From patchwork Wed Jun 14 20:38:22 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Wang, Weilin" X-Patchwork-Id: 108153 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:994d:0:b0:3d9:f83d:47d9 with SMTP id k13csp233208vqr; Wed, 14 Jun 2023 14:19:27 -0700 (PDT) X-Google-Smtp-Source: ACHHUZ78U+kqcsaYhBw5aOny+wBRXfY6r4+rKmfZonJ2irkB8ecGJFJzngMyKL1LZHH8Can7aCjh X-Received: by 2002:a50:ed06:0:b0:516:3261:17d with SMTP id j6-20020a50ed06000000b005163261017dmr9432553eds.20.1686777566995; Wed, 14 Jun 2023 14:19:26 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1686777566; cv=none; d=google.com; s=arc-20160816; b=nY7/LbRrn7b55NALbA1MtIji1tQPcVKJv6yyASl/I8kTRnWzgpQI/YsD1j8P0Lx0R9 DPkzAcXvqCj/hzd8bNdOaXy2kRwOh1VfLcgFQO7vhgfiP295zl7+OZsEyHbpnYCahQO+ +lZcYRO4XsSAvdiFbdezk7x/dgrwDNNmEfhlafqi+YwP8ic4+wiy7pVaRtGkyQrFB/D8 lbub4/LGjwUbbgHKH5Qn6PNSSBKtw55LMCKSyDXc7CNUL6jqyur4qhBbu+busfVg+oxc Ljs6utFMNjc1M9OJVVZk/F4dO/Lgxev7m1hdkQ4SNGozjqYinzrl5N/kfdqi2r1pAUF9 xYJQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:content-transfer-encoding:mime-version :references:in-reply-to:message-id:date:subject:cc:to:from :dkim-signature; bh=t2ustKDGhE06tHi4evGLA5D6WYiVSWOHtwszNwOgi30=; b=SMidmnCf1OhE5v7FNss64fTuhpiG8wRNLZCvGwDlE3wSvDUE4mY7lREzDrKUpqkN3f L687yCRRTCvJbmGCW3ODkyXmbOf0cZ4x6hxFA8VFXn3FsaIswrES8u2c3XidxALpUTZR LeRj7wVFDx5QFZPqL80KbSXxPrKZGk3evTZNpWnKMKGdaJM50K6HC+GiMw3KZvWPm5PJ iNRDQSOIfBBJwkaHSoAUjD+A729e0JGM7l0tQajHrl3Bx+fOS28SqsQnEpKFHj7dx19G AMbu+U0bc1NwjmUDskVcmMaQ4QXnUZdiIRRCddQU8/KPmJmIIn9Um7cD0geRa5f4lE/u EbZA== ARC-Authentication-Results: i=1; mx.google.com; dkim=pass header.i=@intel.com header.s=Intel header.b=L8COK8vk; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=intel.com Received: from out1.vger.email (out1.vger.email. [2620:137:e000::1:20]) by mx.google.com with ESMTP id be12-20020a0564021a2c00b005149e8ae7afsi8618233edb.381.2023.06.14.14.19.02; Wed, 14 Jun 2023 14:19:26 -0700 (PDT) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20; Authentication-Results: mx.google.com; dkim=pass header.i=@intel.com header.s=Intel header.b=L8COK8vk; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=intel.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S235268AbjFNUjF (ORCPT + 99 others); Wed, 14 Jun 2023 16:39:05 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:55752 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S235026AbjFNUix (ORCPT ); Wed, 14 Jun 2023 16:38:53 -0400 Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 092A42686; Wed, 14 Jun 2023 13:38:50 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1686775130; x=1718311130; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=6d/g1iJxdHdtB8pORIwpYQRecfV8E82PURJyPjd0TyI=; b=L8COK8vk0qr5qFSoCz9l4GP7Ir4N5w8axAeUzqvI4RTSrqUr+GGsvoPs x9d/jyhgnY96/P8tdrLV7O+Kf7zTGjbt6sMU9JFyEq8ysu+2cOeNJQKI8 T+6oBhKJQjeDHdCMhzwk70xtjDtARKXki+FPuxwcr24GU0yOkL/vajauq 8Va2Q3dw7eguD4pYpsa2n+E6PUIk34GdNfEzxxw5VF/8DIL47HNQmcNFu rDvI+fU/ZBdcnSYnkU+jThPD0wWwIAFpiRF1g7QBLDuhUt9pEsmdvznqi +OLgw8s95OEH4/j3CYB1VeVh0Wp02C4rQOtPHpKwqqGYrgWCF5IjWvNzh w==; X-IronPort-AV: E=McAfee;i="6600,9927,10741"; a="361209026" X-IronPort-AV: E=Sophos;i="6.00,243,1681196400"; d="scan'208";a="361209026" Received: from orsmga008.jf.intel.com ([10.7.209.65]) by orsmga102.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 14 Jun 2023 13:38:49 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6600,9927,10741"; a="741963190" X-IronPort-AV: E=Sophos;i="6.00,243,1681196400"; d="scan'208";a="741963190" Received: from a0cec87da3f2.jf.intel.com (HELO worker-node-1.jf.intel.com) ([10.165.55.163]) by orsmga008.jf.intel.com with ESMTP; 14 Jun 2023 13:38:49 -0700 From: Weilin Wang To: Peter Zijlstra , Ingo Molnar , Arnaldo Carvalho de Melo , Jiri Olsa , Namhyung Kim , Adrian Hunter , Ian Rogers , linux-perf-users@vger.kernel.org, linux-kernel@vger.kernel.org Cc: Weilin Wang , Kan Liang , Samantha Alt , Perry Taylor , Caleb Biggers Subject: [PATCH v3 1/3] perf test: Add metric value validation test