CuPy记录

因传统的Numpy运算速度较慢，使用其镜像GPU加速版本Cupy

官网

https://docs.cupy.dev/en/stable/index.html

Project Goal

The goal of the CuPy project is to provide Python users GPU acceleration capabilities, without the in-depth knowledge of underlying GPU technologies. The CuPy team focuses on providing:

A complete NumPy and SciPy API coverage to become a full drop-in replacement, as well as advanced CUDA features to maximize the performance.

Mature and quality library as a fundamental package for all projects needing acceleration, from a lab environment to a large-scale cluster.

安装

https://docs.cupy.dev/en/stable/install.html

需要安装CUDA Toolkit，我之前安装PyTorch环境时已经安装过，在NVIDIA控制面板或者运行下列代码查看CUDA版本：

import torch
print(torch.version.cuda)

选择对应CuPy版本进行安装,我这里是11.7（我在Anaconda Prompt中安装）

pip install cupy-cuda11x

之前安装PyTorch时候CUDA还是11.7，不知什么时候升级了到了12.0好像：

验证安装

可以开始使用了！

---

注意：Move array from a device to the host（使用gdal输出的时候）:
https://docs.cupy.dev/en/stable/user_guide/basic.html

x_gpu = cp.array([1, 2, 3])  # create an array in the current device
x_cpu = cp.asnumpy(x_gpu)  # move the array to the host.

x_cpu = x_gpu.get()

numpy 和 cumpy可以相互转换，读取的时候只能被其中一个读取(下边#1 读取了 #2就读取不到内容了)：

f = open(dir, 'rb')
f_int = np.fromfile(f, dtype=np.int16)  # 1
f_int2 = cp.fromfile(f, dtype=cp.int16)  # 2