使用Python Cupy模块加速大规模数值计算实例深究

pip install cupy

import cupy as cp

# 创建一个随机数组
x = cp.random.rand(100)
print(x)

import cupy as cp
# 创建两个数组
arr1 = cp.array([1, 2, 3])
arr2 = cp.array([4, 5, 6])
# 求和
result = arr1 + arr2
print(result)
# 逐元素乘法
result = arr1 * arr2
print(result)

import cupy as cp
# 创建两个随机矩阵
matrix_a = cp.random.rand(3, 3)
matrix_b = cp.random.rand(3, 3)
# 矩阵相乘
result = cp.dot(matrix_a, matrix_b)
print(result)

import cupy as cp
# 创建两个随机矩阵
matrix_a = cp.random.rand(1000, 1000)
matrix_b = cp.random.rand(1000, 1000)
# 使用Cupy进行矩阵乘法
result = cp.dot(matrix_a, matrix_b)
print("矩阵乘法结果:", result)

import cupy as cp
# 创建一个数组
arr = cp.array([1, 2, 3, 4, 5])
# 逐元素求平方
squared = cp.square(arr)
print("数组平方:", squared)
# 索引和切片操作
print("数组的前三个元素:", arr[:3])

import numpy as np
import cupy as cp
import time
# 使用Numpy创建一个大数组
np_arr = np.random.rand(10000, 10000)
# 使用Cupy创建一个大数组
cp_arr = cp.random.rand(10000, 10000)
# 对比 Numpy 与 Cupy 的矩阵乘法性能
start_time = time.time()
np_result = np.dot(np_arr, np_arr)
numpy_time = time.time() - start_time
start_time = time.time()
cp_result = cp.dot(cp_arr, cp_arr)
cupy_time = time.time() - start_time
print("Numpy 矩阵乘法时间:", numpy_time)
print("Cupy 矩阵乘法时间:", cupy_time)