Date: Wed, 14 Jun 2023 13:38:22 -0700 Message-Id: <20230614203824.2895333-2-weilin.wang@intel.com> X-Mailer: git-send-email 2.39.1 In-Reply-To: <20230614203824.2895333-1-weilin.wang@intel.com> References: <20230614203824.2895333-1-weilin.wang@intel.com> MIME-Version: 1.0 X-Spam-Status: No, score=-4.4 required=5.0 tests=BAYES_00,DKIMWL_WL_HIGH, DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,RCVD_IN_DNSWL_MED, RCVD_IN_MSPIKE_H3,RCVD_IN_MSPIKE_WL,SPF_HELO_NONE,SPF_NONE, T_SCC_BODY_TEXT_LINE,URIBL_BLOCKED autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lindbergh.monkeyblade.net Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org X-getmail-retrieved-from-mailbox: =?utf-8?q?INBOX?= X-GMAIL-THRID: =?utf-8?q?1768714474120326743?= X-GMAIL-MSGID: =?utf-8?q?1768714474120326743?= Add metric value validation test to check if metric values are with in correct value ranges. There are three types of tests included: 1) positive-value test checks if all the metrics collected are non-negative; 2) single-value test checks if the list of metrics have values in given value ranges; 3) relationship test checks if multiple metrics follow a given relationship, e.g. memory_bandwidth_read + memory_bandwidth_write = memory_bandwidth_total. Signed-off-by: Weilin Wang --- .../tests/shell/lib/perf_metric_validation.py | 514 ++++++++++++++++++ .../lib/perf_metric_validation_rules.json | 387 +++++++++++++ tools/perf/tests/shell/stat_metrics_values.sh | 30 + 3 files changed, 931 insertions(+) create mode 100644 tools/perf/tests/shell/lib/perf_metric_validation.py create mode 100644 tools/perf/tests/shell/lib/perf_metric_validation_rules.json create mode 100755 tools/perf/tests/shell/stat_metrics_values.sh diff --git a/tools/perf/tests/shell/lib/perf_metric_validation.py b/tools/perf/tests/shell/lib/perf_metric_validation.py new file mode 100644 index 000000000000..81bd2bf38b67 --- /dev/null +++ b/tools/perf/tests/shell/lib/perf_metric_validation.py @@ -0,0 +1,514 @@ +#SPDX-License-Identifier: GPL-2.0 +import re +import csv +import json +import argparse +from pathlib import Path +import subprocess + +class Validator: + def __init__(self, rulefname, reportfname='', t=5, debug=False, datafname='', fullrulefname='', workload='true', metrics=''): + self.rulefname = rulefname + self.reportfname = reportfname + self.rules = None + self.collectlist=metrics + self.metrics = set() + self.tolerance = t + + self.workloads = [x for x in workload.split(",") if x] + self.wlidx = 0 # idx of current workloads + self.allresults = dict() # metric results of all workload + self.allignoremetrics = dict() # metrics with no results or negative results + self.allfailtests = dict() + self.alltotalcnt = dict() + self.allpassedcnt = dict() + self.allerrlist = dict() + + self.results = dict() # metric results of current workload + # vars for test pass/failure statistics + self.ignoremetrics= set() # metrics with no results or negative results, neg result counts as a failed test + self.failtests = dict() + self.totalcnt = 0 + self.passedcnt = 0 + # vars for errors + self.errlist = list() + + # vars for Rule Generator + self.pctgmetrics = set() # Percentage rule + + # vars for debug + self.datafname = datafname + self.debug = debug + self.fullrulefname = fullrulefname + + def read_json(self, filename: str) -> dict: + try: + with open(Path(filename).resolve(), "r") as f: + data = json.loads(f.read()) + except OSError as e: + print(f"Error when reading file {e}") + sys.exit() + + return data + + def json_dump(self, data, output_file): + parent = Path(output_file).parent + if not parent.exists(): + parent.mkdir(parents=True) + + with open(output_file, "w+") as output_file: + json.dump(data, + output_file, + ensure_ascii=True, + indent=4) + + def get_results(self, idx:int = 0): + return self.results[idx] + + def get_bounds(self, lb, ub, error, alias={}, ridx:int = 0) -> list: + """ + Get bounds and tolerance from lb, ub, and error. + If missing lb, use 0.0; missing ub, use