Anforderungen  |   Konzepte  |   Entwurf  |   Entwicklung  |   Qualitätssicherung  |   Lebenszyklus  |   Steuerung
 
 
 
 


Quelle  amd_pstate_trace.py   Sprache: Python

 
#!/usr/bin/env python3
# SPDX-License-Identifier: GPL-2.0-only
# -*- coding: utf-8 -*-
#
""" This utility can be used to debug and tune the performance of the
AMD P-State driver. It imports intel_pstate_tracer to analyze AMD P-State
trace event.

Prerequisites:
    Python version 2.7.x or higher
    gnuplot 5.0 or higher
    gnuplot-py 1.8 or higher
    (Most of the distributions have these required packages. They may be called
     gnuplot-py, phython-gnuplot or phython3-gnuplot, gnuplot-nox, ... )

    Kernel config for Linux trace is enabled

    see print_help(): for Usage and Output details

"""
from __future__ import print_function
from datetime import datetime
import subprocess
import os
import time
import re
import signal
import sys
import getopt
import Gnuplot
from numpy import *
from decimal import *
sys.path.append(os.path.join(os.path.dirname(__file__), "..""intel_pstate_tracer"))
import intel_pstate_tracer as ipt

__license__ = "GPL version 2"

MAX_CPUS = 256
# Define the csv file columns
C_COMM = 15
C_ELAPSED = 14
C_SAMPLE = 13
C_DURATION = 12
C_LOAD = 11
C_TSC = 10
C_APERF = 9
C_MPERF = 8
C_FREQ = 7
C_MAX_PERF = 6
C_DES_PERF = 5
C_MIN_PERF = 4
C_USEC = 3
C_SEC = 2
C_CPU = 1

global sample_num, last_sec_cpu, last_usec_cpu, start_time, test_name, trace_file

getcontext().prec = 11

sample_num =0
last_sec_cpu = [0] * MAX_CPUS
last_usec_cpu = [0] * MAX_CPUS

def plot_per_cpu_freq(cpu_index):
    """ Plot per cpu frequency """

    file_name = 'cpu{:0>3}.csv'.format(cpu_index)
    if os.path.exists(file_name):
        output_png = "cpu%03d_frequency.png" % cpu_index
        g_plot = ipt.common_gnuplot_settings()
        g_plot('set output "' + output_png + '"')
        g_plot('set yrange [0:7]')
        g_plot('set ytics 0, 1')
        g_plot('set ylabel "CPU Frequency (GHz)"')
        g_plot('set title "{} : frequency : CPU {:0>3} : {:%F %H:%M}"'.format(test_name, cpu_index, datetime.now()))
        g_plot('set ylabel "CPU frequency"')
        g_plot('set key off')
        ipt.set_4_plot_linestyles(g_plot)
        g_plot('plot "' + file_name + '" using {:d}:{:d} with linespoints linestyle 1 axis x1y1'.format(C_ELAPSED, C_FREQ))

def plot_per_cpu_des_perf(cpu_index):
    """ Plot per cpu desired perf """

    file_name = 'cpu{:0>3}.csv'.format(cpu_index)
    if os.path.exists(file_name):
        output_png = "cpu%03d_des_perf.png" % cpu_index
        g_plot = ipt.common_gnuplot_settings()
        g_plot('set output "' + output_png + '"')
        g_plot('set yrange [0:255]')
        g_plot('set ylabel "des perf"')
        g_plot('set title "{} : cpu des perf : CPU {:0>3} : {:%F %H:%M}"'.format(test_name, cpu_index, datetime.now()))
        g_plot('set key off')
        ipt.set_4_plot_linestyles(g_plot)
        g_plot('plot "' + file_name + '" using {:d}:{:d} with linespoints linestyle 1 axis x1y1'.format(C_ELAPSED, C_DES_PERF))

def plot_per_cpu_load(cpu_index):
    """ Plot per cpu load """

