# Copyright 2013-2018 ARM Limited # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # # pylint: disable=E1101 import os import re import tempfile import json from collections import defaultdict from wa import ApkUiautoWorkload, Parameter from wa.framework.exception import ConfigError, WorkloadError from wa.utils.misc import capitalize from wa.utils.types import version_tuple class Geekbench(ApkUiautoWorkload): name = 'geekbench' description = """ Geekbench provides a comprehensive set of benchmarks engineered to quickly and accurately measure processor and memory performance. http://www.primatelabs.com/geekbench/ From the website: Designed to make benchmarks easy to run and easy to understand, Geekbench takes the guesswork out of producing robust and reliable benchmark results. Geekbench scores are calibrated against a baseline score of 1,000 (which is the score of a single-processor Power Mac G5 @ 1.6GHz). Higher scores are better, with double the score indicating double the performance. The benchmarks fall into one of four categories: - integer performance. - floating point performance. - memory performance. - stream performance. Geekbench benchmarks: http://www.primatelabs.com/geekbench/doc/benchmarks.html Geekbench scoring methedology: http://support.primatelabs.com/kb/geekbench/interpreting-geekbench-scores """ summary_metrics = ['score', 'multicore_score'] supported_versions = ['4.3.2', '4.3.1', '4.2.0', '4.0.1', '3.4.1', '3.0.0', '2'] package_names = ['com.primatelabs.geekbench', 'com.primatelabs.geekbench3', 'ca.primatelabs.geekbench2'] begin_regex = re.compile(r'^\s*D/WebViewClassic.loadDataWithBaseURL\(\s*\d+\s*\)' r'\s*:\s*(?P\<.*)\s*$') replace_regex = re.compile(r'<[^>]*>') parameters = [ Parameter('version', allowed_values=supported_versions, description='Specifies which version of the workload should be run.', override=True), Parameter('loops', kind=int, default=1, aliases=['times'], description=('Specfies the number of times the benchmark will be run in a "tight ' 'loop", i.e. without performaing setup/teardown inbetween.')), Parameter('timeout', kind=int, default=3600, description=('Timeout for a single iteration of the benchmark. This value is ' 'multiplied by ``times`` to calculate the overall run timeout. ')), Parameter('disable_update_result', kind=bool, default=False, description=('If ``True`` the results file will not be pulled from the targets ' '``/data/data/com.primatelabs.geekbench`` folder. This allows the ' 'workload to be run on unrooted targets and the results extracted ' 'manually later.')), ] is_corporate = False phones_home = True requires_network = True def initialize(self, context): super(Geekbench, self).initialize(context) self.gui.uiauto_params['version'] = self.version self.gui.uiauto_params['loops'] = self.loops self.gui.uiauto_params['is_corporate'] = self.is_corporate self.gui.timeout = self.timeout if not self.disable_update_result and not self.target.is_rooted: raise WorkloadError( 'Geekbench workload requires root to collect results. ' 'You can set disable_update_result=True in the workload params ' 'to run without collecting results.') def setup(self, context): super(Geekbench, self).setup(context) self.run_timeout = self.timeout * self.loops def update_output(self, context): super(Geekbench, self).update_output(context) if not self.disable_update_result: major_version = version_tuple(self.version)[0] update_method = getattr(self, 'update_result_{}'.format(major_version)) update_method(context) def validate(self): if (self.loops > 1) and (self.version == '2'): raise ConfigError('loops parameter is not supported for version 2 of Geekbench.') def update_result_2(self, context): score_calculator = GBScoreCalculator() score_calculator.parse(self.logcat_log) score_calculator.update_results(context) def update_result_3(self, context): outfile_glob = self.target.path.join(self.target.package_data_directory, self.apk.package, 'files', '*gb3') on_target_output_files = [f.strip() for f in self.target.execute('ls {}'.format(outfile_glob), as_root=True).split('\n') if f] for i, on_target_output_file in enumerate(on_target_output_files): host_temp_file = tempfile.mktemp() self.target.pull(on_target_output_file, host_temp_file, as_root=True) host_output_file = os.path.join(context.output_directory, os.path.basename(on_target_output_file)) with open(host_temp_file) as fh: data = json.load(fh) os.remove(host_temp_file) with open(host_output_file, 'w') as wfh: json.dump(data, wfh, indent=4) context.add_artifact('geekout', host_output_file, kind='data', description='Geekbench 