float('inf); missing error, use self.tolerance. + + @param lb: str/float, lower bound + @param ub: str/float, upper bound + @param error: float/str, error tolerance + @returns: lower bound, return inf if the lower bound is a metric value and is not collected + upper bound, return -1 if the upper bound is a metric value and is not collected + tolerance, denormalized base on upper bound value + """ + # init ubv and lbv to invalid values + def get_bound_value (bound, initval, ridx): + val = initval + if isinstance(bound, int) or isinstance(bound, float): + val = bound + elif isinstance(bound, str): + if bound == '': + val = float("inf") + elif bound in alias: + vall = self.get_value(alias[ub], ridx) + if vall: + val = vall[0] + elif bound.replace('.', '1').isdigit(): + val = float(bound) + else: + print("Wrong bound: {0}".format(bound)) + else: + print("Wrong bound: {0}".format(bound)) + return val + + ubv = get_bound_value(ub, -1, ridx) + lbv = get_bound_value(lb, float('inf'), ridx) + t = get_bound_value(error, self.tolerance, ridx) + + # denormalize error threshold + denormerr = t * ubv / 100 if ubv != 100 and ubv > 0 else t + + return lbv, ubv, denormerr + + def get_value(self, name:str, ridx:int = 0) -> list: + """ + Get value of the metric from self.results. + If result of this metric is not provided, the metric name will be added into self.ignoremetics and self.errlist. + All future test(s) on this metric will fail. + + @param name: name of the metric + @returns: list with value found in self.results; list is empty when not value found. + """ + results = [] + data = self.results[ridx] if ridx in self.results else self.results[0] + if name not in self.ignoremetrics: + if name in data: + results.append(data[name]) + elif name.replace('.', '1').isdigit(): + results.append(float(name)) + else: + self.errlist.append("Metric '%s' is not collected or the value format is incorrect"%(name)) + self.ignoremetrics.add(name) + return results + + def check_bound(self, val, lb, ub, err): + return True if val <= ub + err and val >= lb - err else False + + # Positive Value Sanity check + def pos_val_test(self): + """ + Check if metrics value are non-negative. + One metric is counted as one test. + Failure: when metric value is negative or not provided. + Metrics with negative value will be added into the self.failtests['PositiveValueTest'] and self.ignoremetrics. + """ + negmetric = set() + missmetric = set() + pcnt = 0 + tcnt = 0 + for name, val in self.get_results().items(): + if val is None or val == '': + missmetric.add(name) + self.errlist.append("Metric '%s' is not collected"%(name)) + elif val < 0: + negmetric.add("{0}(={1:.4f})".format(name, val)) + else: + pcnt += 1 + tcnt += 1 + + self.failtests['PositiveValueTest']['Total Tests'] = tcnt + self.failtests['PositiveValueTest']['Passed Tests'] = pcnt + if len(negmetric) or len(missmetric)> 0: + self.ignoremetrics.update(negmetric) + self.ignoremetrics.update(missmetric) + self.failtests['PositiveValueTest']['Failed Tests'].append({'NegativeValue':list(negmetric), 'MissingValue':list(missmetric)}) + + return + + def evaluate_formula(self, formula:str, alias:dict, ridx:int = 0): + """ + Evaluate the value of formula. + + @param formula: the formula to be evaluated + @param alias: the dict has alias to metric name mapping + @returns: value of the formula is success; -1 if the one or more metric value not provided + """ + stack = [] + b = 0 + errs = [] + sign = "+" + f = str() + + #TODO: support parenthesis? + for i in range(len(formula)): + if i+1 == len(formula) or formula[i] in ('+', '-', '*', '/'): + s = alias[formula[b:i]] if i+1 < len(formula) else alias[formula[b:]] + v = self.get_value(s, ridx) + if not v: + errs.append(s) + else: + f = f + "{0}(={1:.4f})".format(s, v[0]) + if sign == "*": + stack[-1] = stack[-1] * v + elif sign == "/": + stack[-1] = stack[-1] / v + elif sign == '-': + stack.append(-v[0]) + else: + stack.append(v[0]) + if i + 1 < len(formula): + sign = formula[i] + f += sign + b = i + 1 + + if len(errs) > 0: + return -1, "Metric value missing: "+','.join(errs) + + val = sum(stack) + return val, f + + # Relationships Tests + def relationship_test(self, rule: dict): + """ + Validate if the metrics follow the required relationship in the rule. + eg. lower_bound <= eval(formula)<= upper_bound + One rule is counted as ont test. + Failure: when one or more metric result(s) not provided, or when formula evaluated outside of upper/lower bounds. + + @param rule: dict with metric name(+alias), formula, and required upper and lower bounds. + """ + alias = dict() + for m in rule['Metrics']: + alias[m['Alias']] = m['Name'] + lbv, ubv, t = self.get_bounds(rule['RangeLower'], rule['RangeUpper'], rule['ErrorThreshold'], alias, ridx=rule['RuleIndex']) + val, f = self.evaluate_formula(rule['Formula'], alias, ridx=rule['RuleIndex']) + if val == -1: + self.failtests['RelationshipTest']['Failed Tests'].append({'RuleIndex': rule['RuleIndex'], 'Description':f}) + elif not self.check_bound(val, lbv, ubv, t): + lb = rule['RangeLower'] + ub = rule['RangeUpper'] + if isinstance(lb, str): + if lb in alias: + lb = alias[lb] + if isinstance(ub, str): + if ub in alias: + ub = alias[ub] + self.failtests['RelationshipTest']['Failed Tests'].append({'RuleIndex': rule['RuleIndex'], 'Formula':f, + 'RangeLower': lb, 'LowerBoundValue': self.get_value(lb), + 'RangeUpper': ub, 'UpperBoundValue':self.get_value(ub), + 'ErrorThreshold': t, 'CollectedValue': val}) + else: + self.passedcnt += 1 + self.failtests['RelationshipTest']['Passed Tests'] += 1 + self.totalcnt += 1 + self.failtests['RelationshipTest']['Total Tests'] += 1 + + return + + + # Single Metric Test + def single_test(self, rule:dict): + """ + Validate if the metrics are in the required value range. + eg. lower_bound <= metrics_value <= upper_bound + One metric is counted as one test in this type of test. + One rule may include one or more metrics. + Failure: when the metric value not provided or the value is outside the bounds. + This test updates self.total_cnt and records failed tests in self.failtest['SingleMetricTest']. + + @param rule: dict with metrics to validate and the value range requirement + """ + lbv, ubv, t = self.get_bounds(rule['RangeLower'], rule['RangeUpper'], rule['ErrorThreshold']) + metrics = rule['Metrics'] + passcnt = 0 + totalcnt = 0 + faillist = [] + for m in metrics: + totalcnt += 1 + result = self.get_value(m['Name']) + if len(result) > 0 and self.check_bound(result[0], lbv, ubv, t): + passcnt += 1 + else: + faillist.append({'MetricName':m['Name'], 'CollectedValue':result}) + + self.totalcnt += totalcnt + self.passedcnt += passcnt + self.failtests['SingleMetricTest']['Total Tests'] += totalcnt + self.failtests['SingleMetricTest']['Passed Tests'] += passcnt + if len(faillist) != 0: + self.failtests['SingleMetricTest']['Failed Tests'].append({'RuleIndex':rule['RuleIndex'], + 'RangeLower': rule['RangeLower'], + 'RangeUpper': rule['RangeUpper'], + 'ErrorThreshold':rule['ErrorThreshold'], + 'Failure':faillist}) + + return + + def create_report(self): + """ + Create final report and write into a JSON file. + """ + alldata = list() + for i in range(0, len(self.workloads)): + reportstas = {"Total Rule Count": self.alltotalcnt[i], "Passed Rule Count": self.allpassedcnt[i]} + data = {"Metric Validation Statistics": reportstas, "Tests in Category": self.allfailtests[i], + "Errors":self.allerrlist[i]} + alldata.append({"Workload": self.workloads[i], "Report": data}) + + json_str = json.dumps(alldata, indent=4) + print("Test validation finished. Final report: ") + print(json_str) + + if self.debug: + allres = [{"Workload": self.workloads[i], "Results": self.allresults[i]} for i in range(0, len(self.workloads))] + self.json_dump(allres, self.datafname) + + def check_rule(self, testtype, metric_list): + """ + Check if the rule uses metric(s) that not exist in current platform. + + @param metric_list: list of metrics from the rule. + @return: False