    file_name = 'cpu{:0>3}.csv'.format(cpu_index)
    if os.path.exists(file_name):
        output_png = "cpu%03d_load.png" % cpu_index
        g_plot = ipt.common_gnuplot_settings()
        g_plot('set output "' + output_png + '"')
        g_plot('set yrange [0:100]')
        g_plot('set ytics 0, 10')
        g_plot('set ylabel "CPU load (percent)"')
        g_plot('set title "{} : cpu load : CPU {:0>3} : {:%F %H:%M}"'.format(test_name, cpu_index, datetime.now()))
        g_plot('set key off')
        ipt.set_4_plot_linestyles(g_plot)
        g_plot('plot "' + file_name + '" using {:d}:{:d} with linespoints linestyle 1 axis x1y1'.format(C_ELAPSED, C_LOAD))

def plot_all_cpu_frequency():
    """ Plot all cpu frequencies """

    output_png = 'all_cpu_frequencies.png'
    g_plot = ipt.common_gnuplot_settings()
    g_plot('set output "' + output_png + '"')
    g_plot('set ylabel "CPU Frequency (GHz)"')
    g_plot('set title "{} : cpu frequencies : {:%F %H:%M}"'.format(test_name, datetime.now()))

    title_list = subprocess.check_output('ls cpu???.csv | sed -e \'s/.csv//\'',shell=True).decode('utf-8').replace('\n'' ')
    plot_str = "plot for [i in title_list] i.'.csv' using {:d}:{:d} pt 7 ps 1 title i".format(C_ELAPSED, C_FREQ)
    g_plot('title_list = "{}"'.format(title_list))
    g_plot(plot_str)

def plot_all_cpu_des_perf():
    """ Plot all cpu desired perf """

    output_png = 'all_cpu_des_perf.png'
    g_plot = ipt.common_gnuplot_settings()
    g_plot('set output "' + output_png + '"')
    g_plot('set ylabel "des perf"')
    g_plot('set title "{} : cpu des perf : {:%F %H:%M}"'.format(test_name, datetime.now()))

    title_list = subprocess.check_output('ls cpu???.csv | sed -e \'s/.csv//\'',shell=True).decode('utf-8').replace('\n'' ')
    plot_str = "plot for [i in title_list] i.'.csv' using {:d}:{:d} pt 255 ps 1 title i".format(C_ELAPSED, C_DES_PERF)
    g_plot('title_list = "{}"'.format(title_list))
    g_plot(plot_str)

def plot_all_cpu_load():
    """ Plot all cpu load """

    output_png = 'all_cpu_load.png'
    g_plot = ipt.common_gnuplot_settings()
    g_plot('set output "' + output_png + '"')
    g_plot('set yrange [0:100]')
    g_plot('set ylabel "CPU load (percent)"')
    g_plot('set title "{} : cpu load : {:%F %H:%M}"'.format(test_name, datetime.now()))

    title_list = subprocess.check_output('ls cpu???.csv | sed -e \'s/.csv//\'',shell=True).decode('utf-8').replace('\n'' ')
    plot_str = "plot for [i in title_list] i.'.csv' using {:d}:{:d} pt 255 ps 1 title i".format(C_ELAPSED, C_LOAD)
    g_plot('title_list = "{}"'.format(title_list))
    g_plot(plot_str)

def store_csv(cpu_int, time_pre_dec, time_post_dec, min_perf, des_perf, max_perf, freq_ghz, mperf, aperf, tsc, common_comm, load, duration_ms, sample_num, elapsed_time, cpu_mask):
    """ Store master csv file information """

