-
Notifications
You must be signed in to change notification settings - Fork 1
/
accelergywrapper.py
296 lines (252 loc) · 10.6 KB
/
accelergywrapper.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
from accelergy.plug_in_interface.interface import (
AccelergyPlugIn,
Estimation,
AccuracyEstimation,
AccelergyQuery,
)
from accelergy.plug_in_interface.estimator_wrapper import (
SupportedComponent,
PrintableCall,
)
from typing import Dict, List, Tuple, Union
import os
import sys
# fmt: off for Black formatter
SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
sys.path.append(SCRIPT_DIR)
from scaling import *
from helper_functions import *
# fmt: on
AREA_ACCURACY = 90
ENERGY_ACCURACY = 90
# =============================================================================
# Wrapper Class
# =============================================================================
class LibraryEstimator(AccelergyPlugIn):
def __init__(self):
self.estimator_name = "Library"
super().__init__()
self.components = []
component_files = [
os.path.join(root, f)
for root, _, files in os.walk(os.path.join(SCRIPT_DIR, "library"))
for f in files
]
for k, v in os.environ.items():
if "ACCELERGY_COMPONENT_LIBRARIES" not in k:
continue
for path in v.split(","):
component_files += [
os.path.join(root, f)
for root, _, files in os.walk(path)
for f in files
]
self._load_component_files(component_files)
self._load_reference_files(component_files)
self.logger.info(f"Loaded {len(self.components)} components from library.")
for c in self.components:
c.setdefault("n_instances", 1)
self.action2entry = {}
self.name2entry = {}
for c in self.components:
for action in c["action"].split("|"):
name = c["name"].lower().strip()
action = action.lower().strip()
entry = {**c, **{"name": name, "action": action}}
self.action2entry.setdefault(
(entry["name"], entry["action"]), []
).append(entry)
self.name2entry.setdefault(entry["name"], []).append(entry)
# Make sure all components have a read, write, update, and leak action
for name in self.name2entry:
for action in ["read", "write", "update", "leak"]:
assert (
name,
action,
) in self.action2entry, (
f"Missing {action} action for Library component {name}."
)
def _load_component_lines(self, name: str, lines: List[str]):
"""Loads the component lines into self.components"""
keys = [k.strip() for k in lines[0].split(",")]
last_nonempty = 0
for i, k in enumerate(keys):
if k:
last_nonempty = i
keys = keys[: last_nonempty + 1]
for l in lines[1:]:
if l:
values = [v.strip() for v in l.split(",")]
self.components.append(dict(zip(keys, values)))
self.components[-1]["name"] = name.lower()
def _load_component_files(self, files: List[str]):
"""Loads the component files into self.components"""
component_files = [f for f in files if f.endswith(".csv")]
for f in component_files:
lines = [l.split("#")[0].strip() for l in open(f).readlines()]
lines = [l for l in lines if l.replace(",", "")]
curname = os.path.basename(f).split(".")[0]
curlines = []
for i, l in enumerate(lines):
if "COMPONENT" in lines[i]:
if curlines:
self._load_component_lines(curname, curlines)
curlines = []
curname = l.split(":")[1].strip()
else:
curlines.append(l)
if curlines:
self._load_component_lines(curname, curlines)
def _load_reference_files(self, files: List[str]):
"""Loads the reference files into self.components"""
references = {}
reference_files = [f for f in files if f.endswith("_pointers.txt")]
# Load references
for ref in reference_files:
lines = open(ref).readlines()
for l in lines:
k, v = l.split(":", maxsplit=1)
references[k.strip().lower()] = v.strip().lower()
# Find references in components
for k, v in references.items():
found = False
for c in self.components:
if c["name"] == v:
self.components.append({**c, **{"name": k}})
found = True
if not found:
raise ValueError(
f"Reference {k}->{v} not found. Known components:\n\t"
+ "\n\t".join(c["name"] for c in self.components)
)
def primitive_action_supported(self, query: AccelergyQuery) -> AccuracyEstimation:
success = self.get_energy_or_area(query, log_scaling=False) is not None
return AccuracyEstimation(ENERGY_ACCURACY if success else 0)
def primitive_area_supported(self, query: AccelergyQuery) -> AccuracyEstimation:
success = self.estimate_area(query, log_scaling=False) is not None
return AccuracyEstimation(AREA_ACCURACY if success else 0)
def estimate_energy(self, query: AccelergyQuery, **kwargs) -> Estimation:
return self.get_energy_or_area(query, True, **kwargs)
def estimate_area(self, query: AccelergyQuery, **kwargs) -> Estimation:
return self.get_energy_or_area(query, False, **kwargs)
def get_name(self) -> str:
return self.estimator_name
def match_entry(
self,
query: AccelergyQuery,
entry: Dict[str, str],
target: str,
log_scaling: bool,
) -> Tuple[Union[float, None], int, List[str]]:
"""Matches a query to an entry in the library. Returns the energy/area
scale, the number of matching attributes, and a log."""