when find one metric out in Metric file. (This rule should not skipped.) + True when all metrics used in the rule are found in Metric file. + """ + if testtype == "RelationshipTest": + for m in metric_list: + if m['Name'] not in self.metrics: + return False + return True + + # Start of Collector and Converter + def convert(self, data: list, idx: int): + """ + Convert collected metric data from the -j output to dict of {metric_name:value}. + """ + for json_string in data: + try: + result =json.loads(json_string) + if "metric-unit" in result and result["metric-unit"] != "(null)" and result["metric-unit"] != "": + name = result["metric-unit"].split(" ")[1] if len(result["metric-unit"].split(" ")) > 1 \ + else result["metric-unit"] + if idx not in self.results: self.results[idx] = dict() + self.results[idx][name.lower()] = float(result["metric-value"]) + except ValueError as error: + continue + return + + def collect_perf(self, data_file: str, workload: str): + """ + Collect metric data with "perf stat -M" on given workload with -a and -j. + """ + self.results = dict() + tool = 'perf' + print(f"Starting perf collection") + print(f"Workload: {workload}") + collectlist = dict() + if self.collectlist != "": + collectlist[0] = {x for x in self.collectlist.split(",")} + else: + collectlist[0] = set(list(self.metrics)) + # Create metric set for relationship rules + for rule in self.rules: + if rule["TestType"] == "RelationshipTest": + metrics = [m["Name"] for m in rule["Metrics"]] + if not any(m not in collectlist[0] for m in metrics): + collectlist[rule["RuleIndex"]] = set(metrics) + + for idx, metrics in collectlist.items(): + if idx == 0: wl = "sleep 0.5".split() + else: wl = workload.split() + for metric in metrics: + command = [tool, 'stat', '-j', '-M', f"{metric}", "-a"] + command.extend(wl) + cmd = subprocess.run(command, stderr=subprocess.PIPE, encoding='utf-8') + data = [x+'}' for x in cmd.stderr.split('}\n') if x] + self.convert(data, idx) + # End of Collector and Converter + + # Start of Rule Generator + def parse_perf_metrics(self): + """ + Read and parse perf metric file: + 1) find metrics with '1%' or '100%' as ScaleUnit for Percent check + 2) create metric name list + """ + command = ['perf', 'list', '-j', '--details', 'metrics'] + cmd = subprocess.run(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, encoding='utf-8') + try: + data = json.loads(cmd.stdout) + for m in data: + if 'MetricName' not in m: + print("Warning: no metric name") + continue + name = m['MetricName'] + self.metrics.add(name) + if 'ScaleUnit' in m and (m['ScaleUnit'] == '1%' or m['ScaleUnit'] == '100%'): + self.pctgmetrics.add(name.lower()) + except ValueError as error: + print(f"Error when parsing metric data") + sys.exit() + + return + + def create_rules(self): + """ + Create full rules which includes: + 1) All the rules from the "relationshi_rules" file + 2) SingleMetric rule for all the 'percent' metrics + + Reindex all the rules to avoid repeated RuleIndex + """ + self.rules = self.read_json(self.rulefname)['RelationshipRules'] + pctgrule = {'RuleIndex':0, + 'TestType':'SingleMetricTest', + 'RangeLower':'0', + 'RangeUpper': '100', + 'ErrorThreshold': self.tolerance, + 'Description':'Metrics in percent unit have value with in [0, 100]', + 'Metrics': [{'Name': m} for m in self.pctgmetrics]} + self.rules.append(pctgrule) + + # Re-index all rules to avoid repeated RuleIndex + idx = 1 + for r in self.rules: + r['RuleIndex'] = idx + idx += 1 + + if self.debug: + #TODO: need to test and generate file name correctly + data = {'RelationshipRules':self.rules, 'SupportedMetrics': [{"MetricName": name} for name in self.metrics]} + self.json_dump(data, self.fullrulefname) + + return + # End of Rule Generator + + def _storewldata(self, key): + ''' + Store all the data of one workload into the corresponding data structure for all workloads. + @param key: key to the dictionaries (index of self.workloads). + ''' + self.allresults[key] = self.results + self.allignoremetrics[key] = self.ignoremetrics + self.allfailtests[key] = self.failtests + self.alltotalcnt[key] = self.totalcnt + self.allpassedcnt[key] = self.passedcnt + self.allerrlist[key] = self.errlist + + #Initialize data structures before data validation of each workload + def _init_data(self): + + testtypes = ['PositiveValueTest', 'RelationshipTest', 'SingleMetricTest'] + self.results = dict() + self.ignoremetrics= set() + self.errlist = list() + self.failtests = {k:{'Total Tests':0, 'Passed Tests':0, 'Failed Tests':[]} for k in testtypes} + self.totalcnt = 0 + self.passedcnt = 0 + + def test(self): + ''' + The real entry point of the test framework. + This function loads the validation rule JSON file and Standard Metric file to create rules for + testing and namemap dictionaries. + It also reads in result JSON file for testing. + + In the test process, it passes through each rule and launch correct test function bases on the + 'TestType' field of the rule. + + The final report is written into a JSON file. + ''' + self.parse_perf_metrics() + self.create_rules() + for i in range(0, len(self.workloads)): + self._init_data() + self.collect_perf(self.datafname, self.workloads[i]) + # Run positive value test + self.pos_val_test() + for r in self.rules: + # skip rules that uses metrics not exist in this platform + testtype = r['TestType'] + if not self.check_rule(testtype, r['Metrics']): + continue + if testtype == 'RelationshipTest': + self.relationship_test(r) + elif testtype == 'SingleMetricTest': + self.single_test(r) + else: + print("Unsupported Test Type: ", testtype) + self.errlist.append("Unsupported Test Type from rule: " + r['RuleIndex']) + self._storewldata(i) + print("Workload: ", self.workloads[i]) + print("Total metrics collected: ", self.failtests['PositiveValueTest']['Total Tests']) + print("Non-negative metric count: ", self.failtests['PositiveValueTest']['Passed Tests']) + print("Total Test Count: ", self.totalcnt) + print("Passed Test Count: ", self.passedcnt) + + self.create_report() + return sum(self.alltotalcnt.values()) != sum(self.allpassedcnt.values()) +# End of Class Validator + + +def main() -> None: + parser = argparse.ArgumentParser(description="Launch metric value validation") + + parser.add_argument("-rule", help="Base validation rule file", required=True) + parser.add_argument("-output_dir", help="Path for validator output file, report file", required=True) + parser.add_argument("-debug", help="Debug run, save intermediate data to files", action="store_true", default=False) + parser.add_argument("-wl", help="Workload to run while data collection", default="true") + parser.add_argument("-m", help="Metric list to validate", default="") + args = parser.parse_args() + outpath = Path(args.output_dir) + reportf = Path.joinpath(outpath, 'perf_report.json') + fullrule = Path.joinpath(outpath, 'full_rule.json') + datafile = Path.joinpath(outpath, 'perf_data.json') + + validator = Validator(args.rule, reportf, debug=args.debug, + datafname=datafile, fullrulefname=fullrule, workload=args.wl, + metrics=args.m) + ret = validator.test() + + return ret + + +if __name__ == "__main__": + import sys + sys.exit(main()) + + + diff --git a/tools/perf/tests/shell/lib/perf_metric_validation_rules.json b/tools/perf/tests/shell/lib/perf_metric_validation_rules.json new file mode 100644 index 000000000000..debaa910da9f --- /dev/null +++ b/tools/perf/tests/shell/lib/perf_metric_validation_rules.json @@ -0,0 +1,387 @@ +{ + "RelationshipRules": [ + { + "RuleIndex": 1, + "Formula": "a+b", + "TestType": "RelationshipTest", + "RangeLower": "c", + "RangeUpper": "c", + "ErrorThreshold": 5.0, + "Description": "Intel(R) Optane(TM) Persistent Memory(PMEM) bandwidth total includes Intel(R) Optane(TM) Persistent Memory(PMEM) read bandwidth and Intel(R) Optane(TM) Persistent Memory(PMEM) write bandwidth", + "Metrics": [ + { + "Name": "pmem_memory_bandwidth_read", + "Alias": "a" + }, + { + "Name": "pmem_memory_bandwidth_write", + "Alias": "b" + }, + { + "Name": "pmem_memory_bandwidth_total", + "Alias": "c" + } + ] + }, + { + "RuleIndex": 