Changes
On November 28, 2024 at 1:17:04 PM UTC, admin:
-
Changed value of field
orcid
to0000-0002-3600-6899
in Performance-detective: automatic deduction of cheap and accurate performance models - supplementary material
f | 1 | { | f | 1 | { |
2 | "author": "Schmid, Larissa", | 2 | "author": "Schmid, Larissa", | ||
3 | "author_email": "", | 3 | "author_email": "", | ||
n | n | 4 | "citation": [], | ||
4 | "creator_user_id": "17755db4-395a-4b3b-ac09-e8e3484ca700", | 5 | "creator_user_id": "17755db4-395a-4b3b-ac09-e8e3484ca700", | ||
5 | "doi": "10.35097/1330", | 6 | "doi": "10.35097/1330", | ||
6 | "doi_date_published": "2023", | 7 | "doi_date_published": "2023", | ||
7 | "doi_publisher": "", | 8 | "doi_publisher": "", | ||
8 | "doi_status": "True", | 9 | "doi_status": "True", | ||
n | n | 10 | "familyName": "Schmid", | ||
11 | "givenName": "Larissa", | ||||
9 | "groups": [], | 12 | "groups": [], | ||
10 | "id": "37206c44-f3ad-4cec-b12b-332bd4c98a6c", | 13 | "id": "37206c44-f3ad-4cec-b12b-332bd4c98a6c", | ||
11 | "isopen": false, | 14 | "isopen": false, | ||
12 | "license_id": "CC BY 4.0 Attribution", | 15 | "license_id": "CC BY 4.0 Attribution", | ||
13 | "license_title": "CC BY 4.0 Attribution", | 16 | "license_title": "CC BY 4.0 Attribution", | ||
14 | "metadata_created": "2023-08-04T08:50:33.219568", | 17 | "metadata_created": "2023-08-04T08:50:33.219568", | ||
n | 15 | "metadata_modified": "2023-08-04T09:29:04.998842", | n | 18 | "metadata_modified": "2024-11-28T13:17:04.435204", |
16 | "name": "rdr-doi-10-35097-1330", | 19 | "name": "rdr-doi-10-35097-1330", | ||
17 | "notes": "TechnicalRemarks: Supplementary material for | 20 | "notes": "TechnicalRemarks: Supplementary material for | ||
18 | \"Performance-Detective: Automatic Deduction of Cheap and Accurate | 21 | \"Performance-Detective: Automatic Deduction of Cheap and Accurate | ||
19 | Performance Models\" (DOI 10.1145/3524059.3532391)\r\n\r\n# Step 1: | 22 | Performance Models\" (DOI 10.1145/3524059.3532391)\r\n\r\n# Step 1: | ||
20 | System analysis\r\n\r\nWe provide the processed JSON output of | 23 | System analysis\r\n\r\nWe provide the processed JSON output of | ||
21 | Perf-Taint for the Pace3D and Kripke case studies, as well as the | 24 | Perf-Taint for the Pace3D and Kripke case studies, as well as the | ||
22 | bitcode of Kripke used as input of Perf-Taint. Because Pace3D is | 25 | bitcode of Kripke used as input of Perf-Taint. Because Pace3D is | ||
23 | closed source, we cannot provide the source code to reproduce the | 26 | closed source, we cannot provide the source code to reproduce the | ||
24 | analysis results. \r\n\r\n# Step 2: Experiment design\r\n\r\nWe | 27 | analysis results. \r\n\r\n# Step 2: Experiment design\r\n\r\nWe | ||
25 | provide scripts that: \r\n* check the output of Perf-Taint for | 28 | provide scripts that: \r\n* check the output of Perf-Taint for | ||
26 | parameters not interacting with each other,\r\n* detect iteration | 29 | parameters not interacting with each other,\r\n* detect iteration | ||
27 | parameters potentially linearly influencing the runtime of the | 30 | parameters potentially linearly influencing the runtime of the | ||
28 | calculation\r\n* check the coefficient of variation of single | 31 | calculation\r\n* check the coefficient of variation of single | ||
29 | iterations to confirm the linear influence\r\n\r\n# Step 3: | 32 | iterations to confirm the linear influence\r\n\r\n# Step 3: | ||
30 | Instrumented experiments\r\n\r\n\r\nWe provide the profiles of all | 33 | Instrumented experiments\r\n\r\n\r\nWe provide the profiles of all | ||