    global graph_data_present

    if cpu_mask[cpu_int] == 0:
        return

    try:
        f_handle = open('cpu.csv''a')
        string_buffer = "CPU_%03u, %05u, %06u, %u, %u, %u, %.4f, %u, %u, %u, %.2f, %.3f, %u, %.3f, %s\n" % (cpu_int, int(time_pre_dec), int(time_post_dec), int(min_perf), int(des_perf), int(max_perf), freq_ghz, int(mperf), int(aperf), int(tsc), load, duration_ms, sample_num, elapsed_time, common_comm)
        f_handle.write(string_buffer)
        f_handle.close()
    except:
        print('IO error cpu.csv')
        return

    graph_data_present = True;


def cleanup_data_files():
    """ clean up existing data files """

    if os.path.exists('cpu.csv'):
        os.remove('cpu.csv')
    f_handle = open('cpu.csv''a')
    f_handle.write('common_cpu, common_secs, common_usecs, min_perf, des_perf, max_perf, freq, mperf, aperf, tsc, load, duration_ms, sample_num, elapsed_time, common_comm')
    f_handle.write('\n')
    f_handle.close()

def read_trace_data(file_name, cpu_mask):
    """ Read and parse trace data """

    global current_max_cpu
    global sample_num, last_sec_cpu, last_usec_cpu, start_time

    try:
        data = open(file_name, 'r').read()
    except:
        print('Error opening ', file_name)
        sys.exit(2)

    for line in data.splitlines():
        search_obj = \
            re.search(r'(^(.*?)\[)((\d+)[^\]])(.*?)(\d+)([.])(\d+)(.*?amd_min_perf=)(\d+)(.*?amd_des_perf=)(\d+)(.*?amd_max_perf=)(\d+)(.*?freq=)(\d+)(.*?mperf=)(\d+)(.*?aperf=)(\d+)(.*?tsc=)(\d+)'
                      , line)

        if search_obj:
            cpu = search_obj.group(3)
            cpu_int = int(cpu)
            cpu = str(cpu_int)

            time_pre_dec = search_obj.group(6)
            time_post_dec = search_obj.group(8)
            min_perf = search_obj.group(10)
            des_perf = search_obj.group(12)
            max_perf = search_obj.group(14)
            freq = search_obj.group(16)
            mperf = search_obj.group(18)
            aperf = search_obj.group(20)
            tsc = search_obj.group(22)

            common_comm = search_obj.group(2).replace(' ''')

            if sample_num == 0 :
                start_time = Decimal(time_pre_dec) + Decimal(time_post_dec) / Decimal(1000000)
            sample_num += 1

            if last_sec_cpu[cpu_int] == 0 :
                last_sec_cpu[cpu_int] = time_pre_dec
                last_usec_cpu[cpu_int] = time_post_dec
            else :
                duration_us = (int(time_pre_dec) - int(last_sec_cpu[cpu_int])) * 1000000 + (int(time_post_dec) - int(last_usec_cpu[cpu_int]))
                duration_ms = Decimal(duration_us) / Decimal(1000)
                last_sec_cpu[cpu_int] = time_pre_dec
                last_usec_cpu[cpu_int] = time_post_dec
                elapsed_time = Decimal(time_pre_dec) + Decimal(time_post_dec) / Decimal(1000000) - start_time
                load = Decimal(int(mperf)*100)/ Decimal(tsc)
                freq_ghz = Decimal(freq)/Decimal(1000000)
                store_csv(cpu_int, time_pre_dec, time_post_dec, min_perf, des_perf, max_perf, freq_ghz, mperf, aperf, tsc, common_comm, load, duration_ms, sample_num, elapsed_time, cpu_mask)

            if cpu_int > current_max_cpu:
                current_max_cpu = cpu_int
# Now separate the main overall csv file into per CPU csv files.
    ipt.split_csv(current_max_cpu, cpu_mask)


def signal_handler(signal, frame):
    print(' SIGINT: Forcing cleanup before exit.')
    if interval:
        ipt.disable_trace(trace_file)
        ipt.clear_trace_file()
        ipt.free_trace_buffer()
        sys.exit(0)

trace_file = "/sys/kernel/tracing/events/amd_cpu/enable"
signal.signal(signal.SIGINT, signal_handler)

