test.py

import numpy as np
import re
import sys
import matplotlib.pyplot as plt
from numpy import linalg as lng
from subprocess import Popen, PIPE
import os

num_of_tests = 400
sample_range = 1e6


def main():
    Popen('make').wait()
    if not os.path.exists('tests'):
        os.makedirs('tests')
    gpu_norms = []
    gpu_runtimes = []
    cpu_norms = []
    cpu_runtimes = []
    samples = [10 * i for i in range(1, num_of_tests + 1)]
    for i, n in enumerate(samples, start=1):
        input_path = f'tests/in.txt'
        arr = np.random.uniform(low=-sample_range, high=sample_range, size=(n, n))
        np.savetxt(input_path, arr, delimiter=' ')

        for dev in ['cpu', 'gpu']:
            output_path = f'tests/out{dev}.txt'
            print(f'{dev} test{i}\tstarted (n={n}).')
            f = '-c'
            if dev == 'gpu':
                f = '-g'
            p1 = Popen(['./GJE', f, '-n', str(n), '-f', str(input_path), '-o', str(output_path)],
                       stdout=PIPE, stderr=PIPE)
            stdout, stderr = p1.communicate()
            stdout_str, stderr_str = (stdout.decode("utf-8"), stderr.decode("utf-8"))
            print(stdout_str)
            print(stderr_str, file=sys.stderr)
            runtime = float(re.findall(r"[+-]? *(?:\d+(?:\.\d*)?|\.\d+)(?:[eE][+-]?\d+)?", stdout_str)[0])
            print(f'{dev} test{i}\tfinished.')
            inv = np.array([[float(i) for i in line.split()] for line in open(output_path)])
            norm = lng.norm(np.matmul(arr, inv) - np.identity(n))
            if dev == 'cpu':
                cpu_runtimes.append(runtime)
                cpu_norms.append(norm)
            else:
                gpu_runtimes.append(runtime)
                gpu_norms.append(norm)

    plt.figure(figsize=(10, 7.5))
    plt.plot(samples, cpu_runtimes)
    plt.plot(samples, gpu_runtimes)
    plt.legend(['cpu', 'gpu'], loc='upper left')
    plt.title('runtime plot')
    plt.xlabel('n')
    plt.ylabel('ms')
    plt.show()

    plt.figure(figsize=(10, 7.5))
    plt.plot(samples, cpu_norms)
    plt.plot(samples, gpu_norms)
    plt.legend(['cpu', 'gpu'], loc='upper left')
    plt.title('precision plot')
    plt.xlabel('n')
    plt.ylabel('norm2')
    plt.show()


if __name__ == '__main__':
    main()