python nvidia GPU 測試

2022-11-23

是要做 https://vast.ai/console/host/setup/ 就做一下GPU測試

有 python3 後

pip3 install numba

安裝driver

nvidia 官網下載 CUDA

https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=22.04&target_type=deb_local

安裝完後，先重開機！

nvidia-smi 確認有抓到顯卡

安裝 toolkit

apt install nvidia-cuda-toolkit

安裝完後用以下指令確認版號 nvcc -V

我機器發生 nvidia-driver 與 toolkit版號不同測試了一下解決就是換 driver 版本

apt install nvidia-driver-510

安裝完後，先重開機！

apt install nvidia-cuda-toolkit

測試顯示是:

root@user:~# nvidia-smi
Wed Nov 23 05:19:33 2022
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.85.02    Driver Version: 510.85.02    CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  On   | 00000000:04:00.0 Off |                  N/A |
|  0%   41C    P8    16W / 310W |      1MiB /  8192MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   1  NVIDIA GeForce ...  On   | 00000000:06:00.0 Off |                  N/A |
|  0%   38C    P8     8W / 310W |      1MiB /  8192MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+
root@user:~# nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Thu_Nov_18_09:45:30_PST_2021
Cuda compilation tools, release 11.5, V11.5.119
Build cuda_11.5.r11.5/compiler.30672275_0

GPU 與 CPU 運算比較

from numba import jit, cuda
import numpy as np
# to measure exec time
from timeit import default_timer as timer

# normal function to run on cpu
def func(a):								
	for i in range(10000000):
		a[i]+= 1	

# function optimized to run on gpu
@jit(target_backend='cuda')						
def func2(a):
	for i in range(10000000):
		a[i]+= 1
if __name__=="__main__":
	n = 10000000							
	a = np.ones(n, dtype = np.float64)
	
	start = timer()
	func(a)
	print("without GPU:", timer()-start)	
	
	start = timer()
	func2(a)
	print("with GPU:", timer()-start)

我機器是

without GPU: 1.4943597909999937
with GPU: 0.04574826399999665

用 GPU 來做其他事 …. 如 print

from numba import cuda

def cpu_print(N):
    for i in range(0, N):
        print(i)

@cuda.jit
def gpu_print(N):
    idx = cuda.threadIdx.x + cuda.blockIdx.x * cuda.blockDim.x 
    if (idx < N):
        print(idx)

def main():
    print("gpu print:")
    gpu_print[2, 4](8)
    cuda.synchronize()
    print("cpu print:")
    cpu_print(8)

if __name__ == "__main__":
    main()