class_name = query.class_name.lower()
class_attrs = query.class_attrs
scale, log = 1, []
self.logger.info(f'Checking entry "{entry}')
# Check if we match the attributes. Find those that must be scaled
matching_attrs, attrs_to_scale = [], []
def entrykey(a):
for k in entry:
if str(a).lower() == str(k).lower():
return k
if str(a).lower() in str(k).lower().split("|"):
return k
return None
class2entry = {}
for a in class_attrs:
if entrykey(a) not in class2entry.values():
class2entry[a] = entrykey(a)
else:
class2entry[a] = None
for a in class_attrs:
entry_val = entry.get(class2entry[a], None)
if entry_val is None:
pass
elif (
str(entry_val) == "*"
or str(entry_val).lower() == str(class_attrs[a]).lower()
):
matching_attrs.append(a)
elif not class_attrs.get(f"no_scale_{target}", False):
log.append(f"Scaling {a} from {entry_val} to {class_attrs[a]}")
attrs_to_scale.append(a)
# Scale the attributes that must be scaled
try:
# Try to scale the attributes that must be scaled
for a in attrs_to_scale:
scalefrom = parse_float(entry[class2entry[a]], f"{class_name}.{a}")
scaleto = parse_float(class_attrs[a], f"{class_name}.{a}")
s = scale_energy_or_area(a, scalefrom, scaleto, target)
if s == 1:
matching_attrs.append(a)
scale *= s
if scale and log_scaling:
log.append(
f"Scaled {class_name}.{a} from {scalefrom} to "
f"{scaleto}: {s}x {target}"
)
except (ValueError, ZeroDivisionError) as e:
scale = None
self.logger.info(f"Failed to scale {class_name}: {str(e).strip()}")
return scale, len(matching_attrs), log
def get_energy_or_area(
self,
query: AccelergyQuery,
is_energy: bool = True,
log_scaling: bool = True,
) -> Estimation:
class_name = query.class_name.lower()
# For finding closest-matching component
best_value, best_matches, best_log, best_entry = None, -1, [], {}
target = "energy" if is_energy else "area"
get_value = target
if query.action_name == "leak":
target = "leak"
if is_energy:
action_name = query.action_name.lower()
entries = self.action2entry.get((class_name, action_name), [])
self.logger.info(
f"Found {len(entries)} entries for {class_name}.{action_name}."
)
else:
entries = self.name2entry.get(class_name, [])
self.logger.info(f"Found {len(entries)} entries for {class_name}.")
for entry in entries:
scale, matching_attrs, log = self.match_entry(
query, entry, target, log_scaling
)
if scale is None:
continue
# Scaled successfully! Now get the value
if log_scaling:
log.append(f"{class_name} {target} has been scaled {scale}x")
value = get_value_from_entry(entry, get_value)
self.logger.info(f"{value=}, {matching_attrs=}, {log=}")
if value is not None and matching_attrs > best_matches:
best_value = value * scale
best_matches = matching_attrs
best_log = log
best_entry = entry
if log_scaling:
self.logger.info(f"Best-matching entry: {best_entry}")
for l in best_log:
self.logger.info(l)
if best_value is None:
raise ValueError(f"Could not find {target} for {class_name}")
return Estimation(best_value, "p" if is_energy else "u^2")
def get_supported_components(self) -> List[SupportedComponent]:
supported = []
for c in self.components:
class_names = c["name"].split("|")
c_popped = {k: v for k, v in c.items() if k not in ["name", "action"]}
supported.append(
SupportedComponent(
class_names,
PrintableCall("", [], c_popped),
[PrintableCall(a) for a in c["action"].split("|")],
)
)
return supported
if __name__ == "__main__":
s = LibraryEstimator()