31 | measurements conducted for the training points. \r\n\r\nFor Pace3D, we | 34 | measurements conducted for the training points. \r\n\r\nFor Pace3D, we | ||
32 | had to add one function to the instrumentation filter manually because | 35 | had to add one function to the instrumentation filter manually because | ||
33 | of a bug in Score-P: Functions declared as static inline inside a | 36 | of a bug in Score-P: Functions declared as static inline inside a | ||
34 | header file which are then called from different translation units are | 37 | header file which are then called from different translation units are | ||
35 | not instrumented correctly by Score-P. Thus, the resulting profiles | 38 | not instrumented correctly by Score-P. Thus, the resulting profiles | ||
36 | are incorrect. However, we can work around this bug by including the | 39 | are incorrect. However, we can work around this bug by including the | ||
37 | function calling the static inline function. \r\n\r\n## Training Data | 40 | function calling the static inline function. \r\n\r\n## Training Data | ||
38 | Pace3D\r\n\r\n### Training Data | 41 | Pace3D\r\n\r\n### Training Data | ||
39 | Performance-Detective\r\n\r\nPerformance-Detective derived a minimal | 42 | Performance-Detective\r\n\r\nPerformance-Detective derived a minimal | ||
40 | experiment design with 25 configurations to be measured once, | 43 | experiment design with 25 configurations to be measured once, | ||
41 | resulting in 25 measurements. \r\n\r\nThe data for Extra-P is splitted | 44 | resulting in 25 measurements. \r\n\r\nThe data for Extra-P is splitted | ||
42 | into two folders that contain the same measurements that are labeled | 45 | into two folders that contain the same measurements that are labeled | ||
43 | differently. As we did not implement modeling for the optimized | 46 | differently. As we did not implement modeling for the optimized | ||
44 | subset, we later create two models (one using procs and vol, one using | 47 | subset, we later create two models (one using procs and vol, one using | ||
45 | procs and cubes) and extract the model of each function from the | 48 | procs and cubes) and extract the model of each function from the | ||
46 | respective model (depending on whether the functions relies on vol OR | 49 | respective model (depending on whether the functions relies on vol OR | ||
47 | on cubes). \r\n\r\nThe data for PIM is the same as in the Extra-P | 50 | on cubes). \r\n\r\nThe data for PIM is the same as in the Extra-P | ||
48 | folder, just labeled with all the values measured.\r\n\r\n### Training | 51 | folder, just labeled with all the values measured.\r\n\r\n### Training | ||
49 | Data Full-Factorial\r\n\r\n625 measurements for all possible | 52 | Data Full-Factorial\r\n\r\n625 measurements for all possible | ||
50 | combinations of the five values of procs, vol, and cubes. Each | 53 | combinations of the five values of procs, vol, and cubes. Each | ||
51 | measurement was repeated 5 times. \r\n\r\n### Training Data | 54 | measurement was repeated 5 times. \r\n\r\n### Training Data | ||
52 | Plackett-Burman\r\n\r\n49 samples selected using a random seed and 5 | 55 | Plackett-Burman\r\n\r\n49 samples selected using a random seed and 5 | ||
53 | levels -- effectively a subset of the full-factorial measurements. | 56 | levels -- effectively a subset of the full-factorial measurements. | ||
54 | Each measurement was repeated 5 times.\r\n\r\n\r\n\r\n## Training Data | 57 | Each measurement was repeated 5 times.\r\n\r\n\r\n\r\n## Training Data | ||
55 | Kripke\r\n\r\n### Training Data | 58 | Kripke\r\n\r\n### Training Data | ||
