Q: 你是个计算机工程师，请用Python分别演示用单线程和多线程计算两个矩阵相乘
GPT-4:
在 Python 中，计算矩阵乘法可以用多种方式实现，包括使用单线程和多线程。以下是两种简单的示例：
单线程
def matrix_multiply(A, B):
rows_A, cols_A = len(A), len(A[0])
rows_B, cols_B = len(B), len(B[0])
if cols_A != rows_B:
return "Cannot multiply the two matrices. They have incompatible dimensions."
# Initialize result matrix filled with zeros
C = [[0 for row in range(cols_B)] for col in range(rows_A)]
for i in range(rows_A):
for j in range(cols_B):
for k in range(cols_A):
C[i][j] += A[i][k] * B[k][j]
return C
A = [
[1, 2, 3],
[4, 5, 6],
[7, 8, 9]
]
B = [
[10, 11, 12],
[13, 14, 15],
[16, 17, 18]
]
print("Single-threaded result:", matrix_multiply(A, B))
多线程
使用 Python 的 threading 模块进行多线程操作。
import threading
def threaded_matrix_multiply(A, B, C, row_indices):
cols_A = len(A[0])
cols_B = len(B[0])
for i in row_indices:
for j in range(cols_B):
for k in range(cols_A):
C[i][j] += A[i][k] * B[k][j]
A = [
[1, 2, 3],
[4, 5, 6],
[7, 8, 9]
]
B = [
[10, 11, 12],
[13, 14, 15],
[16, 17, 18]
]
rows_A = len(A)
cols_B = len(B[0])
# Initialize result matrix filled with zeros
C = [[0 for row in range(cols_B)] for col in range(rows_A)]
# Create threads
threads = []
split_indices = [[0], [1], [2]]
for indices in split_indices:
t = threading.Thread(target=threaded_matrix_multiply, args=(A, B, C, indices))
threads.append(t)
t.start()
# Wait for all threads to complete
for t in threads:
t.join()
print("Multi-threaded result:", C)
在这个多线程的例子中，每个线程负责计算结果矩阵的一部分。这样做可以加速计算，尤其是在处理大型矩阵时。
请注意，Python 的 Global Interpreter Lock（GIL）可能会限制多线程代码的效能。对于计算密集型任务，使用多进程或者专门的数值计算库（如 NumPy）通常更加高效。