (x1[0] - x0[0]) / 2
    b = (x2[1] - x0[1]) / 2

    for i, j in ti.ndrange(4, 4):
        k[2 * i, 2 * j] = b / a * (1 + 1 / 3 * l_y[i] * l_y[j]) * l_x[i] * l_x[j] + a / b * (1 - nu) / 2 * (1 + l_x[i] * l_x[j] / 3) * l_y[i] * l_y[j]
    return k

@ti.kernel
def init():
    for i, j in ele:
        ele[i, j][0] = j + i * n_no_x
        ele[i, j][1] = ele[i, j][0] + 1
        ele[i, j][2] = ele[i, j][1] + n_no_x
        ele[i, j][3] = ele[i, j][2] - 1

    for i, j in ti.ndrange(n_no_y, n_no_x):
        init_pos[2 * (j + i * n_no_x)] = j * (l / (n_no_x - 1))
        init_pos[2 * (j + i * n_no_x) + 1] = i * (l / (n_no_x - 1))

    for i, j in ele:
        x0 = ti.Vector([init_pos[2 * ele[i, j][0]], init_pos[2 * ele[i, j][0] + 1]])
        x1 = ti.Vector([init_pos[2 * ele[i, j][1]], init_pos[2 * ele[i, j][1] + 1]])
        x2 = ti.Vector([init_pos[2 * ele[i, j][2]], init_pos[2 * ele[i, j][2] + 1]])
        x3 = ti.Vector([init_pos[2 * ele[i, j][3]], init_pos[2 * ele[i, j][3] + 1]])

        k = k_ele(x0, x1, x2, x3)

init()
#2

Thanks for the question.

Does changing that line to

for i, j in ti.static(ti.ndrange(4, 4)):

solve your problem?

#3

Yes, exactly! Thank you. Also found it mentioned in the document (https://taichi.readthedocs.io/en/stable/meta.html).

1 个赞