2, + "Formula": "a+b", + "TestType": "RelationshipTest", + "RangeLower": "c", + "RangeUpper": "c", + "ErrorThreshold": 5.0, + "Description": "DDR memory bandwidth total includes DDR memory read bandwidth and DDR memory write bandwidth", + "Metrics": [ + { + "Name": "memory_bandwidth_read", + "Alias": "a" + }, + { + "Name": "memory_bandwidth_write", + "Alias": "b" + }, + { + "Name": "memory_bandwidth_total", + "Alias": "c" + } + ] + }, + { + "RuleIndex": 3, + "Formula": "a+b", + "TestType": "RelationshipTest", + "RangeLower": "100", + "RangeUpper": "100", + "ErrorThreshold": 5.0, + "Description": "Total memory read accesses includes memory reads from last level cache (LLC) addressed to local DRAM and memory reads from the last level cache (LLC) addressed to remote DRAM.", + "Metrics": [ + { + "Name": "numa_reads_addressed_to_local_dram", + "Alias": "a" + }, + { + "Name": "numa_reads_addressed_to_remote_dram", + "Alias": "b" + } + ] + }, + { + "RuleIndex": 4, + "Formula": "a", + "TestType": "SingleMetricTest", + "RangeLower": "0.125", + "RangeUpper": "", + "ErrorThreshold": "", + "Description": "", + "Metrics": [ + { + "Name": "cpi", + "Alias": "a" + } + ] + }, + { + "RuleIndex": 5, + "Formula": "", + "TestType": "SingleMetricTest", + "RangeLower": "0", + "RangeUpper": "1", + "ErrorThreshold": 5.0, + "Description": "Ratio values should be within value range [0,1)", + "Metrics": [ + { + "Name": "loads_per_instr", + "Alias": "" + }, + { + "Name": "stores_per_instr", + "Alias": "" + }, + { + "Name": "l1d_mpi", + "Alias": "" + }, + { + "Name": "l1d_demand_data_read_hits_per_instr", + "Alias": "" + }, + { + "Name": "l1_i_code_read_misses_with_prefetches_per_instr", + "Alias": "" + }, + { + "Name": "l2_demand_data_read_hits_per_instr", + "Alias": "" + }, + { + "Name": "l2_mpi", + "Alias": "" + }, + { + "Name": "l2_demand_data_read_mpi", + "Alias": "" + }, + { + "Name": "l2_demand_code_mpi", + "Alias": "" + } + ] + }, + { + "RuleIndex": 6, + "Formula": "a+b+c+d", + "TestType": "RelationshipTest", + "RangeLower": "100", + "RangeUpper": "100", + "ErrorThreshold": 5.0, + "Description": "Sum of TMA level 1 metrics should be 100%", + "Metrics": [ + { + "Name": "tma_frontend_bound", + "Alias": "a" + }, + { + "Name": "tma_bad_speculation", + "Alias": "b" + }, + { + "Name": "tma_backend_bound", + "Alias": "c" + }, + { + "Name": "tma_retiring", + "Alias": "d" + } + ] + }, + { + "RuleIndex": 7, + "Formula": "a+b", + "TestType": "RelationshipTest", + "RangeLower": "c", + "RangeUpper": "c", + "ErrorThreshold": 5.0, + "Description": "Sum of the level 2 children should equal level 1 parent", + "Metrics": [ + { + "Name": "tma_fetch_latency", + "Alias": "a" + }, + { + "Name": "tma_fetch_bandwidth", + "Alias": "b" + }, + { + "Name": "tma_frontend_bound", + "Alias": "c" + } + ] + }, + { + "RuleIndex": 8, + "Formula": "a+b", + "TestType": "RelationshipTest", + "RangeLower": "c", + "RangeUpper": "c", + "ErrorThreshold": 5.0, + "Description": "Sum of the level 2 children should equal level 1 parent", + "Metrics": [ + { + "Name": "tma_branch_mispredicts", + "Alias": "a" + }, + { + "Name": "tma_machine_clears", + "Alias": "b" + }, + { + "Name": "tma_bad_speculation", + "Alias": "c" + } + ] + }, + { + "RuleIndex": 9, + "Formula": "a+b", + "TestType": "RelationshipTest", + "RangeLower": "c", + "RangeUpper": "c", + "ErrorThreshold": 5.0, + "Description": "Sum of the level 2 children should equal level 1 parent", + "Metrics": [ + { + "Name": "tma_memory_bound", + "Alias": "a" + }, + { + "Name": "tma_core_bound", + "Alias": "b" + }, + { + "Name": "tma_backend_bound", + "Alias": "c" + } + ] + }, + { + "RuleIndex": 10, + "Formula": "a+b", + "TestType": "RelationshipTest", + "RangeLower": "c", + "RangeUpper": "c", + "ErrorThreshold": 5.0, + "Description": "Sum of the level 2 children should equal level 1 parent", + "Metrics": [ + { + "Name": "tma_light_operations", + "Alias": "a" + }, + { + "Name": "tma_heavy_operations", + "Alias": "b" + }, + { + "Name": "tma_retiring", + "Alias": "c" + } + ] + }, + { + "RuleIndex": 11, + "Formula": "a+b+c", + "TestType": "RelationshipTest", + "RangeLower": "100", + "RangeUpper": "100", + "ErrorThreshold": 5.0, + "Description": "The all_requests includes the memory_page_empty, memory_page_misses, and memory_page_hits equals.", + "Metrics": [ + { + "Name": "memory_page_empty_vs_all_requests", + "Alias": "a" + }, + { + "Name": "memory_page_misses_vs_all_requests", + "Alias": "b" + }, + { + "Name": "memory_page_hits_vs_all_requests", + "Alias": "c" + } + ] + }, + { + "RuleIndex": 12, + "Formula": "a-b", + "TestType": "RelationshipTest", + "RangeLower": "0", + "RangeUpper": "", + "ErrorThreshold": 5.0, + "Description": "CPU utilization in kernel mode should always be <= cpu utilization", + "Metrics": [ + { + "Name": "cpu_utilization", + "Alias": "a" + }, + { + "Name": "cpu_utilization_in_kernel_mode", + "Alias": "b" + } + ] + }, + { + "RuleIndex": 13, + "Formula": "a-b", + "TestType": "RelationshipTest", + "RangeLower": "0", + "RangeUpper": "", + "ErrorThreshold": 5.0, + "Description": "Total L2 misses per instruction should be >= L2 demand data read misses per instruction", + "Metrics": [ + { + "Name": "l2_mpi", + "Alias": "a" + }, + { + "Name": "l2_demand_data_read_mpi", + "Alias": "b" + } + ] + }, + { + "RuleIndex": 14, + "Formula": "a-b", + "TestType": "RelationshipTest", + "RangeLower": "0", + "RangeUpper": "", + "ErrorThreshold": 5.0, + "Description": "Total L2 misses per instruction should be >= L2 demand code misses per instruction", + "Metrics": [ + { + "Name": "l2_mpi", + "Alias": "a" + }, + { + "Name": "l2_demand_code_mpi", + "Alias": "b" + } + ] + }, + { + "RuleIndex": 15, + "Formula": "b+c+d", + "TestType": "RelationshipTest", + "RangeLower": "a", + "RangeUpper": "a", + "ErrorThreshold": 5.0, + "Description": "L3 data read, rfo, code misses per instruction equals total L3 misses per instruction.", + "Metrics": [ + { + "Name": "llc_mpi", + "Alias": "a" + }, + { + "Name": "llc_data_read_mpi_demand_plus_prefetch", + "Alias": "b" + }, + { + "Name": "llc_rfo_read_mpi_demand_plus_prefetch", + "Alias": "c" + }, + { + "Name": "llc_code_read_mpi_demand_plus_prefetch", + "Alias": "d" + } + ] + }, + { + "RuleIndex": 16, + "Formula": "a", + "TestType": "SingleMetricTest", + "RangeLower": "0", + "RangeUpper": "8", + "ErrorThreshold": 0.0, + "Description": "Setting generous range for allowable frequencies", + "Metrics": [ + { + "Name": "uncore_freq", + "Alias": "a" + } + ] + }, + { + "RuleIndex": 17, + "Formula": "a", + "TestType": "SingleMetricTest", + "RangeLower": "0", + "RangeUpper": "8", + "ErrorThreshold": 0.0, + "Description": "Setting generous range for allowable frequencies", + "Metrics": [ + { + "Name": "cpu_operating_frequency", + "Alias": "a" + } + ] + } + ] +} \ No newline at end of file diff --git a/tools/perf/tests/shell/stat_metrics_values.sh b/tools/perf/tests/shell/stat_metrics_values.sh new file mode 100755 index 000000000000..65a15c65eea7 --- /dev/null +++ b/tools/perf/tests/shell/stat_metrics_values.sh @@ -0,0 +1,30 @@ +#!/bin/bash +# perf metrics value validation +# SPDX-License-Identifier: GPL-2.0 +if [ "x$PYTHON" == "x" ] +then + if which python3 > /dev/null + then + PYTHON=python3 + else + echo Skipping test, python3 not detected please set environment variable PYTHON. + exit 2 + fi +fi + +grep -q Intel /proc/cpuinfo || (echo Skipping non-Intel; exit 2) + +pythonvalidator=$(dirname $0)/lib/perf_metric_validation.py +rulefile=$(dirname $0)/lib/perf_metric_validation_rules.json +tmpdir=$(mktemp -d /tmp/__perf_test.program.XXXXX) +workload="perf bench futex hash -r 2 -s" + +# Add -debug, save data file and full rule file +echo "Launch python validation script $pythonvalidator" +echo "Output will be stored in: $tmpdir" +$PYTHON $pythonvalidator -rule $rulefile -output_dir $tmpdir -wl "${workload}" +ret=$? +rm -rf $tmpdir + +exit $ret +