56 | Performance-Detective\r\n\r\nPerformance-Detective derived a minimal | 59 | Performance-Detective\r\n\r\nPerformance-Detective derived a minimal | ||
57 | experiment design with 5 configurations to be measured once, resulting | 60 | experiment design with 5 configurations to be measured once, resulting | ||
58 | in 5 measurements. \r\n\r\nThe data for Extra-P is splitted into two | 61 | in 5 measurements. \r\n\r\nThe data for Extra-P is splitted into two | ||
59 | folders that contain the same measurements that are labeled | 62 | folders that contain the same measurements that are labeled | ||
60 | differently. As we did not implement modeling for the optimized | 63 | differently. As we did not implement modeling for the optimized | ||
61 | subset, we later create two models (one for procs, one for dirsets) | 64 | subset, we later create two models (one for procs, one for dirsets) | ||
62 | and extract the model of each function from the respective model | 65 | and extract the model of each function from the respective model | ||
63 | (depending on whether the functions relies on procs OR on dirsets). | 66 | (depending on whether the functions relies on procs OR on dirsets). | ||
64 | However, for the single-parameter modeler, modeling based on | 67 | However, for the single-parameter modeler, modeling based on | ||
65 | dependencies from Perf-Taint is not yet implemented. Therefore, the | 68 | dependencies from Perf-Taint is not yet implemented. Therefore, the | ||
66 | measurements are labeled as two parameter-experiments following the | 69 | measurements are labeled as two parameter-experiments following the | ||
67 | policy detailed in the respective folder. \r\n\r\nThe data for PIM is | 70 | policy detailed in the respective folder. \r\n\r\nThe data for PIM is | ||
68 | the same as in the Extra-P folder, just labeled with all the values | 71 | the same as in the Extra-P folder, just labeled with all the values | ||
69 | measured.\r\n\r\n### Training Data Full-Factorial\r\n\r\n125 | 72 | measured.\r\n\r\n### Training Data Full-Factorial\r\n\r\n125 | ||
70 | measurements for all possible combinations of the five values of procs | 73 | measurements for all possible combinations of the five values of procs | ||
71 | and dirsets. Each measurements was repeated 5 times. \r\n\r\n### | 74 | and dirsets. Each measurements was repeated 5 times. \r\n\r\n### | ||
72 | Training Data Plackett-Burman\r\n\r\n10 samples selected using a | 75 | Training Data Plackett-Burman\r\n\r\n10 samples selected using a | ||
73 | random seed and 5 levels -- effectively a subset of the full-factorial | 76 | random seed and 5 levels -- effectively a subset of the full-factorial | ||
74 | measurements. Each measurement was repeated 5 times. \r\n\r\n\r\n\r\n# | 77 | measurements. Each measurement was repeated 5 times. \r\n\r\n\r\n\r\n# | ||
75 | Step 4: Modeling and Evaluation\r\n\r\nData of each case study is in | 78 | Step 4: Modeling and Evaluation\r\n\r\nData of each case study is in | ||
76 | the respective folder. \r\n\r\n## Evaluation | 79 | the respective folder. \r\n\r\n## Evaluation | ||
77 | Measurements\r\n\r\nExtra- and interpolated measurement data in the | 80 | Measurements\r\n\r\nExtra- and interpolated measurement data in the | ||
78 | respective folder. \r\n\r\n## Extra-P\r\n\r\nContains script for | 81 | respective folder. \r\n\r\n## Extra-P\r\n\r\nContains script for | ||
79 | calculating the model errors regarding extra- and interpolated | 82 | calculating the model errors regarding extra- and interpolated | ||