3 output from target.') context.add_metric(namemify('score', i), data['score']) context.add_metric(namemify('multicore_score', i), data['multicore_score']) for section in data['sections']: context.add_metric(namemify(section['name'] + '_score', i), section['score']) context.add_metric(namemify(section['name'] + '_multicore_score', i), section['multicore_score']) def update_result_4(self, context): outfile_glob = self.target.path.join(self.target.package_data_directory, self.apk.package, 'files', '*gb*') on_target_output_files = [f.strip() for f in self.target.execute('ls {}'.format(outfile_glob), as_root=True).split('\n') if f] for i, on_target_output_file in enumerate(on_target_output_files): host_temp_file = tempfile.mktemp() self.target.pull(on_target_output_file, host_temp_file, as_root=True) host_output_file = os.path.join(context.output_directory, os.path.basename(on_target_output_file)) with open(host_temp_file) as fh: data = json.load(fh) os.remove(host_temp_file) with open(host_output_file, 'w') as wfh: json.dump(data, wfh, indent=4) context.add_artifact('geekout', host_output_file, kind='data', description='Geekbench 4 output from target.') context.add_metric(namemify('score', i), data['score']) context.add_metric(namemify('multicore_score', i), data['multicore_score']) for section in data['sections']: context.add_metric(namemify(section['name'] + '_score', i), section['score']) for workloads in section['workloads']: workload_name = workloads['name'].replace(" ", "-") context.add_metric(namemify(section['name'] + '_' + workload_name + '_score', i), workloads['score']) update_result_5 = update_result_4 class GBWorkload(object): """ Geekbench workload (not to be confused with WA's workloads). This is a single test run by geek bench, such as preforming compression or generating Madelbrot. """ # Index maps onto the hundreds digit of the ID. categories = [None, 'integer', 'float', 'memory', 'stream'] # 2003 entry-level Power Mac G5 is considered to have a baseline score of # 1000 for every category. pmac_g5_base_score = 1000 units_conversion_map = { 'K': 1, 'M': 1000, 'G': 1000000, } def __init__(self, wlid, name, pmac_g5_st_score, pmac_g5_mt_score): """ :param wlid: A three-digit workload ID. Uniquely identifies a workload and also determines the category a workload belongs to. :param name: The name of the workload. :param pmac_g5_st_score: Score achieved for this workload on 2003 entry-level Power Mac G5 running in a single thread. :param pmac_g5_mt_score: Score achieved for this workload on 2003 entry-level Power Mac G5 running in multiple threads. """ self.wlid = wlid self.name = name self.pmac_g5_st_score = pmac_g5_st_score self.pmac_g5_mt_score = pmac_g5_mt_score self.category = self.categories[int(wlid) // 100] self.collected_results = [] def add_result(self, value, units): self.collected_results.append(self.convert_to_kilo(value, units)) def convert_to_kilo(self, value, units): return value * self.units_conversion_map[units[0]] def clear(self): self.collected_results = [] def get_scores(self): """ Returns a tuple (single-thraded score, multi-threaded score) for this workload. Some workloads only have a single-threaded score, in which case multi-threaded score will be ``None``. Geekbench will perform four iterations of each workload in single-threaded and, for some workloads, multi-threaded configurations. Thus there should always be either four or eight scores collected for each workload. Single-threaded iterations are always done before multi-threaded, so the ordering of the scores can be used to determine which configuration they belong to. This method should not be called before score collection has finished. """ no_of_results = len(self.collected_results) if no_of_results == 4: return (self._calculate(self.collected_results[:4], self.pmac_g5_st_score), None) if no_of_results == 8: return (self._calculate(self.collected_results[:4], self.pmac_g5_st_score), self._calculate(self.collected_results[4:], self.pmac_g5_mt_score)) else: msg = 'Collected {} results for Geekbench {} workload;'.format(no_of_results, self.name) msg += ' expecting either 4 or 8.' raise WorkloadError(msg) def _calculate(self, values, scale_factor): return max(values) * self.pmac_g5_base_score / scale_factor def __str__(self): return self.name __repr__ = __str__ class GBScoreCalculator(object): """ Parses logcat output to extract raw Geekbench workload values and converts them into category and overall scores. """ result_regex = re.compile(r'workload (?P\d+) (?P[0-9.]+) ' r'(?P[a-zA-Z/]+) (?P