Python #Tools #Dataprocessing #COMSOL

目的

目前对于python处理文字有了一些心得，刚好最近在用COMSOL做数值计算，需要参数扫描。我想分析每次计算COMSOL需要的计算内存和运算时间，防止计算内存爆掉了。

具体操作

COMSOL log文件信息详细记录了每一次扫描的一些进程，一般开头是这样的

*******************************************
***COMSOL 5.4.0.225 progress output file***
*******************************************
Thu Feb 20 21:36:07 CST 2020
<---- Compile Equations: Wavelength Domain {st1} in Study 1 {std1}/Solution 1
      (sol1) {sol1} ------------------------------------------------------------
Started at Feb 20, 2020 9:36:13 PM.
Running on 4 x Intel(R) Xeon(R) Platinum 8163 CPU at 2.50 GHz.
Using 4 sockets with 25 cores in total on 4CPU.

运算完毕之后，会有这样的字样

Memory: 106911/106911 151517/151517
Iter      SolEst     Damping    Stepsize #Res #Jac #Sol   LinErr   LinRes
---------- Current Progress: 100 % - Solving linear system
Memory: 115456/117206 149829/151541
   1        0.79   1.0000000        0.79    1    1    1  2.2e-11  6.9e-12
Solution time: 815 s. (13 minutes, 35 seconds)
Physical memory: 122.42 GB
Virtual memory: 156.74 GB
Ended at Feb 20, 2020 9:50:10 PM.

我需要的就是快计算完毕后的这些字段的信息，Physical memory: 122.42 GB,Solution time: 815 s. (13 minutes, 35 seconds),
其实只需要依次读取log文件的每一行，判断是否有我需要的字段，如果有就将对应的字段提取出来。具体代码如下

import numpy as np
import matplotlib.pyplot as plt


def changetext(filename):
    with open(filename, 'r', encoding='utf-8') as f:
        lines_memo = []  # 创建了一个空列表，里面没有元素
        lines_time = []  # 创建了一个空列表，里面没有元素
        for line in f.readlines():
            if 'Physical memory' in line:
                memo = line[17:-4] # 只读取有用字段
                lines_memo.append(memo)
            if 'Solution time' in line:
                memo = line[15:19] # 只读取有用字段
                lines_time.append(memo)

    return lines_memo, lines_time


if __name__ == "__main__":
    # 服务器一共有三个文件
    dst_dir_1 = '../Log-File/20200220/comsol_progress_11.txt'
    dst_dir_2 = '../Log-File/20200220/comsol_progress_21.txt'
    dst_dir_3 = '../Log-File/20200220/comsol_progress_31.txt'

    lines_memo_1, lines_time_1 = changetext(dst_dir_1)
    lines_memo_2, lines_time_2 = changetext(dst_dir_2)
    lines_memo_3, lines_time_3 = changetext(dst_dir_3)
    print(len(lines_time_1))
    print(len(lines_time_2))
    print(len(lines_time_3))

    num_1 = int((len(lines_time_1)-1)/2)
    num_2 = int((len(lines_time_2)-1)/2)
    num_3 = int((len(lines_time_3)-1)/2)

    x_1 = np.linspace(1, num_1, num_1)
    x_2 = np.linspace(1, num_2, num_2)
    x_3 = np.linspace(1, num_3, num_3)

    memo_mat_1 = np.zeros(num_1)
    time_mat_1 = np.zeros(num_1)

    memo_mat_2 = np.zeros(num_2)
    time_mat_2 = np.zeros(num_2)

    memo_mat_3 = np.zeros(num_3)
    time_mat_3 = np.zeros(num_3)

    for l in range(num_1):
        memo_mat_1[l] = float(lines_memo_1[3*(l)+2])
        time_mat_1[l] = float(lines_time_1[2*(l)+1])
    for l in range(num_2):
        memo_mat_2[l] = float(lines_memo_2[3*(l)+2])
        time_mat_2[l] = float(lines_time_2[2*(l)+1])
    for l in range(num_3):
        memo_mat_3[l] = float(lines_memo_3[3*(l)+2])
        time_mat_3[l] = float(lines_time_3[2*(l)+1])
    # 绘图部分
    fig1 = plt.figure(figsize=(6.4*2, 4.8))
    plt.subplot(121)
    plt.plot(x_1, memo_mat_1, label='4CPU 1')
    plt.plot(x_2, memo_mat_2, label='4CPU 2')
    plt.plot(x_3, memo_mat_3, label='Dell 1')
    plt.xlabel('Loop Counts')
    plt.ylabel('Physical Memory (GB)')
    plt.legend(loc='best')

    plt.subplot(122)
    plt.plot(x_1, time_mat_1, label='4CPU 1')
    plt.plot(x_2, time_mat_2, label='4CPU 2')
    plt.plot(x_3, time_mat_3, label='Dell 1')
    plt.xlabel('Loop Counts')
    plt.ylabel('Simulation Time (s)')

    plt.legend(loc='best')

    plt.show()