forked from Alexey/DAS_2024_1
62 lines
2.3 KiB
Python
62 lines
2.3 KiB
Python
|
import numpy as np
|
|||
|
|
import time
|
|||
|
|
from multiprocessing import Pool
|
|||
|
|
|
|||
|
|
def multiply_matrices_sequential(A, B):
|
|||
|
|
"""Последовательное умножение матриц."""
|
|||
|
|
n = A.shape[0]
|
|||
|
|
C = np.zeros((n, n))
|
|||
|
|
for i in range(n):
|
|||
|
|
for j in range(n):
|
|||
|
|
C[i][j] = np.dot(A[i], B[:, j]) # Используем векторизированное умножение для повышения производительности
|
|||
|
|
return C
|
|||
|
|
|
|||
|
|
def worker(args):
|
|||
|
|
"""Функция для параллельного умножения матриц, которая обрабатывает строки."""
|
|||
|
|
A, B, row_indices = args
|
|||
|
|
C_part = np.zeros((len(row_indices), B.shape[1]))
|
|||
|
|
|
|||
|
|
for idx, i in enumerate(row_indices):
|
|||
|
|
C_part[idx] = np.dot(A[i], B)
|
|||
|
|
|
|||
|
|
return C_part
|
|||
|
|
|
|||
|
|
def multiply_matrices_parallel(A, B, num_workers):
|
|||
|
|
"""Параллельное умножение матриц."""
|
|||
|
|
n = A.shape[0]
|
|||
|
|
C = np.zeros((n, n))
|
|||
|
|
row_indices = np.array_split(range(n), num_workers)
|
|||
|
|
|
|||
|
|
with Pool(processes=num_workers) as pool:
|
|||
|
|
results = pool.map(worker, [(A, B, idx) for idx in row_indices])
|
|||
|
|
|
|||
|
|
# Объединяем результаты
|
|||
|
|
for i, result in enumerate(results):
|
|||
|
|
C[i * len(result): (i + 1) * len(result)] = result
|
|||
|
|
|
|||
|
|
return C
|
|||
|
|
|
|||
|
|
def benchmark(matrix_size, num_workers):
|
|||
|
|
# Генерация случайных матриц
|
|||
|
|
A = np.random.rand(matrix_size, matrix_size)
|
|||
|
|
B = np.random.rand(matrix_size, matrix_size)
|
|||
|
|
|
|||
|
|
start_time = time.time()
|
|||
|
|
if num_workers == 1:
|
|||
|
|
C = multiply_matrices_sequential(A, B)
|
|||
|
|
else:
|
|||
|
|
C = multiply_matrices_parallel(A, B, num_workers)
|
|||
|
|
end_time = time.time()
|
|||
|
|
|
|||
|
|
method = "последовательное" if num_workers == 1 else f"параллельное с {num_workers} потоками"
|
|||
|
|
print(f"{method.capitalize()} умножение матриц {matrix_size}x{matrix_size}: {end_time - start_time:.6f} сек")
|
|||
|
|
|
|||
|
|
if __name__ == "__main__":
|
|||
|
|
# Запуск бенчмарков
|
|||
|
|
sizes = [100, 300, 500]
|
|||
|
|
for size in sizes:
|
|||
|
|
print(f"\nБенчмарк для матриц размером {size}x{size}:")
|
|||
|
|
benchmark(size, 1) # Последовательный
|
|||
|
|
benchmark(size, 4) # Параллельный (4 потока)
|
|||
|
|
benchmark(size, 8) # Параллельный (8 потоков)
|