80 | evaluation measurements. \r\n\r\n### Models\r\n\r\nContains Extra-P | 83 | evaluation measurements. \r\n\r\n### Models\r\n\r\nContains Extra-P | ||
81 | models as well as a converter from Extra-P format to text (creates | 84 | models as well as a converter from Extra-P format to text (creates | ||
82 | PerformanceModel.py), that is used in `calculate_model_errors.py`. It | 85 | PerformanceModel.py), that is used in `calculate_model_errors.py`. It | ||
83 | requires an installation of Extra-P to run. \r\n\r\n\r\n## | 86 | requires an installation of Extra-P to run. \r\n\r\n\r\n## | ||
84 | Performance-Influence Models\r\n\r\nWe use the scripts by [Weber et | 87 | Performance-Influence Models\r\n\r\nWe use the scripts by [Weber et | ||
85 | mance-Influence-Models/tree/main/supplementary-website/code/modeling), | 88 | mance-Influence-Models/tree/main/supplementary-website/code/modeling), | ||
86 | partly modified to account for the configurations of the case studies | 89 | partly modified to account for the configurations of the case studies | ||
87 | and modeling based on known dependencies to functions.\r\n\r\n### | 90 | and modeling based on known dependencies to functions.\r\n\r\n### | ||
88 | Data\r\n\r\nContains the data of the cubex files in csv format. Data | 91 | Data\r\n\r\nContains the data of the cubex files in csv format. Data | ||
89 | is parsed with `create_csv_from_cubex.py`. \r\n\r\n### Model | 92 | is parsed with `create_csv_from_cubex.py`. \r\n\r\n### Model | ||
90 | errors\r\n\r\nContains the csv files with the model errors of the | 93 | errors\r\n\r\nContains the csv files with the model errors of the | ||
91 | respective models. Files are generated by | 94 | respective models. Files are generated by | ||
92 | `learn_method_level_model_with_deps.py` and | 95 | `learn_method_level_model_with_deps.py` and | ||
93 | `learn_method_level_without_deps.py`, respectively.\r\n\r\n# | 96 | `learn_method_level_without_deps.py`, respectively.\r\n\r\n# | ||
94 | License\r\n \r\nThe files `calculate_model_errors.py`, | 97 | License\r\n \r\nThe files `calculate_model_errors.py`, | ||
95 | `extrap_to_text.py`, and `create_csv_from_cubex.py` are modified parts | 98 | `extrap_to_text.py`, and `create_csv_from_cubex.py` are modified parts | ||
96 | of the Extra-P Software (cf. | 99 | of the Extra-P Software (cf. | ||
97 | hub.com/extra-p/extrap/blob/master/extrap/fileio/cube_file_reader2.py) | 100 | hub.com/extra-p/extrap/blob/master/extrap/fileio/cube_file_reader2.py) | ||
98 | and the license file `LICENSE-BSD3-EXTRAP`). \r\n\r\nThe files | 101 | and the license file `LICENSE-BSD3-EXTRAP`). \r\n\r\nThe files | ||
99 | `learn_method_level_model_with_deps.py` and | 102 | `learn_method_level_model_with_deps.py` and | ||
100 | `learn_method_level_model_without_deps.py` are modified parts of the | 103 | `learn_method_level_model_without_deps.py` are modified parts of the | ||
101 | supplementary material of [Weber et | 104 | supplementary material of [Weber et | ||
102 | mance-Influence-Models/tree/main/supplementary-website/code/modeling). | 105 | mance-Influence-Models/tree/main/supplementary-website/code/modeling). | ||
103 | The license file `LICENSE-GPL-PIM` applies to these files.", | 106 | The license file `LICENSE-GPL-PIM` applies to these files.", | ||
104 | "num_resources": 0, | 107 | "num_resources": 0, | ||
105 | "num_tags": 0, | 108 | "num_tags": 0, | ||
n | 106 | "orcid": "", | n | 109 | "orcid": "0000-0002-3600-6899", |
107 | "organization": { | 110 | "organization": { | ||
108 | "approval_status": "approved", | 111 | "approval_status": "approved", | ||
109 | "created": "2023-01-12T13:30:23.238233", | 112 | "created": "2023-01-12T13:30:23.238233", | ||
110 | "description": "RADAR (Research Data Repository) is a | 113 | "description": "RADAR (Research Data Repository) is a | ||
111 | cross-disciplinary repository for archiving and publishing research | 114 | cross-disciplinary repository for archiving and publishing research | ||
112 | data from completed scientific studies and projects. The focus is on | 115 | data from completed scientific studies and projects. The focus is on | ||
113 | research data from subjects that do not yet have their own | 116 | research data from subjects that do not yet have their own | ||
114 | discipline-specific infrastructures for research data management. ", | 117 | discipline-specific infrastructures for research data management. ", | ||
115 | "id": "013c89a9-383c-4200-8baa-0f78bf1d91f9", | 118 | "id": "013c89a9-383c-4200-8baa-0f78bf1d91f9", | ||
116 | "image_url": "radar-logo.svg", | 119 | "image_url": "radar-logo.svg", | ||
117 | "is_organization": true, | 120 | "is_organization": true, | ||
118 | "name": "radar", | 121 | "name": "radar", | ||
119 | "state": "active", | 122 | "state": "active", | ||
120 | "title": "RADAR", | 123 | "title": "RADAR", | ||
121 | "type": "organization" | 124 | "type": "organization" | ||
122 | }, | 125 | }, | ||
123 | "owner_org": "013c89a9-383c-4200-8baa-0f78bf1d91f9", | 126 | "owner_org": "013c89a9-383c-4200-8baa-0f78bf1d91f9", | ||
124 | "private": false, | 127 | "private": false, | ||
125 | "production_year": "2022", | 128 | "production_year": "2022", | ||
126 | "publication_year": "2023", | 129 | "publication_year": "2023", | ||
127 | "publishers": [ | 130 | "publishers": [ | ||
128 | { | 131 | { | ||
129 | "publisher": "Karlsruhe Institute of Technology" | 132 | "publisher": "Karlsruhe Institute of Technology" | ||
130 | } | 133 | } | ||
131 | ], | 134 | ], | ||
t | t | 135 | "related_identifiers": [ | ||
136 | { | ||||
137 | "identifier": | ||||
138 | "https://publikationen.bibliothek.kit.edu/1000146001", | ||||
139 | "identifier_type": "URL", | ||||
140 | "relation_type": "IsIdenticalTo" | ||||
141 | } | ||||
142 | ], | ||||
132 | "relationships_as_object": [], | 143 | "relationships_as_object": [], | ||
133 | "relationships_as_subject": [], | 144 | "relationships_as_subject": [], | ||
134 | "repository_name": "RADAR (Research Data Repository)", | 145 | "repository_name": "RADAR (Research Data Repository)", | ||
135 | "resources": [], | 146 | "resources": [], | ||
136 | "services_used_list": "", | 147 | "services_used_list": "", | ||
137 | "source_metadata_created": "2023", | 148 | "source_metadata_created": "2023", | ||
138 | "source_metadata_modified": "", | 149 | "source_metadata_modified": "", | ||
139 | "state": "active", | 150 | "state": "active", | ||
140 | "subject_areas": [ | 151 | "subject_areas": [ | ||
141 | { | 152 | { | ||
142 | "subject_area_additional": "", | 153 | "subject_area_additional": "", | ||
143 | "subject_area_name": "Computer Science" | 154 | "subject_area_name": "Computer Science" | ||
144 | } | 155 | } | ||
145 | ], | 156 | ], | ||
146 | "tags": [], | 157 | "tags": [], | ||
147 | "title": "Performance-detective: automatic deduction of cheap and | 158 | "title": "Performance-detective: automatic deduction of cheap and | ||
148 | accurate performance models - supplementary material", | 159 | accurate performance models - supplementary material", | ||
149 | "type": "vdataset", | 160 | "type": "vdataset", | ||
150 | "url": "https://doi.org/10.35097/1330" | 161 | "url": "https://doi.org/10.35097/1330" | ||
151 | } | 162 | } |