interval = ""
file_name = ""
cpu_list = ""
test_name = ""
memory = "10240"
graph_data_present = False;

valid1 = False
valid2 = False

cpu_mask = zeros((MAX_CPUS,), dtype=int)


try:
    opts, args = getopt.getopt(sys.argv[1:],"ht:i:c:n:m:",["help","trace_file=","interval=","cpu=","name=","memory="])
except getopt.GetoptError:
    ipt.print_help('amd_pstate')
    sys.exit(2)
for opt, arg in opts:
    if opt == '-h':
        print()
        sys.exit()
    elif opt in ("-t""--trace_file"):
        valid1 = True
        location = os.path.realpath(os.path.join(os.getcwd(), os.path.dirname(__file__)))
        file_name = os.path.join(location, arg)
    elif opt in ("-i""--interval"):
        valid1 = True
        interval = arg
    elif opt in ("-c""--cpu"):
        cpu_list = arg
    elif opt in ("-n""--name"):
        valid2 = True
        test_name = arg
    elif opt in ("-m""--memory"):
        memory = arg

if not (valid1 and valid2):
    ipt.print_help('amd_pstate')
    sys.exit()

if cpu_list:
    for p in re.split("[,]", cpu_list):
        if int(p) < MAX_CPUS :
            cpu_mask[int(p)] = 1
else:
    for i in range (0, MAX_CPUS):
        cpu_mask[i] = 1

if not os.path.exists('results'):
    os.mkdir('results')
    ipt.fix_ownership('results')

os.chdir('results')
if os.path.exists(test_name):
    print('The test name directory already exists. Please provide a unique test name. Test re-run not supported, yet.')
    sys.exit()
os.mkdir(test_name)
ipt.fix_ownership(test_name)
os.chdir(test_name)

cur_version = sys.version_info
print('python version (should be >= 2.7):')
print(cur_version)

cleanup_data_files()

if interval:
    file_name = "/sys/kernel/tracing/trace"
    ipt.clear_trace_file()
    ipt.set_trace_buffer_size(memory)
    ipt.enable_trace(trace_file)
    time.sleep(int(interval))
    ipt.disable_trace(trace_file)

current_max_cpu = 0

read_trace_data(file_name, cpu_mask)

if interval:
    ipt.clear_trace_file()
    ipt.free_trace_buffer()

if graph_data_present == False:
    print('No valid data to plot')
    sys.exit(2)

for cpu_no in range(0, current_max_cpu + 1):
    plot_per_cpu_freq(cpu_no)
    plot_per_cpu_des_perf(cpu_no)
    plot_per_cpu_load(cpu_no)

plot_all_cpu_des_perf()
plot_all_cpu_frequency()
plot_all_cpu_load()

for root, dirs, files in os.walk('.'):
    for f in files:
        ipt.fix_ownership(f)

os.chdir('../../')

Messung V0.5
C=99 H=84 G=91

¤ Dauer der Verarbeitung: 0.10 Sekunden  (vorverarbeitet)  ¤

*© Formatika GbR, Deutschland






Wurzel

Suchen

Beweissystem der NASA

Beweissystem Isabelle

NIST Cobol Testsuite

Cephes Mathematical Library

Wiener Entwicklungsmethode

Haftungshinweis

Die Informationen auf dieser Webseite wurden nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit, noch Qualität der bereit gestellten Informationen zugesichert.

Bemerkung:

Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.






                                                                                                                                                                                                                                                                                                                                                                                                     


Neuigkeiten

     Aktuelles
     Motto des Tages

Software

     Produkte
     Quellcodebibliothek

Aktivitäten

     Artikel über Sicherheit
     Anleitung zur Aktivierung von SSL

Muße

     Gedichte
     Musik
     Bilder

Jenseits des Üblichen ....
    

Besucherstatistik

Besucherstatistik

Monitoring

Montastic status badge