Question

如果导入以下内容，它将编译为预期的.so文件：

import numba as nb
from numba.pycc import CC

cc = CC('t1')


@nb.njit(parallel=True)
@cc.export('diff', '(f4[:], i8)')
def diff(buffer, half_buffer_size):
    for tau in nb.prange(half_buffer_size):
        buffer[tau] = 0


@nb.njit
@cc.export('test', '(f4[:], i8)')
def test(buffer, half_buffer_size):
    diff(buffer, half_buffer_size)

cc.compile()

已经编译了文件，如果我随后运行再次导入t1的脚本，则会出现以下错误：

Traceback (most recent call last):
  File "/Users/me/PycharmProjects/t1/main.py", line 2, in <module>
    import t1
ImportError: dlopen(/Users/me/PycharmProjects/t1/t1.cpython-38-darwin.so, 2): Symbol not found: _do_scheduling_unsigned
  Referenced from: /Users/me/PycharmProjects/t1/t1.cpython-38-darwin.so
  Expected in: flat namespace
 in /Users/me/PycharmProjects/t1/t1.cpython-38-darwin.so

这似乎是在测试内调用diff的结果。如果我对此调用进行注释并重新编译，则可以直接从main.py调用diff，而不会出现任何问题。我该如何解决这个问题？

找不到符号：使用numba时_do_scheduling_unsigned

0 个答案: