comp_analysis.py

#!/usr/bin/python3
import numpy as np
import sys
import matplotlib.pyplot as plt
import config


def get_improvement(hi, lo, performance):
    """
    Get improvement from two execution speeds.
    Arguments:
      hi, lo: (double) execution speeds
      performance: (bool) True if calculating performance increase, False if calculating time reduction.
    Returns: (double) Calculated improvement
    """
    if (lo > hi):
        print('input order incorrect')
        sys.exit()
    return 100 * (hi - lo) / (lo if performance else hi)


def generate_data():
    data = {}
    config.init()
    for filename in config.file_struct.keys():
        raw_data = np.genfromtxt(filename, delimiter=',')
        data[filename] = (np.mean(raw_data), raw_data)
    return data


def generate_txt(data):
    """
    Generate output .txt file
    Arguments:
      data = {filename: (average execution time, [raw data])}
    """
    native_comp_p = get_improvement(
        data[config.native_no_sse_fn][0], data[config.native_sse_fn][0], True)
    wasm_comp_p = get_improvement(
        data[config.wasm_no_simd_fn][0], data[config.wasm_simd_fn][0], True)
    simd_comp_p = get_improvement(
        data[config.wasm_simd_fn][0], data[config.native_sse_fn][0], True)
    no_simd_comp_p = get_improvement(
        data[config.wasm_no_simd_fn][0], data[config.native_no_sse_fn][0], True)

    native_comp_t = get_improvement(
        data[config.native_no_sse_fn][0], data[config.native_sse_fn][0], False)
    wasm_comp_t = get_improvement(
        data[config.wasm_no_simd_fn][0], data[config.wasm_simd_fn][0], False)
    simd_comp_t = get_improvement(
        data[config.wasm_simd_fn][0], data[config.native_sse_fn][0], False)
    no_simd_comp_t = get_improvement(
        data[config.wasm_no_simd_fn][0], data[config.native_no_sse_fn][0], False)

    with open(config.comp_struct[0], 'w') as f:
        f.write("averages: \n")
        f.write("  native without sse2  : " +
                str(data[config.native_no_sse_fn][0]) + "\n")
        f.write("  native with sse2     : " +
                str(data[config.native_sse_fn][0]) + "\n")
        f.write("  wasm without simd128 : " +
                str(data[config.wasm_no_simd_fn][0]) + "\n")
        f.write("  wasm with simd128    : " +
                str(data[config.wasm_simd_fn][0]) + "\n")
        f.write("\n")
        f.write("performance increase percentage [100 * (hi - lo) / lo]: \n")
        f.write("  native without sse2 -> native with sse2     : " +
                str(native_comp_p) + "\n")
        f.write("  wasm without simd128 -> wasm with simd128   : " +
                str(wasm_comp_p) + "\n")
        f.write("  wasm with simd128 -> native with sse2       : " +
                str(simd_comp_p) + "\n")
        f.write("  wasm without simd128 -> native without sse2 : " +
                str(no_simd_comp_p) + "\n")
        f.write("\n")
        f.write("time reduction percentage [100 * (hi - lo) / hi]: \n")
        f.write("  native without sse2 -> native with sse2     : " +
                str(native_comp_t) + "\n")
        f.write("  wasm without simd128 -> wasm with simd128   : " +
                str(wasm_comp_t) + "\n")
        f.write("  wasm with simd128 -> native with sse2       : " +
                str(simd_comp_t) + "\n")
        f.write("  wasm without simd128 -> native without sse2 : " +
                str(no_simd_comp_t) + "\n")


def generate_plt(data):
    """
    Generate output histogram .png file
    Arguments:
      data = {filename: (average execution time, [raw data])}
    """
    bins = np.arange(config.MIN, config.MAX, config.delta)
    plt.hist(data[config.native_no_sse_fn][1], bins, alpha=config.alpha,
             label=config.comp_struct[1][0], edgecolor="black")
    plt.hist(data[config.native_sse_fn][1], bins, alpha=config.alpha,
             label=config.comp_struct[1][1], edgecolor="black")
    plt.hist(data[config.wasm_no_simd_fn][1], bins, alpha=config.alpha,
             label=config.comp_struct[1][2], edgecolor="black")
    plt.hist(data[config.wasm_simd_fn][1], bins, alpha=config.alpha,
             label=config.comp_struct[1][3], edgecolor="black")

    plt.xlabel("Time [s]")
    plt.ylabel("Frequency")
    plt.legend(loc='upper left', fontsize='small')
    plt.title(config.comp_struct[2])
    plt.savefig(config.comp_struct[3])


def generate_bar(data):
    """
    Generate output bar chart .png file
    Arguments:
      data = {filename: (average execution time, [raw data])}
    """
    x_axis = config.comp_struct[5]
    y_axis = np.zeros(4)
    for i in range(4):
        y_axis[i + 1 if i % 2 == 0 else i - 1] = list(data.values())[i][0]
    print(x_axis)
    print(y_axis)
    plt.figure(figsize=config.fig_size)
    plt.bar(x_axis, y_axis)
    plt.xlabel('Implementation')
    plt.ylabel('Time [s]')
    plt.title(config.comp_struct[2])
    plt.savefig(config.comp_struct[4])


def main():
    data = generate_data()
    generate_txt(data)
    generate_plt(data)
    plt.close()
    generate_bar(data)


if __name__ == "__main__":
    main()